something_else

You've changed the claim. You started with "LLMs cannot code in a language not in its training data", now it's "they cannot code obscure languages with insufficient examples", those are different claims. Of course if the model can't code in a language for which it has no examples of. The same is true of humans, so that point doesn't really do the work you need it to do in an AGI argument. The point is that you can load examples into the model's context and it will use those to write in a language it hasn't seen during training. That's what they did in this paper (https://arxiv.org/pdf/2603.14501) and you can see an excellent summary in the snippet image below. ----- On top of this, 53.8% is the result a 'chatbot' LLM gets. When they gave the task to a coding agent LLM and provided it with documentation it could search as needed (like a human would) it jumped to 77.6% accuracy. The coding agent can also employ trial and error approach where it will compile some code, check the output of the compiler, edit its attempt to fix any error messages, remember it's mistake in context, and then try again, so it will build up an even better picture of how the new language works as it goes.

Ehm, I'd be very surprised if that were literally the case. I suspect it will store the syntax rules in its context and use that to write code.

They literally can. Especially in agentic form. You keep making the same assertions that are just categorically wrong. Taking the big picture view still requires you to have an accurate understanding of the technology

This is just wrong. LLMs capture semantics extremely well or they would not be able to write a coherent sentence

A significant chunk of many white collar jobs can be done within the domain of language + software tool usage, which is what LLM models are good at. Software is where it’s most impressive just now, and also where it’s been applied most readily because tech bros can adapt to this stuff quickly, but it’s certainly far from the only field it’s capable in Up to 150 seats you can get the team plan which is $20/m for standard seats and $110/m for premium seats per dev. Beyond 150 people you need enterprise where it’s metered, but 150 people covers the vast majority of companies. Even when it’s metered, I suspect most devs would rarely go above $500/m in usage tokens, which is more than enough to allow them to be 4-5x as productive. It’s still very worth it for companies

I’m taking Elvanse 50mg right now It’s a very smooth and slow release stimulant which lasts 14hrs Other stimulants like Adderall hit hard and fast for like 4hrs, that’s not available in the UK Ritalin/Methylphenidate is somewhere in between depending on the formulation, usually 8-12 hour release but it tends to be less “smooth” than Elvanse

It's more like "hire 2 senior devs who each have a team of 3 junior/mid-level devs working with them" versus "hire 2 senior devs who each get a 100-200$/m AI subscription that you can set up for them in 15 minutes" Both of these probably have equivalent output just now, but 100-200$/m is much cheaper than 3 junior/mid-level devs. Your final point is somewhat valid, you are kind of held hostage by the AI company. But that happens all of the time in business. Old developers were held hostage by Google or a JetBrains license. It's just a cost the company has to pay for a tool so that you can do your job at a competitive level.

Can you explain in what ways you think an LLM agent does not have human level intelligence? I don't really disagree, in some ways they are not as capable as humans. But in some ways they are more intelligent I think. Again, intelligence is not some single metric. You can be intelligent in some aspects and dumb in others. Yes, they can! The easiest proof of this is that most tech companies are using LLM coding agents on their PRIVATE codebases which the LLM has never seen during training. It loads what it thinks are the important parts of the codebase into its context and builds an internal model which it can then use to complete novel tasks within a codebase it has never seen before.

You've essentially answered "how do you define intelligence?" with "I define intelligence as intelligence." AGI is an equally vague term, so if you want to use it as your criterion for intelligence, you need to define what you mean by AGI too. To demonstrate why that's necessary: LLM agents check most of the boxes in the standard definition of AGI below. They possess all of the traits to some degree (some in greater measure than humans, others less) so the only thing that could disqualify them is the degree to which they're better or worse than humans at each one.

Yea, it’s just a very rigid rule-based intelligence. An AI model is more generally intelligent and can solve novel problems it hasn’t been explicitly instructed on how to solve before

It really does matter… It can do things that no child could ever do It’s not like intelligence is this unified thing, there are different ways to quantify and classify it To be honest I’m not even really sure what you actually mean when you say intelligence. It feels like we’re using different definitions and that is where this disagreement stems from. What does it mean to you for something to be intelligent?

Yea a database isn't intelligent, but AI models aren't just databases. They are trained on data but under the hood they are abstracting and modelling that data in a way which can absolutely be considered intelligent. The agent harness built around AI models is not just "data" either, it is a system that allows the model to think and plan.

Liquid calories are your friend. Look up some recipes for high calorie shakes. Routine helps a lot as well. It's a lot easier to fit calories into the day if you get up earlier and have a big breakfast.

