site banner

Culture War Roundup for the week of April 17, 2023

This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.

Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.

We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:

  • Shaming.

  • Attempting to 'build consensus' or enforce ideological conformity.

  • Making sweeping generalizations to vilify a group you dislike.

  • Recruiting for a cause.

  • Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.

In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:

  • Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.

  • Be as precise and charitable as you can. Don't paraphrase unflatteringly.

  • Don't imply that someone said something they did not say, even if you think it follows from what they said.

  • Write like everyone is reading and you want them to be included in the discussion.

On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.

8
Jump in the discussion.

No email address required.

Is the rapid advancement in Machine Learning good or bad for society?

For the purposes of this comment, I will try to define good as "improving the quality of life for many people without decreasing the quality of life for another similarly sized group" an vice versa.

I enjoy trying to answer this question because the political discourse around it is too new to have widely accepted answers disseminated by the two American political parties being used to signify affiliation like many questions. However, any discussion of whether something is good or bad for society belongs in a Culture War threat because, even here on The Motte, most people will try to reduce every discussion to one along clear conservative/liberal lines because most people here are salty conservatives who were kicked out of reddit by liberals one way or another.

Now on to the question: Maybe the best way to discover if Machine learning is good or bad for society is to say what makes it essentially different from previous computing? The key difference in Machine Learning is that it changes computing from a process where you tell the computer what to do with data, and turns it into a process where you just tell the computer what you want it to be able to do. before machine learning, you would tell the computer specifically how to scan an image and decide if it is a picture of a dog. Whether the computer was good at identifying pictures of dogs relied on how good your instructions were. With machine learning, you give the computer millions of pictures of dogs and tell it to figure out how to determine if there's a dog in a picture.

So what can be essentialized from that difference? Well before Machine Learning, the owners of the biggest computers still had to be clever enough to use them to manipulate data properly, but with Machine Learning, the owners of the biggest computers can now simply specify a goal and get what they want. It seems therefore that Machine Learning will work as a tool for those with more capital to find ways to gain more capital. It will allow people with the money to create companies that can enhance the ability to make decisions purely based on profit potential, and remove the human element even more from the equation.

How about a few examples:

Recently a machine learning model was approved by the FDA to be used to identify cavities on X-rays. Eventually your dental insurance company will require a machine learning model to read your X-rays and report that you need a procedure in order for them to cover treatment from your dentist. The justification will be that the Machine Learning model is more accurate. It probably will be more accurate. Dentists will require subscriptions to a Machine Learning model to accept insurance, and perhaps dental treatment will become more expensive, but maybe not. It's hard to say for sure if this will be a bad or a good thing.

Machine learning models are getting very good at writing human text. This is currently reducing the value of human writers at a quick pace. Presumably with more advanced models, it will replace commercial human writing all together. Every current limitation of the leading natural language models will be removed in time, and they will become objectively superior to human writers. This also might be a good thing, or a bad thing. It's hard to say.

I think it's actually very hard to predict if Machine Learning will be good or bad for society. Certain industries might be disrupted, but the long term effects are hard to predict.

I believe there are two broad scenarios for what might happen with ML from an economics/politics perspective.

Scenario 1 is that ML will be a powerful productive tool (ie capital) in the hands of those that can afford it just like many other inventions throughout history.

If this happens the reaction will be along the lines we all know too well. The left will complain that those in power gain even more power and now have novel ways to control and/or extract value from workers. Plus a lot more low-skilled people will become unemployable and redundant so class tensions will probably get worse. On the flip side a few smart early movers will make insane bank and shape the way the next few decades will go. Could be interesting to see how different nation states adopt the new technology.

Scenario 2 is the "things get crazy" scenario. What if ML takes off far quicker than people are expecting, for example by recursively improving itself? I believe in that case we might be unable to fit the development into our usual political lens. If one company has twice as much capital than everyone else combined our systems of power distribution fall apart. If one nation has capabilities that make it effectively invincible our models for foreign relations stop working. If that happens it will be more akin to a scenario where superintelligent aliens have landed on earth and all bets are off.

What if ML takes off far quicker than people are expecting, for example by recursively improving itself? I believe in that case we might be unable to fit the development into our usual political lens. If one company has twice as much capital than everyone else combined our systems of power distribution fall apart. If one nation has capabilities that make it effectively invincible our models for foreign relations stop working.

I'm a little disconcerted at how many people who are working in the industry seem to hold this as the explicit goal and are intentionally maneuvering things so as to prevent anyone from intervening until it's too late.

I expect that "recursively improving itself" will lead to the AI going off into the weeds -- that is, evolving in ways unconnected to the real world. The output will quickly become bizarre and not particularly useful. It works for formal systems like Go because the rules are well-defined, but you can't simulate reality to a sufficient degree of precision.

I think the idea behind recursive self-improvement is more like, a 150 IQ AI should be able to find a way to increase its IQ to 151, a 151 IQ AI should be able to increase its IQ to 152, and so on and so forth until it reaches godhood.

It doesn't necessarily have to simulate large portions of reality, if it's able to find a way to isolate the factors responsible for its g factor and come up with a generalized way of making improvements to those factors. Presumably as part of the cycles of improvement it could interact with the real world in order to get more training and data. But this sort of scenario has its own issues.

Especially if it can spin up various copies of itself and make minute changes to see how that effects performance. Basically massive, parallel experimentation.

I am strongly of the opinion that since neoliberal PMC jobs are the easiest to automatic with AI, there will be incredibly strong regulation banning AI from taking the jobs of the PMC. The power to regulate is the power to destroy, and as incapable of actual productivity the PMC and their legion of bullshit jobs are, they know how to run a grift and bask in their own self importance.

No, what you need to fear from AI is when Facebook fires up an instance of AutoGPT for each user and tasks it with keeping them doom scrolling for as long as is possible. If you thought "the algorithm" was already amoral and sanity shredding, you ain't seen nothing yet. That was a mere baby, feebly hand tuned by meat that thinks (or thinks it thinks). When the AI is fully unleashed on slaving our attention spans to our screens, it's going to be like how Fentanyl turbo charged opioid deaths. You're gonna start seeing people literally starving to death staring at their phones. Actually, nix that, they'll die of dehydration first. I momentarily forgot that nearly always happens first.

I'm gonna register this prediction now too. Apparently Ai has trouble with fingers. You'll know it's gotten loose when there is a new tiktok trend of young people amputating all their fingers. The AI will have decided it's easier to convince us to get rid of our own fingers than figure out how to draw them better. Given the rates of Tiktok induced mental illness, it would probably be right in that assessment.

I am strongly of the opinion that since neoliberal PMC jobs are the easiest to automatic with AI, there will be incredibly strong regulation banning AI from taking the jobs of the PMC. The power to regulate is the power to destroy, and as incapable of actual productivity the PMC and their legion of bullshit jobs are, they know how to run a grift and bask in their own self importance.

This is exactly why the crossbow and handgonnes never took off and why we still live under a feudal system ruled over by our lieges and ladies.

More seriously, this technology is too valuable to not use, anyone who does use it is going to gain a massive advantage over anyone that doesn't, its use is inevitable.

More seriously, this technology is too valuable to not use, anyone who does use it is going to gain a massive advantage over anyone that doesn't, its use is inevitable.

The same is true of nuclear power. It's the only technology that will allow us to hit emission targets and keep the grid stable with cheap, reliable power.

But we've built 3 nuclear power plants in as many decades, and our infrastructure is crumbling and less reliable than ever. Our ruling class simply does not care so long as they can keep living that 0.01% life. Even now they are setting preposterous 10 year EV targets, despite not putting a dime towards building out a domestic EV supply chain or infrastructure. Including upgrading our electric grid to deal with the massive increase in demand all those EVs will create. Which brings us back to the nuclear power they scorn so much.

Your appeals to a reasonable nation performing certain obvious reasonable tasks are pointless. This is clown world. You need to think dumber.

Nuclear power has a lot of benefits, but it takes a significant amount of time and money to get online, with the benefits being generally diffused. The number of organisations that can actually get a nuclear power plant online for long enough that they can start to make a profit is quite small.

AI is comparatively cheap, the changes are quick and easily observable and the pay off for an individual willing to utilise it is substantial. As a class medievial European nobility may have benefited from a complete ban on crossbows and handguns, but the ratio of costs to return of employing these weapons meant that anyone who chose to defect and take up their use would out compete those who did not. The same is true of AI, it cannot be ignored.

Your appeals to a reasonable nation performing certain obvious reasonable tasks are pointless. This is clown world. You need to think dumber.

