Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Bitter Lesson Is Misunderstood

obviouslywrong.substack.com

Post reply on HN
Print view View original

51–60 of 259 posts

Re: The Bitter Lesson Is Misunderstood

#51

Post by jeremyjh »

voxic11 wrote: post #21
jeremyjh wrote: post #20

Earlier quoted context omitted.

AlphaZero trained itself through chess games that it played with itself. Chess positions have something very close to an objective truth about the evaluation, the rules are clear and bounded. Winning is measurable. How do you achieve this for a language model? Yes, distillation is a thing but that is more about compression and filtering. Distillation does not produce new data in the same way that chess games produce…

Synthetic data is already widely used to do training in the programming and mathematics domains where automated verification is possible. Here is an example of an open source verified reasoning synthetic dataset https://www.primeintellect.ai/blog/synthetic-1

Are they actually producing new data though? This is the sort of thing I called "compression and filtering" because it seems to be new information content is not being produced, but LLMs are used to distill the information we already have. We need more raw information.

Re: The Bitter Lesson Is Misunderstood

#52

Post by FloorEgg »

marlott wrote: post #38
FloorEgg wrote: post #13

Earlier quoted context omitted.

Yes, for me both customers and colleagues continually suggested "hey let's just take all these samples of past work and dump it in the magical black box and then replicate what they have been doing". Instead I developed a UX that made it as easy as possible for people to explain what they want to be done , and a system that then goes and does that. Then we compare the system's output to their historical data and ther…

Interesting! Could you give an example with a bit more specific detail here? I take it there's some kind of work output, like a report, in a semi-structured format, and the goal is to automate creation of these. And you would provide a UX that lets them explain what they want the system to create?

Yes, essentially.

There are multiple long-form text inputs, one set is provided by User A, and another set by User B. User A inputs act as a prompt for User B, and then User A analyzes User B's input according to the original User A inputs, producing an output.

My system takes User A and B inputs and produces the output with more accuracy and precision than User As do, but a wide margin.

Instead of trying to train a model on all the history of these inputs and outputs, the solution was a combination of goal->job->task breakdown (like a fixed agentic process), and lots of context and prompt engineering. I then test against customer legacy samples, and inspect any variances by hand. At first the variances were usually system errors, which informed improvements to context and prompt engineering, and after working through about a thousand of these (test -> inspect variance -> if system mistake improve system -> repeat) iterations, and benefiting from a couple base-model upgrades, the variances are now about 99.9% user error (bad historical data or user inputs) and 0.1% system error. Overall it took about 9 months to build, and this one niche is worth ~$30m a year revenue easy, and everywhere I look there are market niches like this... it's ridiculous. (and a basic chat interface like ChatGPT doesn't work for these types of problems, no matter how smart it gets, for a variety of reasons)

So to summarize:

Instead of training a model on the historical inputs and outputs, the solution was to use the best base model LLMs, a pre-determined agentic flow, thoughtful system prompt and context engineering, and an iterative testing process with a human in the loop (me) to refine the overall system by carefully comparing the variances between system outputs and historical customer input/output samples.

Re: The Bitter Lesson Is Misunderstood

#53

Post by Quarrelsome »

kushalc wrote: post #49

Hey folks, OOP/original author and 20-year HN lurker here — a friend just told me about this and thought I'd chime in. Reading through the comments, I think there's one key point that might be getting lost: this isn't really about whether scaling is "dead" (it's not), but rather how we continue to scale for language models at the current LM frontier — 4-8h METR tasks. Someone commented below about verifiable rewards…

> if you can find a way to produce verifiable rewards about a target world

I feel like there's an interesting symmetry here between the pre and post LLM world, where I've always found that organisations over-optimise for things they can measure (e.g. balance sheets) and under-optimise for things they can't (e.g. developer productivity), which explains why its so hard to keep a software product up to date in an average org, as the natural pressure is to run it into the ground until a competitor suddenly displaces it.

So in a post LLM world, we have this gaping hole around things we either lack the data for, or as you say: lack the ability to produce verifiable rewards for. I wonder if similar patterns might play out as a consequence and what unmodelled, unrecorded, real-world things will be entirely ignored (perhaps to great detriment) because we simply lack a decent measure/verifiable-reward for it.

Re: The Bitter Lesson Is Misunderstood

#55

Post by gavmor »

throwaway314155 wrote: post #47

The scaling laws for transformers _deliberately_ factor in the amount of data as well as the amount of compute needed in order to scale. The premise of this article, that data is more important than compute has been obvious to people who are paying attention. Sorry but the unnecessary sensationalism in this article was mildly annoying to me. Like the author discovered some novel new insight. A bit like that doctor wh…

> The premise of this article... has been obvious to people who are paying attention.

Well, forgive me but I feel that the article is a much-needed injection of context into my thinking around the Bitter Lesson. I like the imperative to preface compute requests with data roadmaps.

I'm not an AI guy. Not an ML engineer. I've been studiously avoiding the low-level stuff, actually, because I didn't want to half-ass it when off-the-shelf solutions were still providing tremendous novelty and value for my customers.

So, for most of my career, "compute" has been practically irrelevant! RAM and disk constraints presented more frequent obstacles than processor cycles'. I would have easily told you that data presents more of a bottleneck to value than CPU. But that's just the era of computing I came up in.

The last few years have been different. Suddenly compute is at a premium, again. So it's easy to think, "if only I had more," and "line goes up!" and forget about s-curves and logarithmic scaling.