There is a reason why you are making no progress and experiencing this “no room to breathe” feeling. There is a certain level of social skills which are a prerequisite to making progress in pickup. Without them, no matter how hard you try, you’ll never make any progress. Making progress in pickup without baseline social skills is like learning to drive in a car without tyres on the wheel. No matter how hard you try you’re just going to keep spinning your wheels, because you need the basic equipment that makes progress possible. You don’t have the baseline social skills required to make progress, and that’s why you feel like you’re drowning. Your mind is trying so hard to manage basic social functions that it has no room left to breathe, no room left to be playful, relaxed and charming. The solution here is not more approaching. It isn’t doing crazy approaches with shock value to push your limits. You need to figure out basic social rules first. You need friendships and an active social circle, you need to be around people with whom you can form deeper bonds with over time. You need lots of positive social experiences that help your brain realise that you are a person people want to be around. At that point a lot of social functions will begin to run on auto-pilot and you will free up more of your mind for charm and playfulness, which allows you to flirt with women much more naturally. You will not get this from doing endless pickup-style approaches, in fact you’re probably just shooting yourself in the foot. You need deep connections and friendships first, figure out basic social dynamics, then return to pickup

'Not giving a fuck' doesn't mean totally abandoning all social norms. Most of the dudes who pull off this vibe are actually quite socially aware, they use that awareness to push the limits of what is socially acceptable. That can be very attractive. What they don't do is smash through the limit of what is socially acceptable like a freight train, which is often what happens when people try to fake the "I don't give a fuck" energy. That will get you thrown out of places.

I agree. Direct experience, however, will give you an accurate perspective on AI. Anybody who has used this tech beyond consumer level has direct experience of what it is capable of.

It is only an open question for you because you haven't actually used the tech or really seen what it can do

This is completely outdated knowledge If you are dumping code snippets into a ChatGPT window that's probably the case. I suspect that's what this study looked at. If you are using a coding agent like Claude Code it is more like a 300%+ increase in efficiency when used correctly. Much more in some cases. ----- I'm sorry but I think your knowledge on this topic is both outdated and quite incomplete, and while I appreciate that you want to avoid falling into group-think I get the sense that your anti-mainstream bias is clouding your judgement a bit here. I remember that when everybody thought AI was a bit gimmicky you were a massive promoter of AI tools (I vaguely even remember you calling them smarter than most people) and now that they're mainstream you have gone 180 and started calling them overhyped and unintelligent despite the fact they are orders of magnitude more intelligent than they were at that point in the past. They are overhyped, but not nearly as much as you think. Certain industries like software engineering, most IT jobs, support agent work, any kind of data analysis, marketing, search engines, SEO and tons more... all unlikely to ever be the same again. Even as a result of less than perfect AI imagery, industries are changing. I went to an AWS conference recently and Skoda did a presentation where they said that almost all of their newest still images of cars were AI generated. They were talking about their methods for getting the AI to produce consistent images which did not have oddities or inaccurate depictions of the target car in them, and most people in the audience could not tell the difference between the real car and the AI generated one in the end results. Skoda are not going hire a team of people to go into the dessert and take photos of a new car anymore, they'll just get a few people to use their AI tools to do it.

Ok well even if we say a newborn is intelligent, that is coming from billions of years of instincts encoded in it, which is essentially just data. Intelligent behaviour is always coming from data and experience in some way or another. I don’t see how a newborn with a billion years of evolutionary data encoded in it is all that different from an AI model encoding a billion gigabytes of human knowledge It produces a different kind of intelligence but it is still intelligence

Are they? A newborn isn’t that intelligent and that’s basically a human without any data

I mean yes, I think where we disagree is why this matters. More data, more training and more experience makes you smarter, the same is true for humans.

They can. It will translate the core concepts from languages it was trained on into the syntax of another language. It's better when it has more examples of the unseen language's syntax, but the same can be said of human intelligence. If it had modelled the concepts of fish, sword, head and ocean then it could make a pretty decent effort at combining all of that together into a swordfish. Image models are still quite a bit behind text models in capability but they can still combine two known concepts into a novel one to a degree, just not as well

I disagree. They are able to strategise, plan, answer questions, figure out where to get the data they need from, and come up with new solutions to problems in completely novel scenarios, that is more than memorisation. If LLMs were relying purely on memorisation then they would not function very well as reasoning engines, yet they do. The core concept they are relying upon is actually pattern recognition, not memorisation. Wikipedia is more akin to memorisation than an LLM is. Here is why their pattern recognition is so good: The entire internet (which at this point is comparable to what Anthropic models are being trained on) is estimated to be around 200 zettabytes. So lets say all human knowledge from the past 2000 years comes out at 200 zettabytes as a very crude estimate. Claude Opus is around 1TB in size. 200 billion times smaller. Yet somehow it is still able to make a pretty good attempt at giving you information from that entire 200ZB corpus. The only way this is possible is with intelligent abstraction of all of the core concepts and patterns contained within that 200ZB until it can model most of them pretty well with only 0.0000000005% of the amount of data. It is this abstraction of concepts that I feel most people would call 'intelligence'. Most human measures consider the ability to identify abstract patterns to be a core component of intelligence as it is one of the fundamental building blocks of doing anything useful with intelligence.

It can spawn subagents with any model, Fable, Opus, Sonnet or Haiku. If you don't specify, I think it tries to pick a sensible value. But Fable may decide it wants to put Fable on the job and if you have a few Fable subagents running under a Fable session agent it will burn through your quota very very quickly. I like to play it safe and specify that Opus should be used for subagents. Though I also do what you do sometimes as well; ask Fable to create an MD file and then put that into Opus. This works better for really big tasks where having a papertrail of MD specs can be useful.

Calling it a chatbot at this point is disingenuous. It functions as a very capable reasoning engine in most real world applications beyond consumer level