I'm appealing to human greed and desire for power. You need to think smarter.

The same is true of nuclear power. It's the only technology that will allow us to hit emission targets and keep the grid stable with cheap, reliable power.

Exactly. The general population believes what it was told for 50 years - nuclear power is something immensely dangerous and deadly, something that can explode at any moment, kill millions and turn the whole country into uninhabitable desert full of motocycle riding mutants.

Now, imagine if normies are told:

THE COMPUTER can kill you. Yes, THE COMPUTER can shred you into paperclips, without warning. And not only you, but everyone, everyone in the whole world. Yes, even ordinary computer in your son's room can do it.

Do not wait for your doom. Say something, do something.

The general population believes what it was told for 50 years - nuclear power is something immensely dangerous and deadly, something that can explode at any moment, kill millions and turn the whole country into uninhabitable desert full of motocycle riding mutants.

Again, at least according to this poll, 76% of Americans - the most relevant demographic for this forum - favor nuclear energy. Even the opponents do not necessarily hold the most alarmist and charged view of nuclear as a power source.

The American public won't give up guns, do you think they'll give up computers?

Heck, even if it's just AIs they're told to give up, forces that want to do that will have to move fast, because every passing moment it reaches more hands, and the hands that have it are gonna hold on tight. And at some point soon, we will reach a point of cultural no return on everyone having these tools.

The problem is, we've already had hacker scares for years, I don't know what it would really take for people to realize the threat outside of re-hashed Terminator references.

I'm gonna register this prediction now too. Apparently Ai has trouble with fingers. You'll know it's gotten loose when there is a new tiktok trend of young people amputating all their fingers. The AI will have decided it's easier to convince us to get rid of our own fingers than figure out how to draw them better. Given the rates of Tiktok induced mental illness, it would probably be right in that assessment.

This would be a rad short story. An AI that gets 'frustrated' at its own limitations against the real world and it's solution is to just sand off all the sharp edges that are giving it problems.

Like it genetically engineers all the cows to be spherical so it's physics simulations can be more accurate.

An AI that gets 'frustrated' at its own limitations against the real world and it's solution is to just sand off all the sharp edges that are giving it problems.

I'm obligated to point out that this already happened, the AI was capitalism, the sharp edges were all direct human interactions, and our atomized broken society is the result.

I would be interested in seeing this thought/analogy expanded.

Seeing Like a State plus a broad view of what constitutes a "state," perhaps?

I thought I had seen later Scottposts applying this logic to capitalism.

His Meditations on Moloch sounds like this vein too.

I thought I got this idea from Mark Fisher or Nick Land, but random googling isn't leading me to any obvious writing of theirs on this specific concept. Come to think of it maybe it was one of IlForte's pithier comments. Regardless you should read both of them.

I am strongly of the opinion that since neoliberal PMC jobs are the easiest to automatic with AI, there will be incredibly strong regulation banning AI from taking the jobs of the PMC. The power to regulate is the power to destroy, and as incapable of actual productivity the PMC and their legion of bullshit jobs are, they know how to run a grift and bask in their own self importance.

I highly doubt this will happen. You talk as if the PMC is a giant union where everyone is aligned, which shows you don't understand the social context there and are clearly just poo-pooing your outgroup.

People in the PMC with power have capital, whether it's political, intellectual, or financial. The financial movers and shakers will not agree to regulating AI, at least until they have gotten their piece of the pie. Even if they do, it will take years and years to get everyone to agree on a framework.

You've also got the AI companies themselves. Altman has come out and said he doesn't think regulation at this stage is a good idea, and he's got an incredible amount of political and intellectual capital. Many people in government, for good reason, see Altman as one of the most important figure in the world right now. They don't want to piss him off.

The advent of generative AI heralds the single largest change in the structure of human society since the neolithic revolution (ie. the invention of agriculture and the settled society) 12,000 years ago.

I would actually argue this is a closer parallel to the cognitive revolution, or homo sapiens first discovery of culture, language, and general cognitive technology. The difference is that with the revolution from fire, or the agricultural revolution, or the industrial revolution, or even the internet, the AI revolution deals with intelligence and a paradigm of thinking itself. The Scientific revolution could also be a close contender, since it dramatically increase our ability to think and use our knowledge.

The only thing that seems certain is that it will radically reshape the life of every single human being on earth in the next 5-50 years.

Agree strongly here.

Currently, we're focused on the application of modern LLMs and other generative models to create media (writing, images, video etc) and to perform knowledge roles that involve a combination of text and data manipulation and basic social interaction (ie. the vast majority of PMC labor sometimes derogatorily referred to as 'email jobs'). But current models are so generalizable, and LLMs already appear to translate so well to robotics that even relatively complex physical labor is only a few years behind the automation of the PMC, especially given rapid improvements in battery technology and small motors, which are some of the other major bottlenecks for robotic labor.

The real step change in my opinion is once these models get good at things like drug discovery, mathematical proofs, and building models of physics. We have essentially been locked into a paradigm almost 100 years old in physics, and haven't found many fundamental changes in mathematical or chemical theory since then either, to my knowledge.

In the past, every time we had a major breakthrough in one of these fields it was enough to reshape the world entirely. Chemistry led to the industrial revolution, Newtonian Mechanics led to the scientific revolution. (or was the beginning, whatever.)

There is a (relatively persuasive) case to be made that the invention of agriculture led to a decline in the quality of life for the vast majority of human beings that lasted until the late 19th or early 20th century. It took 11,900 years for the neolithic revolution to pay quality of life dividends, in other words. We can only hope that the period of relative decline in quality of life is shorter this time round, or perhaps avoidable altogether.

As I mention above, I think the comparison to the agricultural revolution falls flat for a number of reasons. Admittedly most revolutions follow a pattern of short term negative issues with long term positive outcomes however.

For the purposes of this comment, I will try to define good as "improving the quality of life for many people without decreasing the quality of life for another similarly sized group" an vice versa.

Tangential, but the term in economics you are touching here is a Kaldor–Hicks improvement I think. It's not Pareto-optimal, but total-wealth increasing, and could theoretically be converted to a Pareto-optimal situation with redistribution from the winners to the losers (assuming such redistribution does not have any externalities itself!).

Personally am very excited for AI improvements. I’m hoping something like ChatGPT will be able to act as a super personal assistant and analyst.

For example in personal life, would love to be able to type into a box that I’m looking to plan a trip with just a few parameters (date, general budget, etc) and have it send me options. I can then have the AI send even more options for what do on the trip and finally book reservations that only require my approval.

That’s just one example but there are plenty of admin type activities that I’d like to offload to an AI. The opportunities in professional life are even greater but I think that may take longer as the aversion to giving the AI access to confidential data may be high (it’s currently banned at my mega corp).

I’m hoping something like ChatGPT will be able to act as a super personal assistant and analyst.

At what level of 'smarts,' however, will an AI that is already training on how you do your job going to stop needing you around to do it?

I mean, you're basically happily accepting an apprentice who will lighten your workload whilst learning your job, except this thing is known to learn 100x faster than your standard human. The assumption that you'll have moved on to bigger and better things (or retired) before the apprentice steps up to take over your job may not hold here.

I think at least in the short/medium term this technology could lead to large productivity gains without corresponding cuts in total headcount.

When I started my career finance teams used to forecast in excel using relatively simple formulas. Now they use coding languages and forecast more frequently, with greater detail, and greater accuracy while working with massive data sets. This hasn’t lead to a huge cut in overall headcount, but it has changed the skill set mix on the teams.

Right, but it's presumably cheaper to spin up more GPT instances or build up more datacenters than it is to train more 'experts' in fields that are susceptible to ML automation.

Hence the question:

At what level of 'smarts,' however, will an AI that is already training on how you do your job going to stop needing you around to do it?

I'm not really doubting that humans will be 'in the loop' for quite a bit longer, but I suspect it will be more 'sanity checking' AI outputs and/or as a backup in case of outages, and there'll be strong downward pressure on wages. Which is fine if productivity gains make things cheaper.

But you're talking about AI as a complement to human skills, but I'm very specifically inquiring about how smart it needs to get to replace given skill sets.

I think at least in the short/medium term this technology could lead to large productivity gains without corresponding cuts in total headcount.

Agreed. It's just psychologically painful to fire people, and especially if companies are making a ton of money from these models I don't think there will be a giant firing spree. As we saw with all the recent layoffs at big tech, when times are good companies are more than willing to keep a bunch of low impact employees on the payroll, especially in tech.

when times are good companies are more than willing to keep a bunch of low impact employees on the payroll

Also, it helps crowd out competition. Why fire a bunch of people when the interest rate is zero?

