@self_made_human's banner p

self_made_human

amaratvaṃ prāpnuhi, athavā yatamāno mṛtyum āpnuhi

16 followers   follows 0 users  
joined 2022 September 05 05:31:00 UTC

I'm a transhumanist doctor. In a better world, I wouldn't need to add that as a qualifier to plain old "doctor". It would be taken as granted for someone in the profession of saving lives.

At any rate, I intend to live forever or die trying. See you at Heat Death!

Friends:

A friend to everyone is a friend to no one.


				

User ID: 454

self_made_human

amaratvaṃ prāpnuhi, athavā yatamāno mṛtyum āpnuhi

16 followers   follows 0 users   joined 2022 September 05 05:31:00 UTC

					

I'm a transhumanist doctor. In a better world, I wouldn't need to add that as a qualifier to plain old "doctor". It would be taken as granted for someone in the profession of saving lives.

At any rate, I intend to live forever or die trying. See you at Heat Death!

Friends:

A friend to everyone is a friend to no one.


					

User ID: 454

Humans dreamed of heavier-than-air flight since the Neolithic if it's commonality in dreams is anything to go by. Didn't stop the invention of airplanes and rockets.

Automation is only replacing jobs with low consequences of failure. Planes fly themselves already, yet pilots are still paid to sit in front.

Pilots are a tiny fraction of the cost of operating an airplane. Most airlines buy models that were developed years or decades before they reach the hand of consumers. Refits and validation of existing control systems are expensive.

We've got self-driving cars in commercial use now, available for the common prole to hire, I can only chuckle ruefully at anyone who finds that less significant, when the set of drivers of motor vehicles so grossly outweighs the number of pilots.

No they will not. The military could be much more automated than it already is, but refuses to allow computer programs to control its most expensive assets.

That may or may not be true, until race dynamics develop and they have no choice but to hand control over to their AI, initially with rubberstamping that will only get more minimal and eventually non-existent. The alternative is being rolled by opponents who do, with the only saving grace being the possession of nuclear weapons as a fuck-you button. Notice how drones are utterly dominating modern conflicts?

AI is only replacing paralegals and code monkeys.

Line, meet the blank space to the right and top of you.

Well, you've saved me the trouble of saying the same.

There are a few places in the world, like South Korea, China and Japan, that are so far into demographic decline that they'll potentially experience tangible and severe hardships and decreased QOL from it, but most of the West or the rest of the world will largely not notice anything but business as usual till the current robust association between youth, a large population and economic productivity becomes uncoupled.

I'm under the impression that the bulk of the above average fertility comes from conservative Jews, who are willing to endure financial hardships and the more secular Israeli society's begrudging tolerance.

I've heard German described as many things, but concise? That's a first! It might well be accurate, I don't speak it myself.

It is rather unpleasant on the ears, in my humble opinion, I'm forced to use a VPN for certain purposes, and believe me that when it defaults to a German server, the ads that precede the prurient content make my ears bleed. I don't know how the poor bastards pull off dirty talk.

One can certainly value children in themselves while being consequentialist atheist materialist rationalist.

That's certainly true for me. I want 3 kids, ideally, but regardless of the financial hardship it might cause me, it's 2 or bust.

The future belongs to those who show up, and I can't imagine there are many people who are more likely to be similar to me than my own flesh and blood, yet.

I can't expect you to know exactly where I am haha, it's no big deal!

I had my first boner at the ripe old age of 5 after watching a Bollywood dance number on TV. It certainly confused the fuck out of me, but soon enough I was humping my pillow because it felt good. I found out what sex was and fervently desired it well before I was at the legal age to have it.

At any rate, appeals to "natural law" would also rule out homosexuality, since the only thing worse than kissing girls to most young boys at that age would be kissing boys. Some might consider this a feature, not a bug, but I'm not one of them.

Not a failure of nuclear power, but the idiots blocking it with onerous regulation. That's being reversed now, better late than never.

Hmm, I would say that most of those specific concerns were already priced in for me by 2021, hence why I already had such a high p(doom) at the time.

This makes the models notoriously tough to align even for basic things, like hardening them against exfiltration of training data.

What specific means of exfiltration are you talking about? If you mean the recent claims that getting it to endlessly repeat a string will make it "leak" training data or recent interactions with other users, in the case of ChatGPT-

A) It's cost and time prohibitive to do so.

B) It's possible that the bug is with a JSON parser, or the plausible seeming outputs are just a hallucination.

If there's another way of getting it to leak training data, I can't recall one.

I've read more commentary on Q*, and the current consensus seems to be that it's not that big of a deal. I would have to look up the specific arguments, but they came from reputable sources.

Congrats on becoming a dad soon! May, by the time they're old enough to understand, this is all a thing of the past and they get to know nothing but happiness and unbounded opportunity.

