abstractcontrol

When I use nonsense like "real AI" I think you very well know what I mean, and don't find it in any way confusing or ambiguous.

As far as I can see, none of the hard problems of intelligence that we had a decade ago got touched. What LLMs ultimately changed is the way I see language. I've underestimated its power, and I will admit, the companies making these models went far further than I could have imagined with them.

But intelligence and AI have a high hurdle to clear, and it won't be done with LLMs.

Revealing the secrets of intelligence will involve new concepts and understanding that that we don't have now. LLMs will be helpful with the plumbing and as replacement for natural evolution, but not so much natural intelligence.

What's really disappointing about current LLM models isn't that they aren't useful, but that they really tell us much about intelligence at all. The techniques used to make LLMs aren't generalizable at all, and I cannot use them to solve problems of interest in the real world.

That wouldn't be the case with real AI. I want better methods and algorithms that could live up to the expectations, and I am sure we'll get them at some point even if we don't have them now.

The history of AI is people deciding that it really means AGI but then getting enamored with some method of the time and calling it AI, before going back to the original meaning. We're just living through the latest iteration of this where LLMs mean AI, and we're forced to invent the term AGI for what used to be AI.

I'll admit that to the people who don't know how LLMs work, which includes most investors, the lie seems very plausible so of course the CEOs running the companies are going to frame it as such.

The thing that actually matters is understanding the problem domain, and thinking about and visualizing the program I am trying to make, so I am happy to be doing less drudge work so I can focus on what really matters.

I've been doing quant research for the past year, and the amount of time (and tedium) LLMs have saved me is huge.

LLMs aren't even AI, if this amount of disruption is causing difficulty, having actual AI will give these people breakdowns.

You could clock out, but I don't think the top performers ever stop thinking about work. Everything you've written here has to be wrong.

One thing that I understand at my age of nearly 39 now is that success comes pretty hard, and luck plays a large role in it. As programmers all we can do is build and develop our skills, even if the world doesn't validate our efforts.

You're asking yourself what you are doing wrong, but you should be asking yourself what your goal is and focus on that. Is it really to just work for other people?

What can you do for yourself?

Opus has been amazingly useful at answering various statistics question that I had for it, and my current idea is a nested auction market theory inspired model. My biggest discovery is that replacing time with volume on the x axis (on a chart) and putting the bar duration on the bottom panel instead of volume normalizes the price movements and makes some of the profitable setups I've seen described in tape reading/price ladder trading courses actually visible on naked charts. A great insight I've gleamed is that variance should be proportional to volume instead of time or trade count. When plotted, it has the effect of expanding high volume areas, and compressing low volatility ones, which exposes trending price action much more readily. It honestly amazing, it's making me think that I could actually win at the trading game.

Lol, who doesn't hate that?

He was a problem gambler, but I think if we looked at top poker players of today, they'd all have some love the gamble in them. Jesse had godly tape reading skills that allowed him to beat the bucket shops at the start of his career.

After being kicked out of the bucket shops, he should have just become a floor trader and in all likelihood, he'd have had lower highs but would have fared a lot better overall. A lot of the trading cliches like cutting trading losses quickly, letting profits run, averaging up rather than down originate from this book. There is a reason people still talk about it 100 years after its publication. It's a good contender for the best trading book of all time.

> Competitive Advantage: The winners in AI won’t be those with the cleverest algorithms, but those who can effectively harness the most compute power.

> Career Focus: As AI engineers, our value lies not in crafting perfect algorithms but in building systems that can effectively leverage massive computational resources. That is a fundamental shift in mental models of how to build software.

I think the author has a fundamental misconception what making best use of computational resources requires. It's algorithms. His recommendation boils down to not do the one thing that would allow us to make the best use of computational resources.

His assumptions would only be correct if all the best algorithms were already known, which is clearly not the case at present.

Rich Sutton said something similar, but when he said it, he was thinking of old engineering intensive approaches, so it made sense in the context in which he said it and for the audience he directed it at. It was hardly groundbreaking either, the people whom he wrote the article for all thought the same thing already.