Sure, you'll save money in the short term, but those workers don't just disappear from the labor market; enterprising competitors will snap them up and end up requiring you to offer them a billion dollar acquihire scheme to shut them down before their product starts taking your marketshare.

Better to just keep them at Bigco. Sure, they won't really develop anything for you, but why drive the state of the art forward when you can just ignore all your customers, keep your competitors down, and rake in the cash from your ad business?

At what level of 'smarts,' however, will an AI that is already training on how you do your job going to stop needing you around to do it?

At some point soon we will at least increase productivity by 1.5-2x per person. At that point why don't we collectively demand a 3 or 4 day workweek?

Humans don't 'collectively' demand things because generally there's a massive divergence in values at scale. Coordination problems abound.

And put simply, if you can make $4000 for a 4 day work week, and $5500 for a 5 day work week, then there are plenty of rational reasons to just work an extra day.

The choice to do or not do so comes down to, I'd say, values, as above. If you have high time preference and thus value leisure and 'fun' things, you'll try to minimize the time spent working as much as you can.

The markets will balance supply of labor and demand for labor, as they always do, unless we actually do achieve fully automated gay luxury space communism.

As usual, WTF Happened in 1971 is a fitting reference. Productivity and compensation stopped correlating in 1971, and we haven't (effectively) collectively demanded a reduced work week yet.

We could have transitioned to three day work weeks way before 1971. The flaw in Keynes's famous prediction is that, past the point of basic subsistance, economic utility is relative. People don't want to make $20,000 or $50,000 or $100,000 or $200,000 inflation-adjusted household income to be happy. They want more than their peers. They want to have class-markers that low status people don't, not the luxuries that those class-markers manifest themselves in. It's why the canard about modern trailer trash having it better than kings in 1900 is so ridiculous.

If whatever happened in 1971 never happened, people would still be working as much as ever. The hedonic treadmill would just be moving faster.

First ask yourself this: why do you not already have a 3 day workweek?

Because I'm too poor.

I often think of the possibility that ML is right now our best and maybe only chance to avoid some massive economic downturns due to a whole hell of a lot of chickens coming home to roost all at the same time.

I will ignore the AI doomer arguments which would suggest protracted economic pain is preferable to complete annihilation of the human species for these purposes.

I am in a state of mind where I'm not sure whether we're about to see a new explosion in productivity akin to a new industrial revolution as we get space-based industry (Starship), broad-scale automation of most industries and boosted productivity, and a massive boost in human lifespans thanks to bio/medical breakthroughs... OR

Maybe we're about to see a global recession as energy prices spike, the boomer generation retires and switches from production and investment to straight consumption or widespread unrest as policies seek to avert this problem, international relations (and thus trade) sour, even if there's no outright war, and a general collapse in living standards in virtually everywhere but North America.

How the hell should one place bets when the near-term future could be a sharp downward spike OR a sharp exponential curve upwards? Yes, one should assume that things continue along at approximately the same rate they always have. Status quo is usually the best bet, but ALL the news I'm seeing is more than sufficient to overcome my baseline skepticism.

But the possible collapse due to demographic, economic, and geopolitical issues seems inevitable in a way that the gains from Machine Learning do not.


The problem, which you gesture at, is that this world is going to be very heavily centralized and thus will be very unequal at the very least in terms of power and possibly in terms of wealth.

ALREADY, ChatGPT is showing how this would work. Rather than a wild, unbounded internet full of various sites that contain information that you may want to use, and thus thousands upon thousands of people maintaining these different information sources, you've got a single site, with a single interface, which can answer any question you may have just as well.

Which is great as a consumer, except now ALL that information is controlled by a single entity and locked away in a black box where you can only get at it via an interface which they can choose to lock you out of arbitrarily. If you previously ran a site that contained all the possible information about, I dunno, various strains of bananas and their practical uses, such that you were the preferred one-stop shop resource for banana aficionados and the banana-curious, you now cannot possibly hope to compete with an AI interface which contains all human-legible information about bananas, but also tomatoes, cucumbers, papayas, and every other fruit or vegetable that people might be curious about.

So you shut down your site, and now the ONLY place to get all that banana-related info is through ChatGPT.

This does not bode well, to me.

And this applies to other ML models too. Once there's a trained model that is better at identifying cavities than almost any human expert, this is now the only place anyone will go to get opinions about cavities.

The one thing about wealth inequality, however, is that it's pretty fucking cheap to become a capital-owner. For $300 you can own a piece of Microsoft. See my aforementioned issues about being unsure where to bet, though. Basically, I'm dumping money into companies that are likely to explode in a future of ubiquitous ML and AI models.

Of course, if ML/AI gets way, WAY better at capital allocation than most human experts, we hit a weird point where your best bet is to ask BuffetGPT where you should put your money for maximum returns based on your time horizon, and again this means that the ONLY place people will trust their money is the the best and most proven ML model for investment decisions.

Actually, this seems like a plausible future for humanity, where competing AI are unleashed on the stock market and are constantly moving money around at blinding speeds (and occasionally going broke) trying to outmaneuver each other and all humans can do is entrust one or several of these AIs with their own funds and pray they picked a good one.

Yep.

In retrospect, I actually begin to wonder if the increasing tendency to throw up paywalls for access to various databases and other sites which used to be free access/ad supported was because people realized that machine learning models were being trained on them.

This also leads me to wonder, though, is there information out there which ISN'T digitized and accessible on the internet? That simply can't be added to AI models because it's been overlooked because it isn't legible to people?

If I were someone who had a particularly valuable set of information locked up in my head, that I was relatively certain was not something that ever got released publicly, I would start bidding out the right to my dataset (i.e. I sit in a room and dictate it so it can be transcribed) to the highest bidder and aim to retire early.

Is there a viable business to be made, for example, going around and interviewing Boomers who are close to retirement age for hours on end so you can collect all the information about their specialized career and roles and digitize it so you can sell it and an AI can be trained up on information that would otherwise NOT be accessible?

This also leads me to wonder, though, is there information out there which ISN'T digitized and accessible on the internet? That simply can't be added to AI models because it's been overlooked because it isn't legible to people?

There is actually a ton of information that has not been digitized and only exists in, for example, national archives or similar of various countries or institutions.

I hadn't actually realized that this was the case until I started listening to the behind the scenes podcast for C&Rsenal - they're trying to put together a comprehensive history or the evolution of revolver lockwork, and apparently a large amount of the information/patents are only accessible via going there in person.

This is fascinating and it suggests that training AI on 'incomplete' information archives could lead to it making some weird inferences or blind guesses about pieces of historical information is simply never encountered.

I now have to wonder if there are any humans out there with a somewhat comprehensive knowledge of the evolution of revolver lockwork.

And now we have to wonder just HOW LARGE the corpus of undigitized knowledge is, almost by definition we can't know how much there is because... it's not documented well enough to really tell.

This is fascinating and it suggests that training AI on 'incomplete' information archives could lead to it making some weird inferences or blind guesses about pieces of historical information is simply never encountered.

Well this is basically how C&Rsenal started their revolver thing... doing episodes on multiple late 19th century European martial revolvers and realizing that the existing histories are incomplete.

I now have to wonder if there are any humans out there with a somewhat comprehensive knowledge of the evolution of revolver lockwork.

Probably the best one right now would be Othais from C&Rsenal.

And now we have to wonder just HOW LARGE the corpus of undigitized knowledge is, almost by definition we can't know how much there is because... it's not documented well enough to really tell.

I would guess that a huge amount of infrequently requested data is totally undigitized still.

Actually, another area that demonstrates this: I frequently watch videos about museum ships on youtube and so much of the stuff they talk about is from documents and plans that they just kinda found in a box on the ship. So much undigitized.

Probably the best one right now would be Othais from C&Rsenal.

And this is my thought now, that he has a potentially valuable cache of information in his head he could sell the rights to digitize for use training an AI.

I don't know that he can really monopolize it--on the C&Rsenal website itself, there is a publicly-available page where they've put together a timeline of revolver patents. I think Othais's passion as a historian outweighs his desire to secure the bag.

How do you compile and digitize it effectively?

THAT question seems to be answered already. Audio recordings fed to an AI that can transcribe to digital words gets you there.

Plus, a lot of it might just be self-aggrandizing nonsense.

I mean, the internet pretty much thrives on that sort of information, which is what the ML algos are trained on anyway.