Is the article unnecessarily sensationalist? I don't know, maybe you've been overestimating how much the rest of us are "paying attention."[0]

0. https://xkcd.com/2501/

Re: The Bitter Lesson Is Misunderstood

#56

Post by FloorEgg »

kushalc wrote: post #49

Hey folks, OOP/original author and 20-year HN lurker here — a friend just told me about this and thought I'd chime in. Reading through the comments, I think there's one key point that might be getting lost: this isn't really about whether scaling is "dead" (it's not), but rather how we continue to scale for language models at the current LM frontier — 4-8h METR tasks. Someone commented below about verifiable rewards…

10+ years ago I expected we would get AI that would impact blue collar work long before AI that impacted white collar work. Not sure exactly where I got the impression, but I remember some "rising tide of AI" analogy and graphic that had artists and scientists positioned on the high ground.

Recently it doesn't seem to be playing out as such. The current best LLMs I find marvelously impressive (despite their flaws), and yet... where are all the awesome robots? Why can't I buy a robot that loads my dishwasher for me?

Last year this really started to bug me, and after digging into it with some friends I think we collectively realized something that may be a hint at the answer.

As far as we know, it took roughly 100M-1B years to evolve human level "embodiment" (evolve from single celled organisms to human), but it only took around ~100k-1M for humanity to evolve language, knowledge transfer and abstract reasoning.

So it makes me wonder, is embodiment (advanced robotics) 1000x harder than LLMs from an information processing perspective?

Re: The Bitter Lesson Is Misunderstood

#57

Post by dr_dshiv »

jacobolus wrote: post #37
dr_dshiv wrote: post #31

Earlier quoted context omitted.

Let’s keep in mind that we don’t have most of the renaissance through the early modern period (1400-1800) because it was published in neolatin with older typefaces— and only about 10% is even digitized. We probably don’t have most of the Arabic corpus either — and barely any Sanskrit. Classical Chinese is probably also lacking — only about 1% of it is translated to English.

The volume of text in English and digitized from the past few years dwarfs the volume of Latin text from all time. Unless you are wondering about a very niche historical topic there’s more written in English than Latin about basically everything.

Well, if you are looking for diversity of perspective— temporal diversity may be valuable.

Marsilio Ficino was hired by the Medici to translate Plato and other classical Greek works into Latin. He directly taught DaVinci, Raphael, Michelangelo, Toscanelli, etc. I mean to say that his ideas and perspectives helped spark the renaissance.

Insofar as we hope for an AI renaissance and not an AI apocalypse, it might benefit us to have the actual renaissance in the training data.

Re: The Bitter Lesson Is Misunderstood

#59

Post by FloorEgg »

socalgal2 wrote: post #40
FloorEgg wrote: post #10

Earlier quoted context omitted.

Very true. Living cells are ~4-5 orders of magnitude more functional-information-dense than the most advanced chips, and there is a lot more living mass than advanced chips. But the networking potential of digital compute is a fundamentally different paradigm than living systems. The human brain is constrained in size by the width of the female pelvis. So while it's expensive, we can trade scope-constrained robustnes…

> Living cells are ~4-5 orders of magnitude more functional-information-dense than the most advanced chips, In what sense is this true? That sounds suspiciously like cubic meter of dirt is more advanced than an iPhone because there are 6-7 orders of magnitude more atoms in the dirt.

Well, sort of, but functional-information is a specifically defined term that you can google. "Advanced" is vague and in this context not very helpful.

functional information is basically the amount of data (bits) necessary to explain all the possible functions matter can perform based on its unique configuration (in contrast to random). I am sure I partially butchered this explanation... but hopefully its close enough to catch my drift.

Life is optimized to process and learn from the real world, and it is insanely efficient at it and functional-information dense. (It might even be at the theoretical limit) Our most advanced technology is still 4-5 orders of magnitude behind it.

The capabilities of your iPhone are extremely narrow when compared to a handful of dirt. To you it may seem the opposite, but you are probably mixing up utility to you with functional capability. Your iPhone is has more functional utility to you, but the same amount dirt has way more general functional utility. (Your iphone isn't capable of self-replication, self-repair, and self-nonself distinction aka autopoiesis)

Re: The Bitter Lesson Is Misunderstood

#60

Post by kushalc »

FloorEgg wrote: post #56
kushalc wrote: post #49

Hey folks, OOP/original author and 20-year HN lurker here — a friend just told me about this and thought I'd chime in. Reading through the comments, I think there's one key point that might be getting lost: this isn't really about whether scaling is "dead" (it's not), but rather how we continue to scale for language models at the current LM frontier — 4-8h METR tasks. Someone commented below about verifiable rewards…

10+ years ago I expected we would get AI that would impact blue collar work long before AI that impacted white collar work. Not sure exactly where I got the impression, but I remember some "rising tide of AI" analogy and graphic that had artists and scientists positioned on the high ground. Recently it doesn't seem to be playing out as such. The current best LLMs I find marvelously impressive (despite their flaws), a…

Not a robotics guy, but to extent that the same fundamentals hold—

I think it's a degrees of freedom question. Given the (relatively) low conditional entropy of natural language, there aren't actually that many degrees of (true) freedom. On the other hand, in the real world, there are massively more degrees of freedom both in general (3 dimensions, 6 degrees of movement per joint, M joints, continuous vs. discrete space, etc.) and also given the path dependence of actions, the non-standardized nature of actuators, actuators, kinematics, etc.

All in, you get crushed by the curse of dimensionality. Given N degrees of true freedom, you need O(exp(N)) data points to achieve the same performance. Folks do a bunch of clever things to address that dimensionality explosion, but I think the overly reductionist point still stands: although the real world is theoretically verifiable (and theoretically could produce infinite data), in practice we currently have exponentially less real-world data for an exponentially harder problem.

Real roboticists should chime in...

Post reply on HN