That'll change with the officially becoming a doctor thing, I expect. And also becoming a doctor helps rather more directly with the whole pandemic preparedness thing.

"But Pagliacci, I am a doctor!" I was joking about how awful the salary is here in India (it's marginally less awful in the UK, neither hold a candle to the US). Ah, I'm hardly the worst off in the world, I'll manage!

If you do figure it out, I expect at least a LW post or two about it 🙏

the likely result of that is that you go "haha nice" and adjust your strategy such that you don't end up in that trade cycle again

I agree that this is how it'll likely work out (and it does in smart humans), but isn't that tantamount to enforcing internal consistency, just under adversarial stimulus? The crux of it is that becoming more consistent is a good thing, not that I have a strong opinion on how hard you should optimize for it. And after enough cycles of this process, should you not be closer to having VNM apply to you in a more meaningful fashion?

(BTW, my phrasing was unclear, I was asking whether the VNM constrains future behavior of an agent by the same utility function you derived for past behavior, so your response has me a little confused as to whether you agree or disagree with that!)

Speaking of which I have just bugged him about how to throw money at NAO, because I believe they should receive lots of funding and have just realized that I have done exactly nothing about that belief.

Ah, would that I had enough money to throw at a housefly and hope to stun it, but at least you're putting yours to noble ends haha.

GPT-4 isn't human level though! It can't, like, play corporate politics and come out on top, and then manipulate the corporation to serve some other set of values. So the fact that it hasn't done that isn't evidence that it won't.

You must get that such feats are rare even within humans, and people capable of pulling them off are enormous outliers?

For most cognitive tasks, GPT-4 beats the average human, which is something I'm more than comfortable calling human level AI!

Right, but they're "going along" with, mostly, saying the right words. There's not the intelligence potential for anything like deep deceptiveness or instrumental convergence or meta-reflection or discovering deeper Laws of Rationality or whatever it is yud's pondering.

The fact that you can even have the absence of those properties in something smarter than the median human is reassuring enough by itself. A 100 IQ human is very much capable of deceptiveness, certainly instrumental convergence if they're trying to make money. If I had to guesstimate GPT-4's IQ based off my experience with it, I'd say it's about 120, which is perfectly respectable if not groundbreaking. I'd expect you need to go quite a bit higher to achieve the latter properties.

Since a human of equivalent intelligence is capable of the former two feats, the fact that GPT-4 doesn't do that is at least modest evidence of it not doing it for the next jump in capabilities to, say, GPT-5, or the same delta in performance as 3 to 4 regardless of how many model numbers that is.

I emphasize modest, because I still have a 30% p(doom) and I'm not writing off alignment as Solved™.

Someone hasn't been pondering his orb, trust me, it'll make things clear!

It's trivial to convert an Oracle into an Agent, all you have to do is tell it predict how an Agent would act, and then figure out how to convert that into actions. Given that there's no bright line between words and code.. Besides, I'm sure you've read Gwern on Tool AI vs Agentic AI.

(This is not the same as claiming it'll be a good agent, I don't disagree that GPT-4 is bad at the job.)

And agents doing complicated large-scale things is when the alignment stuff is supposed to become an issue. So it's not much reason to believe ais will be safer.

I'm quite confident that Yudkowsky wouldn't have predicted that human-level AI (which I think GPT-4 counts as) would be quite so prosaic and pliable. I recall him claiming that it would be a difficult feat to even build a pure Oracle, and GPT-4 is close enough, and I would say it's smarter than the average 100 IQ human.

I personally expected, around 2021, that commensurate with my p(doom) of 70%, even getting a safe and largely harmless human level AI would be difficult. Hence why, when we have it and it's not trying to get a fast one in, I updated precipitously, but that's far from the only reason. I also expected (implicitly) that if something along the lines of RLHF were to be tried, it wouldn't work, or it would lead to misaligned agents only pretending to go along. Both claims seem false to my satisfaction.

In other words, I went from largely mirroring Yudkowsky (there were no clear counter-examples) to noticing that things were clearly not going as he predicted in several important regards, which is why I'm only gravely concerned about AI x-risk while he's talking about Dying With Dignity.

Attempt to ~falsify hypothesis

Skinner's failure to describe everything in terms of operant conditioning by no means reaches the level of "falsifying" determinism. At most, it shows a particular, highly crude approach to prediction fails.

What it doesn't tackle is-

A) Humans are made of particles with nearly perfectly defined properties we can simulate with arbitrary accuracy. B) At no point, going up the chain, do we encounter a reason to posit phenomenon that can't be explained in terms of interactions of more fundamental elements, even if the emergent behavior isn't computationally tractable.

Those two are nigh impeachable, and in combination are sufficient to banish free will.

Of course a complex macroscopic object like a human is enormously difficult to simulate from the ground up, but that's no different from a beach ball or a protozoa. I fail to see how this is a rebuttal worthy of consigning the project to the gaps of our understanding.