To get a bit Lao Tzu, the information that can be collected and digitized isn't the real, valuable information.

At some point LLMs may be able to speak the True Dao. Their whole shtick is essentially building an object that contains multiple dimensions of information about one concept, yes?

There may be a viable but difficult business there anyways; you'd basically be doing the same work as an old folklorist gathering stories as cultures die. How do you craft the questions to know what to ask? How do you compile and digitize it effectively?

The AI can craft the questions. The AI can ask them too. It's already a more attentive and engaged listener than many humans (me included).

I know something the superintelligent AI doesn't? It would like to learn from me? What an ego boost!

I do wonder if we'll create a framework where places like OpenAI need to pay fraction of cents for each token or something. It would hit their profitability but would still make things fine if they achieve AGI.

Otherwise I agree that the open structure would be tough.

I'm not talking about after they train, I'm basically saying that in order to train on data or scrape it period, they would have to pay. Otherwise all data would be walled off. (Not sure if we could do this to only LLMs without making the internet closed again - that's a concern.)

or the existence of Peru

Is there anyone in the English-speaking world who didn't learn about the existence of Peru from Paddington Bear?

Me

Once there's a trained model that is better at identifying cavities than almost any human expert, this is now the only place anyone will go to get opinions about cavities.

It seems unlikely that there would only be one, though, unless there are barriers to entry e.g. the US government makes severe AI alignment requirements that only Microsoft can meet. Even Google, at its peak, was not the only search engine that people used.

I am amenable to this thought.

But if there's one ML model that can identify cavities with 99.9% accuracy, and one that 'merely' has a 98.5% accuracy, what possible reason could there be for using the latter, assuming cost parity.

Microsoft is an interesting example of this since they have 75% market share on PC OS. If they successfully integrate AI into windows I can see that going higher.

But if there's one ML model that can identify cavities with 99.9% accuracy, and one that 'merely' has a 98.5% accuracy, what possible reason could there be for using the latter, assuming cost parity.

Depends on how much the first ML model exploits its advantage. Also, firms often push for monopolistic competition rather than straight imitation, so the firm marketing the 98.5% model might just look for some kind of product differentiation, e.g. it identifies cavities and it tells funnier jokes.

The key difference in Machine Learning is that it changes computing from a process where you tell the computer what to do with data, and turns it into a process where you just tell the computer what you want it to be able to do.

I think there is yet another point to make here. With current Large Language Models, we have systems that treat Natural Language as a code, that is where revolution comes from. Even before LLMs, there were multiple "revolutions" where instead of working directly with machine code you could work with higher level languages utilizing concepts more suitable for humans as opposed to "data" in its raw form. This made programming incrementally more accessible to wider population. Even things like invention of graphical user interface for operating systems enabled people to tell computers what to do with data in more natural way without some arcane knowledge.

Also on the level of let's say algorithms creating novel things on some simple inputs, there was procedural generation around for a long time. Giving the computer system some simple parameters and computer running the simulation to confirm/falsify end result was a standard thing in the past. Again, the key difference is that we now have a very powerful system that can treat natural language as a code.

You might be interested in this post on LessWrong which discusses Scaffolded LLM's as natural language computers.. I'd be curious on @DaseindustriesLtd's take on this as well.

Key points:

What we have essentially done here is reinvented the von-Neumann architecture and, what is more, we have reinvented the general purpose computer. This convergent evolution is not surprising -- the von-Neumann architecture is a very natural abstraction for designing computers. However, if what we have built is a computer, it is a very special sort of computer. Like a digital computer, it is fully general, but what it operates on is not bits, but text. We have a natural language computer which operates on units of natural language text to produce other, more processed, natural language texts. Like a digital computer, our natural language (NL) computer is theoretically fully general -- the operations of a Turing machine can be written as natural language -- and extremely useful: many systems in the real world, including humans, prefer to operate in natural language. Many tasks cannot be specified easily and precisely in computer code but can be described in a sentence or two of natural language.

The LLM itself is clearly equivalent to the CPU. It is where the fundamental 'computation' in the system occurs. However, unlike the CPU, the units upon which it operates are tokens in the context window, not bits in registers. If the natural type signature of a CPU is bits -> bits, the natural type of the natural language processing unit (NLPU) is strings -> strings.

The RAM is just the context length. GPT4 currently has an 8K context or an 8kbit RAM (theoretically expanding to 32kbit soon). This gets us to the Commodore 64 in digital computer terms, and places us in the early 80s.


The obvious thing to think about when programming a digital computer is the programming language. Can there be programming languages for NL computers? What would they look like? Clearly there can be. We are already beginning to build up the first primitives. Chain of thought. Selection-inference. Self-correction loops. Reflection. These sit at a higher level of abstraction than a single NLOP. We have reached the assembly languages. CoT, SI, reflection, are the mov, leq, and goto, which we know and love from assembly. Perhaps with libraries like langchains and complex prompt templates, we are beginning to build our first compilers, although they are currently extremely primitive.

I find this framing extremely persuasive, and awesome in the true sense. If transformers can actually act as a new type of general purpose computer using natural language, the world will become strange indeed very quickly.

As @2rafa and others have mentioned, ML will be a step change in how human society creates value and interacts with the world more generally. Once we've achieved AGI, roughly defined as having an AI that can act at the level of an ordinary human, our ability to solve problems will drastically increase.

Intelligence is the generic solver for essentially any problem. Sam Altman himself has said that by 2030 he envisions a world where every product and service will either have integrated intelligence, or be angling towards that. This means that our phones, laptops, PCs, will all obviously be intelligent. However what most people don't realize is this technology will also effect our coffeemakers, stoves, thermostats, glasses, and practically every other technology you can think of. I'm sure adaptive clothing will exist soon with camoflauge like capabilities. People will be able to get realtime instructions into headphones telling them exactly how to complete each task.

Even these predictions only scratch the surface. If the true promise of AGI comes out it will also let us break through issues in hard mathematics, create brand new drugs, find extremely dense and powerful new materials. It will help navigate endless layers of bureaucracy, effortlessly pruning through the thousands of regulations that hold up large projects, helping us pinpoint ruthlessly where cost is added to solve the cost disease problem, and generally help unstick our public works. We could be building scintillating skyscrapers of filament-thin materials with bridges across the sky that glisten in the air, all in a decade. The future is truly difficult to even envisage, let alone predict.


In terms of comparisons to other revolutions, @2rafa says below:

There is a (relatively persuasive) case to be made that the invention of agriculture led to a decline in the quality of life for the vast majority of human beings that lasted until the late 19th or early 20th century. It took 11,900 years for the neolithic revolution to pay quality of life dividends, in other words. We can only hope that the period of relative decline in quality of life is shorter this time round, or perhaps avoidable altogether.

I agree that the agricultural revolution led to issues, a la Scott Alexander's review of Against the Grain.. That being said, I find the comparison of the AI revolution to agriculture as facile. Ultimately the reason the agricultural revolution proved bad for us was that we shifted our lifestyles from nomadic culture to a static culture - which inherently leads to problems of physical fitness, freedom, social control, and cultural institutions have to rapidly shift.

With the AI revolution, we have no idea how far it will go. The possibility space is far beyond what could have existed for any previous revolution. As doomers say, we could all die. We could all transcend our fleshly forms and become gods in ten years. China may create an ASI and lock us all into a totalitarian doom state forever.

The stakes here are far higher than the agricultural revolution, and I highly doubt our situation will parallel that trajectory.

At the end of the day if we can survive the AI revolution without any horrible outcomes of the x-risk or s-risk variety, I think it would be ridiculous to posit any sort of negative future. With intelligence at our fingertips, we will be able to finally achieve our potential as a species.

People will be able to get realtime instructions into headphones telling them exactly how to complete each task.

Where have I heard this one before?

Seriously, this seems too specific to be a coincidence. Was it a deliberate reference?

Nope. I actually don’t like referencing that story because I think it’s pretty short sighted, although does have some interesting ideas.

This is a very common thought in any hard sci fi that has AI. Manna is by no means original just popular in the rat sphere.

Thought your link led to this

Frankly at this point I'm just riding the tides. Whatever happens, happens. This will be like the fifth once in a generation event I've lived through, and like the 20th doomsday scenario. I don't have the energy to care anymore. I have apocalypse fatigue.

But there could be a utopia! Unlike Nuclear and other scenarios, I think it's likely this moves us far closer to a utopia, soon.

There could be! And that would be nice! But like with giving in to doom-mongering, I'm also not going to get my hopes up, either. Realistically, whatever's going to happen is going to happen regardless of whether I get hyped up or stressed out about it.