People like the author of this article don't understand the context and are taking his words as gospel. There is no reason not to think that there won't be different machine learning methods to supplant the current ones, and it's certain they won't be found by people who are convinced that algorithmic development is useless.

Underneath it all, there is some hope that an innovation might come about to keep the wave going, and indeed, a new branch of ML being discovered could revolutionize AI and actually be worthy of the hype that LLMs have now, but that has nothing to do with the LLM craze.

It's cool that we have them, and I also appreciate what Stable Diffusion has brought to the world, but in terms of how much LLMs influenced me, they only shorted the time it takes for me to read the documentation.

I don't think that machines cannot be more intelligent than humans. I don't think that the fact that they use linear algebra and mathematical functions makes the computers inferior to humans. I just think that the current algorithms suck. I want better algos so we can have actual AI instead of this trash.

Some of the stuff in this playlist might be relevant to you, though it is mostly about programming GPUs in a functional language that compiles to Cuda. The author (me) sometimes works on the language during the video, either fixing bugs or adding new features.

Edit: Nwm, I saw you worked for 90 days without pay. Ack.

Staged Functional Programming In Spiral

I am doing a fully fused ML GPU library along with a poker game to run it on in my own programming language that I've worked on for many years. Currently, right at this very moment, I am trying to optimize compilation times along with register usage by doing more on the heap, so I am creating a reference counting Cuda backend for Spiral.

Both the ML library and the poker game are designed to run completely on GPU for the sake of getting large speedups.

Once I am done with this and have trained the agent, I'll test it out on play money sites, and if that doesn't get it eaten by the rake, with real money.

I am doing fairly sophisticated functional programming in the videos, the kind you could only do in the Spiral language. Many parts of the series involve me working and improving the language itself in F#.

However, in the Spiral series, I aim to go beyond just making an ML library for running NN models and break new ground.

Newer GPUs actually support dynamic memory allocation, recursion, and the GPU threads have their own stacks, so you could in fact treat them as sequential devices and write games and simulators directly on them. I think once I finish the NL Holdem game, I'll be able to get over 100x fold improvements by running the whole program on the GPU versus the old approach of writing the sequential part on a CPU and only using the GPU to accelerate a NN model powering the computer agents.

I am not sure if this is a good answer, but this is how GPU programming would be helpful to me. It all comes down to performance.

The problem with programming them is that the program you are trying to speed up needs to be specially structured, so it utilizes the full capacity of the device.

Even so, creating all the abstractions needed to implement even regular matrix multiplication in Spiral in a generic fashion took me two months, so I'd consider that good enough exercise.

You could do it a lot faster by specializing for specific matrix sizes, like in the Cuda examples repo by Nvidia, but then you'd miss the opportunity to do the tensor magic that I did in the playlist.

I spent 2 months implementing a matmult kernel in Spiral and optimizing it.

I find this a very questionable business decision.

Back in 2015 I thought this would be the dominant model in 2022. I thought that the AI startups challenging Nvidia would be about that. Instead, they all targetted inference instead of programmability. I thought that a Tenstorrent hardware would be about what you are talking about - lots of tiny cores, local memory, message passing between them, AI/matmult intrinsics.

I've been hyped about Tenstorrent for a long time, but now that it is finally coming out with something, I can see that the Grayskulls are very overpriced. And if you look at the docs for their low-level kernel programming, you will see that Tensix cores can only have four registers, have no register spilling, and also don't support function calls. What would one be able to program with that?

It would have been interesting had the Grayskull cards been released in 2018. But in 2024 I have no idea what the company wants to do with them. It's over five years behind what I was expecting.

My expectations for how the AI hardware wave would unfold were fit for another world entirely. If this is the best the challengers can do, the most we can hope for is that they depress Nvidia's margins somewhat so we can buy its cards cheaper in the future. As we go towards the Singularity, I've gone from expecting revolutionary new hardware from AI startups to hoping Nvidia can keep making GPUs faster and more programmable.

Ironically, that latter thing is one trend that I missed, and going from Maxwell cards to the last generation, the GPUs have gained a lot in terms of how general purpose they are. The range of domains they can be used for is definitely going up as time goes on. I thought that AI chips would be necessary for this, and that GPUs would remain as toys, but it has been the other way around.