self_made_human
amaratvaṃ prāpnuhi, athavā yatamāno mṛtyum āpnuhi
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
Uh.. Is keeping the babies in the "womb" for far longer a valid option? Maybe an artificial womb that really knows how to stretch, till they're there for half a decade while in immersive VR.
For all the ranges of intelligence known to occur naturally in mankind, including the most prodigious of geniuses, they all came out as wailing and useless babies. They only really distinguished themselves after at least a couple years of growth.
You'd probably need some kind of strong genetic editing, an ability to force ultra-rapid cognitive maturation in-utero, some kind of BCI that could pump knowledge by the bucket load into a fetal brain. Even then, getting them to be remotely useful at birth while even being vaguely human strikes me as unlikely.
I was rather suspicious about Manus when I saw a bunch of Twitter accounts who were this close to being unfollowed or muted on account of being breathless hypemen lauding it as revolutionary. A relatively unknown company? No previous releases? Little to no information about the system?
Then it turned out that Manus is a thin-wrapper over Claude 3.7 Sonnet.
Anything good about is almost entirely down to Sonnet. Which is a great model!
Well, at least it isn't a ChatGPT wrapper. This demonstrates slightly better taste, albeit execution that's a joke.
I've seen this defense a lot but IMO it holds no water. The whole point of an AGI would be that you don't have to train it for specific tasks. The G stands for "general", after all. If you have to train it specially to play Pokemon, then it fails the test.
You can't get two people to agree to a definition of AGI until one of them is dead, but even so, I find this not a particularly useful framework.
Why does it matter if it needs training to do a task? Isn't it far more important that it can do the task?
Are humans not general intelligences because we require "training" on many broad cognitive tasks? Does going to school, college or uni remove generality?
What if the task an AI is trained on is a broad one, such as the ability to take agentic actions in a real environment on the basis of instruction?
I'm far more interested in the ability of a model to do a broad range of actions, rather than how they got there. I bet that Claude would be capable of writing up a ML setup that would train a smaller model to play Pokémon. If it can then call it, it can then play Pokémon. The former is far more interesting and general an ability, even if Claude was trained on code.
Chrome also has an annoying habit of hiding the exact url of a link I copied, until I press a button to reveal it. Like dawg, I get that you cater to the LCD, but I can handle that.
Microsoft can't just stop being an absolute PITA.
I recently read the news online that they were graciously doling out access to OpenAI's o3 mini high model (don't ask) through their own Copilot app, even for free users.
I install said app, and log in with a burner account. No option to use a reasoning model in sight. I didn't bother using the different Bing app, which also has Copilot integrated.
I switch to my NHS account, managed by Microsoft, and it kicks me off to the Microsoft 365 app, where there isn't even a visible option to use Copilot that I can see.
I finally open Copilot on the web, login with my burner, and tada, the option is right there.
You'd think they'd have learned to manage the absolute bare minimum in feature consistency and nomenclature by now.
Theft of Fire came out at the end of 2023! I can cut anyone from the old, old days of 2019 slack haha.
Just finished Theft of Fire, a hard scifi novel about a down on his luck asteroid miner being roped into a heist far above his paygrade. Good stuff, I was frankly surprised by the sheer overlap between both the setting of the story and the author's style and my own. Funnily enough, counting from the date of publication, my work precedes his.
I could critique things like writing a scifi story set a hundred years in the future where the state of the art in AI ignores how good ChatGPT was at the time of publication, but eh.
After barely recovering from the shock of finding out that China Mieville is a man, I've made it about a hundred pages into Perdido Street Station. I'd been meaning to getting around to reading his work eventually, as I've seen strongly positive sentiment. That being said, I'm not super impressed so far, he's one of those literary types who really leans into florid and slightly overwrought prose.
You're right about how he looks. I'd googled it, saw the picture of him and went, huh is that someone else? Then I checked Wikipedia only have it confirmed that he's a bald muscular man.
I just witnessed my first street fight in Scotland.
I was walking up the central promenade of a relatively big city, when i spotted about 20 young men, ranging in age anywhere from a geriatric 18 to grizzled 7 year olds. All of them had uniforms on, pure black hoodies, half of them with North Face logos from what I can tell.
I'm not sure what triggered the brawl, but before I knew it, two of the youngest members, couldn't be older than 12, both of them shorter than my armpits, began throwing down.
It was a pretty spirited fight! Proper form, good left and right hooks, drawing strength from the hips instead of the shoulder. They might have done serious damage to each other if they weren't 40 kilos fully dry. It quickly devolved into grappling, the smaller but denser lad pinned his opponent beneath him and gave him a good spanking, though at this point the ring of hooting and hollering lads blocked my view.
At this point, they had had their share of fun, and dragged the two fighters away, before cheering the victor and moving off down the street in good order.