Yeah I try and keep a cool head as well. I'd love to quit my job and party till the singularity comes but it may not be the best idea...

People thought this about Nuclear and the other ones too if you remember.

There will be no utopia, because utopia is not a thing that exists. Our lives might get better and worse in various ways, but the idea of a perfect society, and by extension of moving towards a perfect society, has always been delusional.

No utopia, just a shifted technological landscape.

There could be a utopia but it could only be achieved by either

  1. Changing the human race fundamentally to remove the desire for accomplishment or status

  2. Hiding the true nature of reality and creating a unique simulation for each human that would provide a fulfilling life path for that person

Why do you think this is the case? And what does a 'fundamental' change mean?

The goal of the axial revolution has always been to improve ourselves. We are slowly becoming better, in my opinion. Less violent, more understanding, more focused on technical accomplishment. If we continue on that path and eventually eschew (most) status, is that a fundamental change or an incremental one?

but the idea of a perfect society, and by extension of moving towards a perfect society, has always been delusional.

Some delusions are worth chasing my friend. Chasing the delusion of truth, intellectual honesty, and rigor led us to the Scientific revolution, which brought us where we are today. Just because you don't think it's likely doesn't mean those seeking utopia are fools.

Utopia doesn't exist the same way any other ideal doesn't exist. Does that mean you shouldn't strive to be kind or love others?

Once you accept that these are forces which you can't individually impact, the path forward becomes pretty clear.

Just set things up to maximize your chances of living to see whatever crazy future we end up with.

And maybe have some fun along the way.

  1. Go to church

  2. Have kids

  3. Buy land

  4. Acquire chickens

Simple as.

I truly think people are almost embarrassingly overstating the importance of the AI apocalypse. Maybe an apocalypse for twitter and other online spaces, maybe an apocalypse “just a barely intelligent warm body” call center jobs, maybe an apocalypse for bootcampers making $300k/yr gluing JavaScript frameworks with cute names together.

Not an apocalypse for anybody with a skill set that can exist completely independent of the internet, not an apocalypse for the people who understand computer programming from first principles.

In the sense the AI will bankrupt the people who have been mining the good out of society while contributing absolutely nothing of value to it, it is a massive net good. I absolutely welcome our AI overlords. Show me who is posting the MOST human-passing-but-totally-useless-garbage on twitter, or trapping the MOST ethical non-monogamist coombrained Reddit atheism posters into pointless time wasting arguments and I will either go work for them for free, or donate compute time to them.

Let’s fucking go.

Acquire chickens

Skipped 1, but I'm on 4. Chickens are about 2 weeks old, and I'm assessing the plans for the coop I plan to build. At least, after I finish ripping out the stupid Cyprus trees the last owner planted everywhere.

Based and eggpilled.

Seriously love chickens. They are equally stupid and annoying, and beautiful. They also make fantastic babysitters for #2 and will entertain them for HOURS. Highly recommend.

Chickens are raging assholes that go everywhere they're not supposed to and refuse to die when their time is up.

Ducks are much easier to manage. The eggs are tastier, too.

I was wondering if we were going to get the chicken vs duck argument going. I have a coworker who has ducks and recommends them. I have a neighbor with chickens, although they might have gotten rid of them, or at least the roosters.

I want to try guinea fowl next year.

Our previous neighborhood had feral peacocks, and they give off this great jungle call in the middle of the night, and every once in a while I hear them here too, from a half mile or so away.

I didn't know such arguments were infamous.

All I know is, after having to deal with both, I'll take the ducks.

Nah, ducks turn their ponds into swamps and give you a rash when you cuddle them. Chickens are much more convenient. (We have both.)

Ducks require too much feed. Geese can graze most of the day.

Has anyone considered…pet pigeons?

Pigeon eggs can be eaten too!

coo coo

I had about twenty white homing pigeons as a teen for 4-H. They're great, but are terribly difficult to get rid of. Homing ability is both impressive and obnoxious.

Not an apocalypse for anybody with a skill set that can exist completely independent of the internet, not an apocalypse for the people who understand computer programming from first principles.

In the sense the AI will bankrupt the people who have been mining the good out of society while contributing absolutely nothing of value to it, it is a massive net good.

I can't tell if this comment is a spoof?

Sure, go back to your farm and use tools like tractors, fertilizers, modern crop rotation techniques, plates, silverware, cups, etc which have been created by the larger society. Created, distributed and improved by people who are supposedly 'mining the good out of society.'

Society is a team effort, bud. Your fantasies of living scott-free totally 'independent' on your plot of land are just that - fantasies. You wouldn't make it a week without the collective wisdom and knowledge society has gifted you and your family. Have some respect for the people who came before you, and the people who help you live a cushy life now.

I say:

go to church

start a family

And you internet this as “isolate yourself from society and pay no respect to the people who came before you”?

Just to be clear when i say “go to church”, I mean specifically a Catholic Church. There could not exist another institution on planet earth that is more of a strong indicator that you should stand in the shoulders of the people who came before you.

The people mining the good out of society are people running porn websites, and AB testing headlines and algorithmic content feeds to see which ones make people hate each other more, and then buy the products that they’re selling. Onlyfans is mining the good out of society, blackrock is mining the good out of society, McKinsey consulting is mining the good out of society

Porn websites and management consulting agencies did not invent pottery, crop rotation, iron smelting, or anything else. The fact that you either think otherwise or think that “go to church and start a family” somehow means “throw away every good discovery ever made by mankind” is certainly telling of something.

The people mining the good out of society are people running porn websites, and AB testing headlines and algorithmic content feeds to see which ones make people hate each other more, and then buy the products that they’re selling.

This is a small fraction of people in modern society, and if history tells anything I'd imagine they will be hurt less by AGI because this class of people is good at finding BS niches to milk value out.

I'm just not a fan of broad statements talking about how an ill-defined outgroup is milking everything from society while you and yours are the ones building it. Thanks for clarifying.

The people mining the good out of society are people running porn websites, and AB testing headlines and algorithmic content feeds to see which ones make people hate each other more, and then buy the products that they’re selling. Onlyfans is mining the good out of society, blackrock is mining the good out of society, McKinsey consulting is mining the good out of society

Those people will be doing more of all that and better (or rather "more efficiently" - nothing about it will be better for the audience), with higher profit margin since they'll no longer need to pay the grunts in call centers.

I think this comment is an example of "inferential distance." Your meaning of "people mining the good out of society" is porn sites, investors, and engagement-optimizers, whereas Dag's interpretation was "all the smart people who brought us modern technology."

@firmamenti also engaged in the classic Motte and Bailey to my mind. His Bailey is:

Not an apocalypse for anybody with a skill set that can exist completely independent of the internet

Basically claiming that anyone who relies on the Internet is gonna get fukt, and they should cry about it.

Then when challenged he retreated to the much more specific claim of:

people running porn websites, and AB testing headlines and algorithmic content feeds to see which ones make people hate each other more,

I'm not impressed with this sort of rhetoric.

Not an apocalypse for anybody with a skill set that can exist completely independent of the internet

Basically claiming that anyone who relies on the Internet is gonna get fukt, and they should cry about it.

Just pointing out, your interpretation there doesn't quite check out logically. It would only be a motte/bailey when mischaracterized like that.

I'm not impressed with this sort of rhetoric.

You cut one of my sentences in half to make your point, and then you accused me of bad faith argument.

The rest of the statement which you cut off was: "not an apocalypse for the people who understand computer programming from first principles."

This is not a motte and bailey. You either didn't read the rest of my comment, or you are being deliberately misleading in your characterization of it.

Either way: don't do this.

Eh, I cut it out for brevity but I see where you’re coming from. Either way I see you slicing the populace into such a chunk as to be making a ridiculously callous and egotistical statement.

I’m happy to discuss further which chunk of humanity deserves to have their lives be destroyed and suffer unnecessarily, but I generally find that type of rhetoric to be unsavory. I apologize if I mischaracterized your stance.

I’m really getting the urge to grow out my neckbeard and get euphoric up in this bitch. Postrats are converting to Mormonism now, Mormonism! At least with wokeness you have to go outside and observe the world to realize that it’s false. Most of these religions don’t even make sense on their own terms.

It’s cope is what it is, cope. It makes you feel good, and it’s useful (so it seems), so you believe it.