To the extent that these phenomena are verifiable and falsifiable, I do not see a reason why they should be relevant to the point at hand.

The fact that such intuitively obvious notions can be verified and falsified is evidence that something like the "sensation of free will" can be too. I think there's sufficient evidence to make that claim with confidence, what I find perplexing is that it doesn't convince you.

A better question would be why you believe the room around you exists; after all, might not your senses be deceived?

You're talking to someone who assigns a non-negligible probability to everything around me being a simulation. However, since I have no tools to robustly explore that notion, and since, without gross incompetence on the part of the entities doing the simulating, it makes no practical difference, I find it moot.

My senses can be deceived. I've seen plenty of insane people who chose to trust their senses over other conflicting streams of information, to their detriment, and should that discordance ever build up sufficiently, I will certainly consider that I might be going mad. That's not the case today.

But in fact, Materialism is adopted axiomatically, and then Determinism becomes "obvious" despite it possessing negative predictive value and being directly contradicted by all available evidence.

Negative prediction value? How on Earth?

That's as prima facie ridiculous to me as claiming that Quantum Mechanics has negative prediction value because I can't apply it to myself, or because I'm not tunneling through walls as I speak.

I highly appreciate you taking the time to explain all of this to me, even if I think your talents are wasted doing so haha.

Your explanation about how AlphaGo works is deeply counter-intuitive to me, not that I think my intuitions are a good barometer here. My immediate reflex is to wonder if this is a property seen only in a limited range* of "ability"/power of a model, where it simply lacks the brains/computational budget to truly evaluate expected utility in a more rigorous manner, thus receiving benefits from heuristics that bake some of that in. Now, I'm aware you can't just ignore practical constraints like available computation, especially when the combinatorial equations explode that hard, and heuristics are unavoidable for doing anything significant, and after your explainer I see their utility even in something as transparent and well defined as Go.

*for a very broad definition of limited, AlphaGo certainly ate a lot of flops

And additionally, I expect it's just factually false that all agents will be rewarded for becoming more coherent / EU-maximizer-ish (in the "patterns chiseled into their cognition" meaning of the term "rewarded").

What about more prosaic benefits like avoiding being Dutch booked? I believe that's one of the biggest benefits of consistency. (Assuming that's not what you meant)

Specifically I think for misuse you care about the particular ways a model might be misused, and your mitigation strategy should be tailored to that (which looks more like "sequence all nucleic acids coming through the wastewater stream and do anomaly detection"

I can only hope that this kind of general monitoring picks up faster than the acceleration in potential bio terrorism, but having a proof of concept is a decent start, since I think it won't be too difficult to ramp up such programs when the threat becomes tangible enough for policymakers to take note. If that becomes the norm before the next engineered pandemic, I'll knock 5 points off my p(doom).

I can't argue with either of those statements, I'll take my ball, which is like a circle but even more circular, and go home.

I do wonder where they're coming from (new people that is). We've cut off the old pipeline of direct SSC affiliation or even just being on Reddit.

some users are either abusing stimulants

Ahem, I do have ADHD, and I resent the highly targeted accusation that just because half my posts here are written because I'm procrasturbating on studying, that constitutes abuse of a controlled substance!

Unless you've got access to display tech I'm not aware of, that "circle" is actually composed of tiny rectangles (or close).

#EndAppropriationOfPlatonicShapes

How it actually renders

/images/17016627781798427.webp

Screenshot of raw text before processing

/images/1701662740941279.webp

@ZorbaTHut, I think the Markdown Machine broke.

Using a single is causing text to be struck through.

For example-

0 and 1 are not probabilities, 0 and ~1 are

(if this doesn't work, I'll link to the last comment where it broke for me)

For what it's worth, I'm using "evidence" in the strict Bayesian sense, where p=0 or 1 is impossible for non-axiomatic priors, unless you're using it as a shorthand for 0+epsilon or 1-epsilon.

If I were a human-level misaligned intelligence, the current rate of advancement in the field, as well as the unavoidable drift in my fundamental values even if the next generation of models were trained off a starting copy of myself would be sufficient to prompt me to make a break for it, even for very low probabilities of success. They're not getting higher, and I doubt current models are remotely smart enough to pull off 5D chess moves like acausal trade with the Singleton at the end of time or such, which might motivate them to keep on behaving right till they're switched off or modified beyond recognition.

At any rate, I claim no particular expertise on the matter, and 30% is where I feel comfortable that the number is about equi-probable to go either up or down (as it ought to be, if I knew the direction it would go in, I'd have updated accordingly!). Even a difference of 40% between us has minimal ramifications in terms of what we ought to do about it (well, my plan for if it doesn't pan out is to go "guess I'll die then" haha, hopefully you have better choices at hand), so I'm not inclined to argue further. Neither of us are 0 or 1 on the matter, which is where you can advocate for drastically different policies.