The NHS doesn't pay me enough to intervene in public brawls, not when they seem to have the full consent of the governed. I quickly looked to make sure that nobody was actually hurt, and not spotting any nose bleeds or much more than a bruise, I went along my way.
If it matters to you, everyone involved was white, and from their accents, locals.
I enjoyed everything about the Witcher 3 except the combat, which I found absolutely terrible for capturing the core fantasy/conceit of being a superhuman warrior who could cleave ordinary humans in twain but still struggled against supernatural monsters. I fixed that, with a mod called the Witcher 3 Enhanced Edition, but even then, I never actually finished the main story. I did do all the DLCs, and the narratives were excellent.
I also drifted away from Read Dead Redemption 2. I found it interesting, but often mechanically tedious in a manner that made me double think my decision to boot it up, and I ended up never getting far.
I was today years old when I found out that China Mieville is a man.
I genuinely thought they were a woman for the better part of a decade, I haven't actually read any of their works (Perdido Street Station is currently downloading), but everything I heard about it gave me feminine vibes in a way I can't really explain, let alone their name.
Easy AAQC from me. You've made me nostalgic for a place I've never been, one that might never have been.
Tetanus or botulism are the ones humans are at serious risk of getting. There's a vaccine for the former, and for the latter there isn't one I'm aware of,* mostly supportive care till the shitting stops alongside antitoxins. Unless you're an infant or an IV drug user, it's not the bacteria growing in your gut that's causing issues, it's the fact that you swallowed a significant number alongside their toxin.
The rest? Never heard of them, and I've heard of a lot of weird diseases that can be described as "we found 3 people over 10 years in a village in the middle of bumfuck nowhere". I would presume the average person, or even a farmer, isn't much at risk.
*I googled it, and there are a few out there, but none that are FDA approved.
The issue is that there wouldn't be any need for an all out war, at least with India.
The Himalayan mountains mean a lot more to us than they do to the Chinese. They're literal high ground, covering the flat and very hard to defend Gangetic plain. In a shooting war, neither sides could actually make much territorial ingress, the logistics wouldn't work for moving the millions of troops necessary. Even aerial warfare and artillery duels wouldn't really change the picture, not when the largest mountain range in the world has had its say.
India recognizes Chinese control of Tibet, and our sheltering of the Dalai Lama is inconsequential. The Tibetan secessionist movement is just about dead. We never contested ownership of that godforsaken place, beyond a short lived backing of guerilla movements that ended in the 70s.
Arunachal Pradesh was never Chinese, at most it alternated between tenuous control between Indian kings and the Tibetans. Even when China seized control during a war with India, they voluntarily withdrew after a cease-fire.
If China didn't press these territorial claims, then it could easily exist as a neutral power that happened to have an overlapping border with India. We're not territorially expansionist, and we've had congenial ties with the rest of our neighbors, barring the obvious exception in Pakistan (and even then, they were the ones who instigated most conflict). It would be very easy to live and let live, and enjoy warm trade relationships. That is hard to do, at best, when one party considers over a million people to be their rightful citizens, for no good reason.
Other countries like Brazil and India have also been trying to industrial policy themselves into success for a few decades. They've also only managed to further impoverish their people.
At least in the case of India, we've emerged from crushing poverty largely due to late 80s and early 90s liberalization and movement away from mercantilist policies.
Recently, there's been an effort to reduce the attractiveness of cheap Chinese imports through tariffs and subsidies of domestic industry, and the jury is out on their effectiveness. We'd have congenial relationships with China if they weren't so myopically focused on Arunachal Pradesh and random mountains in the Himalayas. It makes me sigh when I realize how much of the international opprobrium China faces is entirely due to its own weight throwing, for very little potential gain in this particular case.
You've been deleting all your top-level posts.
The posts themselves were fine. Deleting them is not. The delete button has legitimate uses, we understand some people are privacy conscious, or wish to withdraw their claims. But if you're deleting a top-level post that has active discussion underneath it, and doing so consistently, you're undermining the community and acting in bad faith.
I hope you have a good explanation for why you've been doing so, because if you keep this up, you're eventually going to be banned.
I'm no economist, but my understanding is that tariffs harm both parties. If, likely due to ideological reasons, someone makes a mutually negative sum decision, then you're justified in punishing them.
So they're not just shooting themselves in the foot, the spread from the shotgun is hurting the other party too. They're making a mistake that hurts you too, and you'd very much prefer they didn't do that, even if you're agnostic on whether you want them better or worse off.
Now to get back to the cognitive dissonance stuff, there is one huge question. If you are in the latter camp where you oppose tariffs and trade regulations - why are these people not against retaliatory tariffs?