Choosing to believe (or act as if you believe) useful things seems very rational to me. I have an old coworker who was an atheist and cynically became a Mormon in order to marry a Mormon wife and live in a close-knit community. He now lives in Idaho and has 4 kids and by all accounts is very satisfied with the outcome. Who's more rational, him or a depressed medicated outspokenly atheist Bay area tech worker who's the least-liked member of his drama-cursed polycule?

If you rational long enough, you're eventually going to rational about rationality, and you'll see that beliefs are instrumental like anything else. There's no God of Integrity who laid down the law that you must profess true beliefs.

The short answer is, it fucks up your epistemology. It’s probably worth a whole post going through exactly why that’s so bad. Perhaps the old atheism arguments from the early 2000s need updating for the TikTok generation.

I disagree. You can be rational when the situation calls for it, and be religious on a meta level.

It definitely deserves a longer treatment than one sentence, but I'm fond of "once you've told a lie all truth is your enemy". Or something about lightning, I guess. Intentionally professing beliefs in falsehoods because they are useful is the epistemic equivalent of the doctor killing their patients to donate their organs -- it may sound like it does more good then harm in the short term, but you wouldn't want to live in a place where that's the rule.

So... I have a church, yard, kids, and chickens. Also it's Bright Week. Al masih qam!

Yet here I am, typing away on The Motte about AI. And here you are.

Plausibly I should work on my in-person network. A local church has installed ten Russian bells on a new building they've been working on these past two years. I watched the video of the blessing, and it sounds really good. The acequias association is supposed to be flushing the irrigation ditches tomorrow. My husband walked down the street and gave eggs to a neighbor last week, and has resolved to do that again, because it was a good experience. My daughter is now old enough to walk to the village church if we ever get our act together on time. People wave, and are out by the street cleaning their ditches. I can and should make physical art out of wool and wax for next year's local studio tour and art markets.

And yet here we are, even so.

Is the rapid advancement in Machine Learning good or bad for society?

Option C: neither. It's just a tool, neither good nor bad in itself. What will make it good or bad is how we use it, which remains to be seen.

Anything that takes us closer to post-scarcity is good from my perspective. I disagree with some people I otherwise respect, such as Ilforte, on the fundamental benevolence (or rather, absence of malevolence) of the ruling class, especially the ones that will end up wielding the power put in their hands by AGI. It will cost them very little indeed to at least maintain the standards of living of everyone alive today, and little more to improve everyone's to First World upper middle class levels.

Upload everyone into VR, and it's quite possible that everyone can experience eudaimonia on a 10 watt budget.

Now, I'm not a happy person. I've been struggling with depression so long that I've forgotten what it might have ever felt like to not be under a cloud, I feel fundamentally burned out at this point, experiencing something in between learned helplessness and nihilism regarding AI advances. What'll happen will happen, and everyone here is only running commentary on the impending Apocalypse.

Back when I was entering med school, I consoled myself that the suffering was worth it because medicine was likely to be among the last fields to be automated away. Can't say that I feel very vindicated, because the automation overhang is here, and I see the Sword of Damocles dangling overhead when I think about further professional advancement.

It seems awfully clear to me that medicine is about to be automated, GPT-4 is a good doctor. Probably not the best possible doctor, but already outperforming the average in an already incredibly competitive and cognitively demanding profession. I only look at the further slog of psychiatry training ahead for me and shiver, because there's absolutely no way that by the time I'm done, I'll be employed by the graces of anything other than regulatory inertia instead of genuine competitiveness.

Instead of a gradual deployment (over like 2 or 3 years, I had short timelines even then) where AI came for Radiologists, then Opthalmology, all the way to Surgery and then Psych, it seems to me that the pressure will mount until regulatory bodies cave, and overnight everyone from the lowliest janitor to the highest ranking neurosurgeon will find themselves out on their arse in short order.

What pisses me off further is that this is also a slamming shut of the clearest pathway to betterment and improved quality of life I have, namely emigration to the First World. Not a consideration for the average person here, since you're already living there, but simply imagine how fucking terrible it is to face the wall of obsolescence without having a government that can even in theory maintain living conditions by redistribution of wealth.

As a concrete example, the NHS is largely propped up by foreign doctors, with a large fraction of the locals fleeing to greener shores such as the US or Australia. Pay has stagnated for a decade, prompting serious strikes, currently ongoing, to achieve inflation based pay restoration.

Even today, when automation is merely imminent, the British government has publicly stated it's intent to automate as much of medicine as it can to stomp down on them uppity doctors who aren't content with sub-market pay from a monopsony employer. You think those cheap bastards will hesitate for more than a microsecond to get rid of doctors or at least their pay, when the moment finally arrives?

I see British doctors mocking those claims today, as much as I support their attempts at pay restoration for selfish reasons, neither I nor they will be laughing much longer.

Maybe American doctors will hold out a little longer, you lot clearly aren't very concerned with efficiency in your healthcare expenses, but places like India, or the slightly whiter version of the Indian subcontinent, will end up clamoring to get rid of any expenses for their state-run public health services.

I'm fucked, clearly out of good options, and now picking the least bad ones.

On the note of doctors, the medical guild has always been the most robust, perhaps other than lawyers, at defending its monopoly. I would be willing to bet doctors still resist automation through regulatory barriers for quite a while.

Even if that doesn’t shake out, it could be a scenario where human augmentation rolls out relatively slowly. You, being a transhumanist, Should greatly benefit in a lot of those scenarios. I imagine the vast majority of people alive today will be unwilling to augment them selves for purity-based reasons. Not having that hangup alone would be a huge competitive advantage.

If all else fails you can always mortgage your future computing space for a loan or some thing and hope to jump up into the immortal class. I for one hope can you make it, although I will admit that I am not the most optimistic when it comes to proles getting access to longevity technology.

Doctors have successfully defender their guild (albeit more so in the US than the UK, by a large margin) because they were indispensable. Training replacements to disgruntled doctors would take a great deal of time, and while medical education isn't perfect, you can't really circumvent most of it without ending up with noticeably worse practitioners.

That changes greatly when human doctors become outright obsolete, speaking in the UK context, I have little doubt that the government would happily tell all involved to take a hike if that was the cost of "saving" the NHS or even saving money.

Doctors in the UK have been cucked to put it mildly haha. They've only recently grown a backbone after the wage decreases have become unbearable.

The UK government(s) have historically relied on immigrant doctors to prop up the NHS when the locals started getting fed up about it. I can't complain about this too much, given that I intend to emigrate soon, but this certainly is responsible in part for their depressed wages.

A government willing to sideline its populace with immigrants will happily do so with AI as and when feasible, and they've already stated that that's their intent.

I could live with postponing the singularity a few years till we get it right, but that's seemingly not on the table.

(I mildly disagree that most people won't avail of transhuman upgrades. Eventually they'll end up normalized, in much the same way nobody really makes a fuss about glasses, hearing aids or pacemakers.)

That changes greatly when human doctors become outright obsolete

This is where we disagree - I don't see human doctors becoming obsolete anytime soon. Perhaps from a medical perspective, sure, but for the majority of laypeople I'd imagine a large part of a doctor's job is comforting the person they're treating.

Now I do think that like with almost all knowledge work, doctors will be able to become more productive. Especially those that don't see patients most of the day. But my understanding is that the vast majority of, say, a primary care physician's job is to go from 30 min patient visit to 30 min patient visit, hearing what people have to say and writing it down, then telling them they're going to be okay and the doctor can help.

Even if we can prove that LLMs give better medical advice than doctors 100% of the time, I don't think the majority of people would be comfortable hearing it from a non-doctor for quite a while.

I could live with postponing the singularity a few years till we get it right, but that's seemingly not on the table.

You don't think accelerating progress now could be the best way to reach alignment?

I mildly disagree that most people won't avail of transhuman upgrades. Eventually they'll end up normalized, in much the same way nobody really makes a fuss about glasses, hearing aids or pacemakers.

Depends on the speed of the takeoff, I suppose.

but for the majority of laypeople I'd imagine a large part of a doctor's job is comforting the person they're treating.

Is that true? I don't think I know anyone who thinks that, or anything even remotely close to it.

Every time I've interacted with medical professionals over the past several years, there has been no emotional component at all, or mildly negative. Doctors are able to diagnose, proscribe, and conduct operations, otherwise people would stay far away.

For instance: family member was pretty sure he had pneumonia. Went to a hospital, got an x-ray. Yep, that's pneumonia alright, here are two antibiotics that might help, come back if you're just as bad or worse in a week (edit: these were, as I remember, not actually given at the hospital. We had to drive to the pharmacy for them). The antibiotics worked, hooray. In addition to $500 upfront and $1,000 from insurance, there was another $1,000 surprise charge, botched and shuttled about through bill collection, which took six months to resolve. Next time family member has pneumonia, he'll probably hold out even longer before attempting to interface with the medical system.

