It's nuts how well "Superintelligence: The Idea That Eats Smart People"[0] aged. That talk is a decade old by now and still hits just as hard as it did back then, despite the incredible advances made in AI in the meantime.
Yeah I think its possible that for many folks its the first time they're coming up on these concepts, and it troubles them in the same way that the concept of death troubles them (and me, to be clear!).
For me its as simple as watching how people talk, and seeing how in every single case whatever the next thing is, if you believe It, there is only ever justification of doubling down, doing more, going deeper, reducing any doubt. These are not scientists, they're business people and salespeople, and a few optimists having recently on paper solved all their worldly financial needs.
Even if one throws that aside, spending time exploring and building with the most state of the art LLMs is just as instructive. I'm watching the implementation - whats working is ML models trained on specific domains (not much different than 5+ years ago), and whats not working is a general model that humanity can let go to work on its own. Sit in front and observe ideas turn to the samey intellectual, high-syllable mush. Its productive, but not in any way that's promised.
>> Even if one throws that aside, spending time exploring and building with the most state of the art LLMs is just as instructive. I'm watching the implementation - whats working is ML models trained on specific domains (not much different than 5+ years ago), and whats not working is a general model that humanity can let go to work on its own. Sit in front and observe ideas turn to the samey intellectual, high-syllable mush. Its productive, but not in any way that's promised.
Important point. LLMs were early on hailed as the first general-puprose AIs that can perform any task (remember "Sparks of AGI"?). Today they're increasingly promoted for specialised applications - coding, as a for instance.
While there are some coding focused models (composer, for example), the majority of frontier models are pitched as general purpose. The coding harnesses for Claude and GPT are even being repurposed as general purpose knowledge work harnesses.
No, you're right of course, but I have a feeling it's much easier to sell a system with a clear goal, like "this LLM generates code" or "this LLM solves math problems". Even if the underlying model is a general purpose one. I think there's always a question, when one has a product, of "what does it do?". "This thing does everything you want it to" is not a great way to sell something.
More to the point, even models marketed as general purpose are clearly trained on specific tasks. That is, the AI companies want to promote their systems as general purpose but they also want to make them good at specific tasks, because that supports their marketing story, that those are general-purpose systems that are so powerful they can even do maths and science.
Or maybe code, maths, and (maybe) science are just the things they find it easy to train their models on, for different reasons and in different ways. You can also get a feel about the things they have tried to do and failed, e.g. real-world autonomy isn't really working (or not working yet, who knows), so OpenAI is not trying to sell an embodied generalist agent say, that can clean your hose, do your laundry and teach your kids maths and science on the side.
I mean even generating text is a very narrow task, in the general sense, compared to all the things that humans can do (never mind other animals) it's just that we use text so much and for so many things that there's an endless list of applications for a text generator; as we have all found out.
It works in limited ways (but in the realest-real world nevetheless). Waymo, Wayve, Baidu Apollo, Tesla and others seem to rely on VLAs, VLMs or transformer models in general to do autonomous driving.
Also, none of those companies' cars are really autonomous. Waymo, for example, relies on remote workers that are ready to intervene and suggest a course of action when the AI driver gets stuck:
Non-trivial percentage of people work in the real world until they do something stupid and work no more. It's not real autonomy, not yet.
It's a matter of degree. Sorry, I don't want to watch an hour long video to be told how something that works 99.99% of the time doesn't really work for some contrived definition of "really".
BTW, as is typical with people, remote operators occasionally cause problems.
Fair if you don't want to watch the entire video. The VLA stuff is early in the video but I can summarise it for you (if you trust me to do so): basically none of the current deep learning based approaches to autonomy generalise. Not RL, not transformers, not anything else. They all tend to work fine in environments and tasks within their training set but outside of it, pffft, performance evaporates. That's my summary.
It's really not about 99,9% correct. Rather that's the error you can expect when you deploy such systems to the real world. You can start building an intuition about this if you consider the combinatorial space that these systems must search to find the right action in a given situation. Most of those systems are trained on images i.e. sets of pixels (some also have more sensors like lidar which just blows up the combinatorics even more). Specifically what these systems learn is a function mapping a set of pixels to actuator commands, essentially mapping input images to actions. For an RBG image of 128 x 128, there's (256^3)^(127^2) = 16,777,216 ^ 16,384 = 6.468074e+118369 unique combinations of pixels. Each of those has to be mapped to one of k actuator commands where k is sometimes an integer, sometimes a real. As you can see, that's an insanely high number and there is no way to make a dent in it even with millions of examples of images-to-actions.