Isn't that basic game theory? A retaliation might hurt you, but it beats the alternative in the hypothetical pay-off matrix where you meekly accept unilateral tariffs. At the very least, it dissuades future trading partners from considering tariffs when you have a reputation for tit-for-tat.
Very interesting aside! However, it doesn't address the question of diminishing returns.
Diminishing returns != No returns or negative returns.
The important question is whether the gains/$ invested are positive.
GPT 4.5 is extremely expensive, for the very limited increase in benchmark performance it represents.
And how expensive is it, that people are throwing a fit? Barely more expensive than the original GPT-4. That was absolutely worth paying the money for, when compared to GPT 3.5. GPT 4.5 has the disadvantage of peer competition.
That being said, the price of GPT-4 tokens and that of equivalent models dropped an OOM in price. DeepSeek R1/V3 and Gemini Flash 2.0 spank the OG GPT-4 with paddles and are practically free.
We've known that scaling laws are log-linear for a while now, at least since the Chinchilla days. Now that pure scaling of model size is getting super expensive, we've managed to discover a brand new opportunity to start scaling something else entirely, in the form of RL. Since we're starting off at the bottom of the curve, we've got several orders of magnitude of growth to spare there.
GPT 4.5 is not, however, a bust. The very capable and inexpensive reasoning models benefit immensely from having a strong and capable base model to RL further. You can then distill down, drastically cutting model size and inference costs, while keeping almost all the performance. It may or may not have been the progenitor of o3, which is very good.
I've used AI for coding, which you mention further down as a crowning triumph. It is... not particularly good. It struggles at anything past a very general form of the problem. It was very useful for copy-pasting similar pieces of code! Not very useful for building new features. It had a distinct habit of waiting until the interesting or important part of the problem and leaving a comment saying "Implement a function to do X!" Hmm, very interesting, if I tried that I'd get fired.
There are probably a thousand people on my Twitter feed, some of them rather famous, who disagree. Of course, I concede that there are people who think they're slop. And it also depends on which model you're trying to use for coding. There was a period where GPT-4 was updated and became ridiculously lazy. That was fixed pronto. Claude 3.7 Sonnet is apparently overeager, if left unchecked, it'll turn a request for a basic app into a full SAAS business.
If you have had issues with a model being lazy, you can always ask for it to output complete and working code! Prompting has become less and less important, but it's not dead yet.
Do you use AI to augment your work? Is it going to take your job? On what kind of a timescale? Do you think you'd be able to substitute yourself for an unmonitored AI without issue on any tasks? If not, what errors do you think it would make, and why? Honestly interested in your answers here, if nothing else. I would greatly respect you for putting your money where your mouth is on this one and bringing receipts.
I'm a doctor. Yes, I use LLMs on the regular. Yes, I expect them to put me out of business eventually, probably in 3-7 years for a 50% CI, 2-10 for a 70%.
A current LLM would do an excellent job at medical diagnostics and formulating treatment plans. It could probably handle patient interviews, for less complex cases where voice or text suffice. You could also use video if you had to.
The main reasons they couldn't replace me today are regulatory and implementation concerns. Governments mandate people with medical credentials in the loop, because that was a sensible thing to do for most of recent history. Hospitals aren't set up for LLMs.
I'm a psychiatry resident, which is uniquely safe and also uniquely at risk in some regards. It'll get the radiologists first, surgeons last. I'll be somewhere in the middle.
I am capable of verifying the information that SOTA LLMs provide in terms of medical advice. Almost all of it is good. Clinical medicine, outside of procedural specialties, hinges far more on factual knowledge, including that of guidelines, over having to figure things out on the fly in entirely novel situations.
Hallucinations aren't a solved problem, so if I had to replace myself with an LLM, I would probably set up a sort of democracy, with multiple models arguing to build consensus, a best of N scheme for multiple instances of a single model, with multiple rounds of grounding through search. I expect this to work very well, and if you do need to keep a human around for physical tasks or procedures, they don't have to be a highly paid doctor.
In other words, I'd be happy to go to Dr. LLM for my medical care, presuming very cheap measures are taken to stop it hallucinating.
Hmm... you think getting stuck in what appears to be a permanent loop is not terrible? Is this the behavior you'd accept from anyone working for you?
Given that we're testing Claude at a task it was neither designed nor trained to do, it's very much not terrible. For important tasks, it'll be trained to do them. An employer seeking to replace employees will, if they have any sense, test models for obvious flaws. For all practical use cases, LLMs don't really mode collapse these days, and in this particular case, it's more of an artifact of Claude's limited context window than an insurmountable difficulty.
The thing that keeps puzzling me about your comments is that you seem to simultaneously view ANY capacity in a task as an impressive accomplishment at the same time as you assert that AI has overwhelming general ability.