I'm glad that for a couple of hours of wretched interactions, trying to hand write forms alone and delirious, and two week's pay, family member was able to get needed medicine. This is better than the vast majority of times and places. But if there were an automated scanner that dispensed antibiotics, that would be vastly better experience.

I also gave birth during the ending phase of Covid restrictions. I'm glad that there are medical interventions to deal with complications and manage pain. But there is not really any comforting being done that couldn't be replaced with a recorded voice stating what's on the fetal monitor and what it means.

there has been no emotional component at all

The flat affect, 'no emotional component' is what I mean. They are giving a sort of impartial authority to their diagnosis to make you feel okay.

I disagree with doctors, but many of the people I know in the middle-class PMC take their word as Truth.

All the people I know generally think of your average medical care professional as an opponent that you have to outsmart or out-research before you are permitted bodily autonomy and usually know less about your body than you do if you have an IQ over 120.

They'd drop them for an uncensored medical expertise AI in a second.

I would drop doctors as well but I’m trying to model the modal human. Maybe I’m failing but I think people here are far into an intelligence/tech literate bubble.

Yes, reassurance and a good bedside manner are important aspects of a doctor's role! That being said, I can see AI doing all of that too:

  1. Humans will anthromorphize anything, so a cutesy robot face on a monitor or even a deepfaked one might work. Proof of concept: Telemedicine.

  2. Otherwise unskilled individuals who are simply conveying the information provided by an AI, such as a deskilled doctor or nurse, only there as a pretty face. Still utterly catastrophic for the profession.

  3. People get used to anything, eventually when the public cottons onto the fact that AI doctors are faster, cheaper and better than humans, they'll swallow their discomfort and go with it.

Hmm, deepfakes for telemedicine would be concerning. I get your point with #2 as well, although I think that'll take some time to roll out.

I see what you mean I suppose the medical profession might be on the way out. I was supposed to be the optimistic one! Alas.

Is the rapid advancement in Machine Learning good or bad for society?

Over what time horizon?

I expect the deployment of machine learning to follow approximately the same path as every other labor saving technology humans have developed. In the short term it will be somewhat of a mixed bag. On the one hand we'll be able to produce the same/more goods at lower costs than before. On the other hand this savings will likely come with impact to the people and companies that used to produce those things. Over the long term I expect it will make people much better off.

Creative destruction!

Do you not see any difference between this paradigm shift and previous ones?

Not with respect to the fact that it will be net beneficial to humanity over the long run.

I agree with you, for what it’s worth.

I'd like to believe that, as it follows a well-established pattern. But honestly, what really happens if there's no more work left for people to do anymore? It seems that we'd have to really count on some redistribution of wealth, UBI, etc to ensure that the gains of the new automation doesn't just go to the owners of the automation (as much as I never thought I'd ever say that), or else people simply will not have the means to support themselves. Or if the job destruction is localized to just upper-class jobs, then everyone will have to get used to living like lower-class, and there may not even be enough lower-class jobs to go around. The carrying capacity of society would be drastically reduced in either situation.

In other words, what if

On the other hand this savings will likely come with impact to the people and companies that used to produce those things.

means the death of large swaths of society?

I'd like to believe that, as it follows a well-established pattern. But honestly, what really happens if there's no more work left for people to do anymore?

As others have said, there will always be work to do! As long as humans have any problems whatsoever, there will be work.

The carrying capacity of society would be drastically reduced in either situation.

How the heck does AGI reduce the carrying capacity of society? You'll have to explain this one to me.

Well, I'm hypothesizing that potentially, all (or almost all) of the solutions to all of the problems humans have may be covered by AI. If the AI is owned by a very limited number of people, then those people would be the ones who are the gatekeepers, and the ones that get most of the benefit of AI. Everyone will be paying these limited numbers of people for basically everything, and no one else would be able to make a living.

This is almost like imagining Karl Marx's worse nightmare regarding the proletariat owning all means of production, ratcheted up to unbelievable proportions. I'm no communist, nor socialist, so like I said, I never thought I'd say this. But this is a fear of mine, that AI puts everyone out of work, meaning that no one can support themselves.

If the AI is owned by a very limited number of people, then those people would be the ones who are the gatekeepers, and the ones that get most of the benefit of AI.

This doesn't really scare me. Elites generally enjoy the society they're in, enjoy feeling useful, and above others. I think the vast majority of people who could create an AGI would use it to solve most of their problems, get really rich, then use it to solve everyone else's problems with a fraction of their incredible wealth.

Going into the future things could get very nasty indeed, but at that point all problems relevant to humans right now will be solved. It'll be an issue for the next stage of intelligence in our species' life, hopefully, and I'd imagine we'll be better suited to solve it then.

But honestly, what really happens if there's no more work left for people to do anymore?

That would be awesome! People (mostly) don't work because work is awesome and they want to do it. People work because there are things we want and we need to work to get the things we want. No work left for people to do implies no wants that could be satisfied by human labor.

It seems that we'd have to really count on some redistribution of wealth, UBI, etc to ensure that the gains of the new automation doesn't just go to the owners of the automation (as much as I never thought I'd ever say that), or else people simply will not have the means to support themselves.

This paragraph seems in tension with the idea of lacking work for people to do, to me. If a bunch of people are left with unfulfilled wants, why isn't there work for people to do fulfilling those wants? This also seems to ignore the demand side of economics. You can be as greedy a producer of goods as you want but if no one can afford to buy your products you will not make any money selling them.

Or if the job destruction is localized to just upper-class jobs, then everyone will have to get used to living like lower-class, and there may not even be enough lower-class jobs to go around.

I think there's an equivocation between present wages and standard of living to post-AI wages and standard of living that I'm not confident would actually hold. Certain kinds of jobs have certain standards of living now because of the relative demand for them and people's capability to do them and the costs of satisfying certain preferences etc. In a world with massively expanded preference satisfaction capability (at least along some dimensions) I'm not sure working a "lower-class" job will entail having what we currently think of as a "lower-class" standard of living.

The carrying capacity of society would be drastically reduced in either situation.

I'm a little unclear what the "carrying capacity of society" is and how it would be reduced if we had found a new way to generate a lot of wealth.

I'm not an economist, and I know very little about econ, so it's very possible that there is something major I'm missing.

If a bunch of people are left with unfulfilled wants, why isn't there work for people to do fulfilling those wants?

This is the part of my hypothesis that's tripping me up. Could you walk me through it?

Basically, let's say that we do fundamentally believe in capitalism (because I do), that a person should have to pay for any good or service that he receives.

And let's say that there's a person who is dying of starvation, because he has no job, because AI does everything better and cheaper than he can. Therefore, no one wants to come to him to do these tasks, because they'd rather go to the owner of the AI. How does this person get the money he needs to get the food he needs?

And let's say that there's a person who is dying of starvation, because he has no job, because AI does everything better and cheaper than he can. Therefore, no one wants to come to him to do these tasks, because they'd rather go to the owner of the AI. How does this person get the money he needs to get the food he needs?

So, for this kind of situation to arise it needs to be the case that the marginal cost for providing this person the necessities of life is below the marginal value their labor can generate for others.

Notice there is nothing AI specific about this scenario. It can (and does) obtain in our society even without large scale AI deployment. We have various solutions to this problem that depend on a variety of factors. Sometimes people can do useful work and just need a supplement to bring it up to the level of survival (various forms of welfare). Sometimes people can't do useful work but society would still like them to continue living for one reason or another (the elderly, disabled, etc). The same kinds of solutions we already deploy to solve these problems (you mention some in your comment) would seem to be viable here.

It's also unclear to me how exactly AI will change the balance for a persons marginal value vs marginal cost. On the one hand the efficiency gains from AI mean that the marginal cost of provisioning the means of survival should fall. Whether directly due to the influence of AI or do to a reallocation of human labor towards other things. On the other hand it will raise the bar (in certain domains) for the marginal value one has to produce to be employed.

Partially this is why I think it will be a long term benefit but more mixed in the short term. There are frictions in labor markets and effects of specialization that can mean it is difficult to reallocate labor and effort efficiently in the short and medium term. But the resulting equilibrium will almost certainly be one with happier and wealthier people.

There exist people today who, due to disabilities or other conditions, are unable to support themselves financially. They depend on the charity of others, and in richer countries they may also get tax-funded disability benefits. If the development of AI caused a significant number of people to become unemployable, there is no reason why we couldn't just include them in that category.