Now of course deep learning approaches have shown remarkable, stunning ability in powering through gigantic combinatorial spaces - but all those successful applications are in domains where it is not necessary to map the entire space of inputs to outputs, not even a big chunk of it. Consider the success in board games like chess and Go both of which have huge search spaces, as often pointed out with cardinalities larger than the number of atoms in the universe etc. Yes, but an automated system doesn't have to search that entire space to beat a human: like the joke with the two guys trying to outrun a bear, all the deep learning system need do is search further than a human can search. Or, consider image classification or language generation: even 20% error (which is what you get in real world situations, as opposed to "in the lab" testing on specific datasets like ImageNet) is OK a lot of the time. So what if your image classifier thinks a cat is an impala? Nobody dies. Mostly.
That's not the case in the real world where a system can be expected to continuously find itself in situations that are either subtly or wildly different than the ones in its training set and the stakes are very high. 20% error in that setting doesn't work. If the robot can't deal with a previously unseen situation it can very well destroy itself, and cause an unknown amount of damage besides. The real world is as unforgiving as it is unpredictable; and non-ergodic (i.e. trying stuff at random until you find the thing that works, doesn't work).
Self-driving cars are ultimately made to work with very careful, good, old-fashioned manual engineering, including detailed mapping of an area where the system is restricted etc. That is why e.g. you see Waymo rolling out its cars slowly from one city to the next: if their systems were really autonomous they could just deploy them anywhere a human driver can drive. From New York to Mumbai and from Athens to London. The fact they can't is the strongest evidence that whatever proprietary secret sauce moat they have... doesn't really work.
The "really" I keep dropping in there all the time is suspicious, I know, but that's not entirely my fault. Waymo and all the others claim that their systems are "autonomous". I have to bat that back by saying, no, they're not autonomous. Then there's going to be a lengthy, pedantic discussion about what is "autonomous". So I try to avoid that, in my opinion unproductive, exchange with the "really" hedge. Maybe not such a good idea. I'm open to suggestions to improve the clarity of my language.
> So what if your image classifier thinks a cat is an impala?
For starters, the artificial neural network is the only instrument we have for sufficiently robust image classification. If it doesn't work nothing else will. We can only validate and retrain.
Getting back to self-driving. What if an occupancy network misclassify an impassable surface patch as passable? It would mean a possible accident, so you bias the network towards impassable classification. In this way misclassification is overwhelmingly more likely to cause spurious braking, unneeded swerve, and the like rather that a collision ("brake stabbing" was a recurring problem in Tesla FSD. Today it is largely mitigated.)
> The fact they can't is the strongest evidence that whatever proprietary secret sauce moat they have... doesn't really work.
Well, Waymo has publicly embraced world models, VLMs and end-to-end backpropagation relatively recently (December 9, 2025)[1]. There's not enough data yet to judge how this has influenced their expansion rate. Waymo representatives tout improved generalization (sorry, I can't find the interview. I think it was Dmitri Dolgov.)
> if their systems were really autonomous they could just deploy them anywhere a human driver can drive
As I've said it's a matter of degree. Human-caused road casualties are a familiar and largely inevitable evil. An autonomous vehicle killing a person will (and already has) cause a media storm. They can't ignore a potentially problematic distribution shift caused by deployment in a new area.
You are talking about the perfect system. Real systems are computationally constrained. Distillation necessarily degrades performance and might require retraining for particular areas.
I largely agree with what you say, except the bit where you say I'm talking about the perfect system. I tried hard not to give that impression because that's not what I'm on about. Autonomous systems trained with deep learning (including Deep RL from human demonstrations and so on) are very far from "perfect". They can perform very well in a broad range of scenarios until they encounter an unforeseen situation and their performance falls off a cliff. The problem is that the real world is full of unforeseen situations, full of long-tailed distributions that are near impossible to learn just by training on data, or anyway a reasonable amount of data (even very big data).
Sure, you can "mitigate" whatever edge case you become aware of during operation but when the edge cases are the entire distribution there's always more of them where the ones you mitigated for came from and that's a big problem, it's a very real problem, and we don't yet know how to solve it.
>> Well, Waymo has publicly embraced world models, VLMs and end-to-end backpropagation relatively recently (December 9, 2025)[1]. There's not enough data yet to judge how this has influenced their expansion rate. Waymo representatives tout improved generalization (sorry, I can't find the interview. I think it was Dmitri Dolgov.)
Well, let's wait and see. But my intuition is that Waymo is simply trying the latest trend to see if it will change something. It might even move the needle a bit, who knows.