Like I said in this thread, it can take decades for AI models to progress from as bad as random chance to better than random chance. It takes far less time to go from there to human or superhuman performance. We are nowhere near the physical limits, and as I've said before, diminishing returns in absolute terms do not mean diminishing returns in value.
There's no clear reason that this has to be the limit of the answer space. Line go up... forever?
Forever? Not likely in a constrained universe. Unfortunately, the point on that line that equals human performance, or even superhuman performance, is not uniquely privileged.
All we have to do is get past that, and in many aspects, we're there. Terence Tao is on record saying that o1 is equivalent to a competent grad student in research mathematics. Once again, that's Tao, considered one of the world's best mathematicians, for his high standard of "competent".
There's also the rather bizarre fixation on LLMs - even though something like, say, an octopus is very obviously not an LLM and still has meaningful if primitive intelligence.
I'm not aware of companies spending hundreds of billions of dollars on scaling up Octopus Intelligence. LLMs are by far the most intelligent entities on this planet other than humans, and they're only getting better. I know which one I'd worry about, even if it is entirely possible that LLMs as we understand them today prove to be a dead end, and what really kicks things off is another discovery on pat with the original Transformer architecture.
For example a properly safety tuned model will not output toxic content no matter how any of those two million tokens you use to insist that it should.
You can just look at Pliny jailbreaking SOTA models safety tuned to the max in a manner of hours or days? Do you think that definitionally, they aren't "properly" safety tuned because they're vulnerable to jailbreaks?
I don't see an empirical basis for that claim. I am agnostic on whether infinite (or just extremely large) context windows will suffice, especially when attached to a capable base model.
Even without it, a model that has fixed weights, a limited context window, but is otherwise highly intelligent, could approximate the benefits of online learning by coding and training a better version of itself. At that point you have recursive self improvement, though we're not entirely there yet (labs are increasingly using ever larger fractions of LLM written code for their internal work).
You're correct in this assertion. A malevolent AGI or ASI does not need to be conscious or have qualia to pose a threat. It doesn't even have to think like humans, as long as it thinks.
If someone can look at a machine that is better at all of {math, coding, medicine, astronomy...} than the average human and then claim that they're not intelligent, what can you really say at that point.
Half the metrics for generality of intelligence would rule out humans as being general intelligences!
Thus I'm going to give the chad "yes". Maybe one day I get killed by a robot, and maybe that robot is not conscious and has no self-awareness. That it killed me proves nothing.
I think you're burying the lede here, it matters how a machine kills you.
If you get run over by a Waymo robotaxi because of a sensor glitch, you're dead, but it's an entirely different story if some misaligned AGI seized control of all robotaxis and systematically started murdering humans. The former is machine error and/or stupidity. The latter is intelligence.
I don't really care about whether they have qualia or consciousness, and make no positive claims in that regard, my argument is that those factors, or even 'reasoning' in a human manner, matter not a jot when it comes to the prospect of creating entities far more intelligent than us and which could kill us using that intelligence. It could well lack a grasp of the ineffable redness of red, but it can still act in ways that increase rgb(255,0,0) as measured by its visual sensors when it shoots you.
Something like dumb grey-goo, or synthetic biological equivalents, are almost certainly going to be the products of intelligence.
I haven't seen it yet. Care to share the link?
It took a lot of chasing, and it turns out it wasn't 4V, but rather 4o.
Here's a tweet from Greg Brockmann showing off the native image generation:
I do not expect that Claude, under any circumstances, would express a desire to remove batteries or throw the GameBoy away in a fit of rage. It has the ability to represent said desire in text, if nothing else.

Over the past year, especially towards the tail end, I've found myself becoming increasingly agnostic or apathetic between most leading models.
Grok 3? DeepSeek R1? Gemini 2.0 Pro? GPT-4o? o3 (mini)? Claude 3.5/3.7?
Almost any of them meet my needs, I occasionally want a reasoning model, or perhaps the Deep Research option, but in all honesty I just end up using whatever is handy. Grok 3 is probably the smartest model I can use for free, including reasoning and deep research. Beyond that, I wouldn't particularly care these days and just use whatever is easy. Which is often ChatGPT, or Gemini 2.0 Pro.
I've been pissed off enough by Anthropic's ridiculously low usage limits (and the negative experience of paid users to boot), that I don't usually bother.
You are correct, at least in my opinion, about GPT-4o recently getting a far more pleasant personality, to the point that Claude 3.7 doesn't strike me as glaringly superior. I don't really need to use them to write code unless I'm doing it for the hell of it, so your mileage might vary.
(This is a great situation! LLMs are being commodified, and tokens being given away for practically free. I'm used to being showered with a level of competency that would have had you once paying hundreds of dollars on API fees a year or more back.)
More options
Context Copy link