If the claim that "a person should have to pay for any good or service that he receives" is to be interpreted literally, then that's not "capitalism", that's some extreme form of libertarianism, verging on parody. That would make even charity immoral. Real-life libertarians believe, at most, that people should be free to do what they want with their money, including giving it to charity. Maybe Andrew Ryan of Bioshock believes that donating to the poor is bad because it keeps them alive even though they deserve to die, but I doubt you could find a real libertarian who believes that.

I, too, "believe in capitalism", that is, I believe that a free market with some (limited) state intervention is the optimal form of social organization from a utilitarian perspective in the current technological environment. I don't believe that there is a universal moral law that people have to work for everything. If robots take all the jobs, taxing the robots' owners to provide income to the newly-unemployed would clearly be the right decision from a utilitarian perspective.

If the claim that "a person should have to pay for any good or service that he receives" is to be interpreted literally, then that's not "capitalism", that's some extreme form of libertarianism, verging on parody. That would make even charity immoral.

I don't believe that there is a universal moral law that people have to work for everything.

When I say "a person should have to pay for any good or service that he receives", I don't believe it as a moral thing, for the most part. I don't think it's immoral if someone gets something through charity. But I also don't think people should count on charity. Partly this is out of my own fears. I would hate living a life in which I was entirely dependent on someone else's charity to stay alive, where I had no control over my own destiny, no ability to provide for myself. I'd be terrified of starving to death all the time!

Also, even if I don't think it's "immoral", I do at least have an aversion to people believing that it is incumbent upon other people to provide for you (let's say if you're older than 18 and able). I'm against most of the arguments saying it's immoral for people to be rich, or saying that it's perfectly fine to just take their wealth by force, or painting rich people as monsters. However, true AGI may be where I would have to draw the line on some of my beliefs, due to the sheer magnitude of people who could be put out of work by AGI. In that case, we may have to put capitalism aside and move to a new model that works better in a post-scarcity world.

Not my article but: https://www.rintrah.nl/the-end-of-the-internet-revisited/

I'm not sure the machine learning/AI revolution will end up being all it's hyped up to be. For local applications like identifying cavities, sure. For text generation however, it seems much more likely to make the internet paradoxically much more addictive and completely unusable. There's so much incentive (and ability) to produce convincing scams, and chatGPT has proved to be both easy to jailbreak and/or clone, that any teenager in his basement can create convincing emails/phone calls/websites to scam people out of their money. Even without widespread AI adoption, this is already happening to some extent. I've had to make a second email account because the daily spam (that gets through all the filters) has made using it impossible, and Google search results have noticeably decayed throughout the course of my lifetime. On the other side of the coin, effectively infinite content generation, that could be tailored specifically to you, seems likely to exacerbate the crazy amount of time people already spend online.

Another thing I'm worried about with the adoption of these tools is a loss of expertise. Again this is already happening with Google, I just expect it to accelerate. One of the flaws of argument that knowledge-base on the internet allows us to offload our memorization and focus on the big picture, is that you need to have the specifics in your mind to be able to think about them and understand the big picture. The best example of this in my own life is python: I would say I don't know python, I know how to google how to do things in python. This doesn't seems like the kind of knowledge that programmers in the past, or even the best programmers today have. ChatGPT is only going to make this worse: you need to know even less python to actually get your code to do what you want it to, which seems good on the surface, but increasingly it means that you are offloading more and more of your thinking onto the machine and thus becoming further and further divorced from what you are actually supposed to be an expert in. Taken to the extreme, in a future where no one knows how to code or do electrical engineering, asking GPT how to do these things is going to be more akin to asking the Oracle to grant your ships a favorable wind than to talking to a very smart human about how to solve a problem.

I'm not sure I really like what I see to be honest. AI has the potential to be mildly to very useful, but the way I see it being used now is primarily to reduce the agency of the user. For example, my roommate asked us for prompts to feed to stable diffusion to generate some cool images. He didn't like any of our suggestions, so instead of coming up with something himself, he asked ChatGPT to give him cool prompts.

The best days of the internet are behind us. I think it's time to start logging off.

ChatGPT is only going to make this worse: you need to know even less python to actually get your code to do what you want it to, which seems good on the surface, but increasingly it means that you are offloading more and more of your thinking onto the machine and thus becoming further and further divorced from what you are actually supposed to be an expert in. Taken to the extreme, in a future where no one knows how to code or do electrical engineering, asking GPT how to do these things is going to be more akin to asking the Oracle to grant your ships a favorable wind than to talking to a very smart human about how to solve a problem.

We have been offloading thinking to tools forever, I highly doubt we will reach some breaking point now. We absolutely do lose knowledge when we gain this, but we trade it for more efficiency. Is it bad that we have calculators everywhere?

I'm not sure I really like what I see to be honest. AI has the potential to be mildly to very useful, but the way I see it being used now is primarily to reduce the agency of the user.

I agree with this on the advertising portion. I'm becoming increasingly concerned that targeted advertising could lead to terrifying outcomes, like a small group controlling public opinion. (actually that already exists, but still)

As i read your comment, ive just completed the mass effect legendary collection. Also spoilers. ||So anyway for those who have never played the general gist of the games is this: there is a sentient race of machines called reapers and they are cleansing the galaxy of all advanced life every 50000 years. During the 3rd game, some notable things happen, mainly:

You have A personal AI on the ship you command in the game named EDI, she gets her own body, and is relatively harmless. She also evolves: she learns things like sacrifice, attempts to date the pilot, and tries to find meaning in her own existence generally.

This goes back a bit farther then game 3, however there is an AI race called Geth, that were made by a different alien species. Long story short, the game is a RPG where your decisions impact the story, and you choices impact how things with the Geth and their creators play out. Quarians basically tried to destroy the geth out of fear, but later on as you learn about the geth. They really just want to exists and be left alone, and they even help you fight the reapers. The game gives you the choice to destroy the geth, or you can humanize them and give them basic human decency. There is a scene in the game where a Quarian tries to experiment on one of the geth, and you can basically shut it down and tell the quarian not to.

You meet another alien race in the game that are responsible for the reapers, that basically tell you that they made an AI that is responsible for the reapers, it was ironically created to prevent computers from destroying organic species. The AI turns on them, converts them into robots, and procedes to take over the galaxy in hopes to preserve organic species forever in robot form. Near the end of the game you meet the reaper AI and he basically gives you 3 options: Destroy them, Control them, or Synthesis (you can also just flat out not choose)

Destroy and control are pretty straight forward, however synthesis is where you become one entity with the machines. Its essentially transhumanism. Its suppose to be the "ideal" solution.||

Now mind you, mass effect 3 got a lot of shit when it was released because the endings were abhorrent, however i could see any one of these happening when real AI gets created, maybe we'll control it and everything turns out OK ish the AIs end up being neutral or benevelont like EDI or the Geth, in a slim chance we end up successfully destroying it if things go wrong. Or we reach some perfect transhumanist state. I think with the current things going however, its arguably more likely that we'll become, well, ill let the video speak for itself (most reliable data suggests this current trajectory)

Finally, concrete plan how to save the world from paperclipping dropped, presented by world (in)famous Basilisk Man himself.

https://twitter.com/RokoMijic/status/1647772106560552962

Government prints money to buy all advanced AI GPUs back at purchase price. And shuts down the fabs. Comprehensive Anti-Moore's Law rules rushed through. We go back to ~2010 compute.

TL;DR: GPU's over certain capability are treated like fissionable materials, unauthorized possession, distribution and use will be seen as terrorism and dealt with appropriately.

So, is it feasible? Could it work?

If by "government" Roko means US government (plus vassals allies) alone, it is not possible.

If US can get China aboard, if if there is worldwide expert consensus that unrestricted propagation of computing power will kill everyone, it is absolutely feasible to shut down 99,99% of unauthorized computing all over the world.

Unlike drugs or guns, GPU's are not something you can make in your basement - they are really like enriched uranium or plutonium in the sense you need massive industrial plants to produce them.

Unlike enriched uranium and plutonium, GPU's were already manufactured in huge numbers, but combination of carrots (big piles of cash) and sticks (missile strikes/special forces raids on suspicious locations) will continue dwindling them down and no new ones will be coming.

AI research will of course continue (like work on chemical and biological weapons goes on), but only by trustworthy government actors in the deepest secrecy. You can trust NSA (and Chinese equivalent) AI.

The most persecuted people of the world, gamers, will be, as usual, hit the hardest.

The trick now is t