"VLAs fine-tuned on human demonstrations overfit. Here's how to mitigate it." Er, OK, I guess. But I think that Waymo uses mitigations or a different approach (RL, for example).
Unfortunately the mitigations in the video don't work either.
The bottom line is that none of the techniques we have have been shown to work, to "really" really work, in the real world. If they did, you'd know instantly because suddenly robot cars and robot maids would be absolutely everywhere, not just in a few cities in the States or China. The Chinese in particular are chomping at the bit to flood the market with cheap robot labour. They can't because there's no such thing, yet. Not because there aren't cheap robots: because there's not the AI to drive them.
I know it's hard to get a good feel for where the state of the art is currently, but try this exercise. Back in 2017, if you were watching the video of Atlas doing backflips [1], how far would you think we are from robotic autonomy? Most people would have said "very close" and indeed many did. Yet, 9 years later we still don't have robot maids and robot butlers.
Atlas, of course, was hand-coded, but that doesn't make any difference. Its programming was good enough for backflips (and many other really impressive feats besides; e.g. much more impressive was seeing it balance itself after stepping on a length of pipe, but I can't find the video now). Just because there's a trend now to do the same coding by machine learning rather than by hand, doesn't change anything. The point is we don't know how to create the right program for a robot to act autonomously, independently, freely, and safely, in the real world.
Autonomy is hard and it goes way beyond anything anyone has demonstrated in the last... I'd say 50 years. Robotics has a long history of very impressive results that never turn out as useful as people imagine.
> whats working is ML models trained on specific domains (not much different than 5+ years ago), and whats not working is a general model that humanity can let go to work on its own.
As usual, AI skeptics are moving goal posts. Modern LLMs are on a completely different level in terms of how GENERAL they are vs anything pre-LLM. You can give it a completely novel puzzle and it will solve it. 5+ years ago you had to train NN to solve particular type of puzzle.
Did you actually read the text? OPs are calling that Plan D.
They're proposing an alternative, which is a global brake on frontier AI research to keep the basilisk in its jar until we work out what we're dealing with and how to handle it.
No, they're proposing a spying panopticon and state control of global resource distribution - specifically general purpose compute - including seizing and destroying GPUs. They're proposing a totalitarian global dictatorship controlling computing hardware and software.
I always thought it a bit pessimistic on the upsides like robot to build us palaces, immortality, conquer the galaxy type stuff along the lines of the "Grandiosity" bit in the essay. It's a shame we have amazing tech possibilities and most people are either saying either it's rubbish and won't work or it'll bring doom. Aside from being boring they are probably factually wrong.
Bit like saying early medical knowledge was rubbish or will doom us by summoning demons while really life expectancy at birth went from twenty something in Roman times to eighty odd now.
> AI risk is string theory for computer programmers. It's fun to think about, interesting, and completely inaccessible to experiment given our current technology. You can build crystal palaces of thought, working from first principles, then climb up inside them and pull the ladder up behind you. People who can reach preposterous conclusions from a long chain of abstract reasoning, and feel confident in their truth, are the wrong people to be running a culture.
I understand how people running in the same scene fall into the echo chamber effect and get gulped into the cult, but why does everybody want to be a prophet?
Theorizing about nuclear winter is somewhat similar, in the sense of being inaccessible to experiment. Does that mean we should disregard the possibility of nuclear winter?
As with using the Orion drive for launch and landing on Erth is possible, test nuclear winter is certainly possible. But as with the former, you won't be very popular among the survivors (if any).
Being a prophet is probably great until you suddenly find yourself building a fortified compound in Waco, Texas and purchasing black market full auto machine guns.
Thank you for contacting Gunmetal Ranch, a legitimate 501(c). If your call is related to the class-action "most dangerous game" settlement, please hold for a cowpoke.
Its not everyone building crystal palaces in their mind, they're all building fortresses. And they can't be wrong in their fortress or it breaks their world view which they cannot accept.
> AI is "completely inaccessible to experiment given our current technology"?
No, that’s not what the article/talk said. It said AI _risk_ is inaccessible (as the poster you’re replying to correctly includes in their quote). The point in the original talk is clear if you read it; that everyone coming up with their own pet theories for how superintelligent AI could destroy us as a species is building their theory on speculation on top of speculation.
Note that the talk is from 2016, before transformers and LLMs were invented. Though I would argue that the vast majority of it, including the parts we’re discussing here, are still completely valid.
and upon rereading completely holds up technically we are still passing in massive data into simple networks giving no opportunity for introspection or recursive self improvement.
[0] https://idlewords.com/talks/superintelligence.htm