What we know about LLMs • Hacker News Board

Re: What we know about LLMs

#101

Post by KronisLV » Tue, Jul 25, 2023, 4:40 PM UTC

killernap wrote: post #13

ChatGPT was announced November, 2022 - 8 months ago. Time flies. Question for HN: Where are we in the hype cycle on this? We can run shitty clones slowly on Raspberry Pi's and your phone. The educational implementations demonstrate the basics in under a thousand lines of brisk C. Great. At some point you have to wonder... well, so what? Not one killer app has emerged. I for one am eager to be all hip and open minded…

> Not one killer app has emerged.

I mentioned the Stack Overflow Developer Survey once already today, but at the risk of sounding like a broken record, it has some data on this as well: https://survey.stackoverflow.co/2023/#ai

To save someone a click, around 44% of the respondents (some 39k out of 89k people) are already using "AI" solutions as a part of their workflow, another 25% (close to 23k people) are planning to do so soon.

The sentiment also seems mostly favorable, most aim to increase productivity or help themselves with learning and just generally knock out some more code, though there is a disconnect between what people want to use AI for (basically everything) and what they currently use it for (mostly just code).

There's also a section on the AI search tools in particular, about 83% of the respondents have at least had a look at ChatGPT, which is about as close to a killer app as you can probably get, even if it's cloud based SaaS: https://survey.stackoverflow.co/2023/#section-most-popular-t...

> Where are we in the hype cycle on this?

I'm not sure about the specifics here, but the trend feels about as significant as Docker and other container technologies more or less taking the industry by storm and changing a bunch of stuff around (to the point where most of my server software is containers).

That said, we're probably still somewhere in the early stages of the hype cycle for AI (the drawbacks like hallucination will really become apparent to many in the following years).

Honestly, the technology itself seems promising for select use cases and it's still nice that we have models that can be self hosted and somehow the software has gotten decent enough that you can play around with reasonably small models on your machine even without a GPU: https://blog.kronis.dev/tutorials/self-hosting-an-ai-llm-cha...

I'm cautiously optimistic about the current forms of LLM/AI, but fear that humanity will misuse the tech (as a cost cutting measure sometimes, without proper human review).

Re: What we know about LLMs

#102

Post by FishInTheWater » Tue, Jul 25, 2023, 4:42 PM UTC

flangola7 wrote: post #68

FishInTheWater wrote: post #28

Earlier quoted context omitted.

You're missing the point, there is a difference; The answers are often wrong, and more-wrong the more complex the question gets. They're only able to answer simple (relative-to-the-model's-size) straightforward reasoning questions. Which is a nice party trick, but not broadly useful. They can however tell you how to convert that problem into steps that can be run in an algebra system. Usually they can't do that very…

OpenAI is experimenting with process supervision, which corrects many of these issues: Problem Let x8+3x4−4=p1(x)p2(x)⋯pk(x),x8+3x4−4=p1(x)p2(x)⋯pk(x), where each non-constant polynomial pi(x)pi(x) is monic with integer coefficients, and cannot be factored further over the integers. Compute p1(1)+p2(1)+⋯+pk(1).p1(1)+p2(1)+⋯+pk(1). Model attempt Here, GPT-4 successfully performs a complex series of polynomial factoriz…

It's an impressive result, but shouldn't be seen as "correction". Framing it as a (drastic) reduction in mistakes is more useful here.

If the model is productionized (read: dumbed down so it isn't as expensive to run), the reasoning abilities drastically decline again.

And these reasoning abilities are still around a language model, rather than around abstract models.

This is a very effective party trick for general math, whose language quite directly maps onto these abstract concepts, but there are some holes. Information about e.g. which values may be zero isn't encoded in the language, and so this approach is liable to blundering around division-by-zero issues.

If you want a particular example to toy around with, LLMs are not fond of quaternions and their conversion to other representations.

Re: What we know about LLMs

#103

Post by KronisLV » Tue, Jul 25, 2023, 4:46 PM UTC

IKantRead wrote: post #54

killernap wrote: post #13

ChatGPT was announced November, 2022 - 8 months ago. Time flies. Question for HN: Where are we in the hype cycle on this? We can run shitty clones slowly on Raspberry Pi's and your phone. The educational implementations demonstrate the basics in under a thousand lines of brisk C. Great. At some point you have to wonder... well, so what? Not one killer app has emerged. I for one am eager to be all hip and open minded…

> Where are we in the hype cycle on this? Can we stop acting like the Gartner "hype cycle" is anything more than a marketing gimmick created Gartner to validate their own consulting/research services? While you can absolutely find cases that map to the "hype cycle", there is nothing whatsoever to validate this model as remotely accurate or valid for describing technology trends. Where is crypto in the "hype cycle"? I…

> The "hype cycle" is a neat idea but doesn't really map to reality in a way that makes it useful.

What do you propose as a more accurate alternative, or do you think that the whole idea should be scrapped? Because personally I feel like certain tech/practices certainly go through multiple stages, where initially people expect too much from them and eventually figure out what they're good for and what they're not.

Not always a single linear process, like NFTs/crypto refusing to die despite numerous scams out there and projects that seem to go nowhere, yet people still falling for the scams due to promised profits. However, the amount of people critiquing the blockchain as a crappy database seems to suggest at least some lessons learnt along the way and hopefully some actually decent use cases.

Re: What we know about LLMs

#104

Post by pseudonom- » Tue, Jul 25, 2023, 4:47 PM UTC

swyx wrote: post #99

> Transformers can be generally categorized into one of three categories: “encoder only” (a la BERT); “decoder only” (a la GPT); and having an “encoder-decoder” architecture (a la T5). Although all of these architectures can be rigged for a broad range of tasks (e.g. classification, translation, etc), encoders are thought to be useful for tasks where the entire sequence needs to be understood (such as sentiment class…

There's some discussion in section 3.2 of https://arxiv.org/pdf/1910.10683.pdf

Re: What we know about LLMs

#105

Post by dsubburam » Tue, Jul 25, 2023, 4:50 PM UTC

swyx wrote: post #99

> Transformers can be generally categorized into one of three categories: “encoder only” (a la BERT); “decoder only” (a la GPT); and having an “encoder-decoder” architecture (a la T5). Although all of these architectures can be rigged for a broad range of tasks (e.g. classification, translation, etc), encoders are thought to be useful for tasks where the entire sequence needs to be understood (such as sentiment class…

'Formal Algorithms for Transformers'[1] is a proper account of the architectures and what tasks they naturally lend themselves to, by authors from DeepMind. See sections 3 (Transformers and Typical Tasks) and 6 (Transformer Architectures).

Not much on empirical observations, though.

[1]https://arxiv.org/abs/2207.09238

Re: What we know about LLMs

#106

Post by bilsbie » Tue, Jul 25, 2023, 4:52 PM UTC

bradfox2 wrote: post #94

bilsbie wrote: post #91

Earlier quoted context omitted.

Thanks. That helps. So how can you update weights without doing back-propagation? Or is it still back propagation but with a different metric?

It's still loss being backproped, but the loss is calculated over a different criteria

Ok that makes a lot of sense.

Why do they call it reinforcement learning then? Is it not traditional RE such as Q learning?

Re: What we know about LLMs

#107

Post by RC_ITR » Tue, Jul 25, 2023, 4:53 PM UTC

bilsbie wrote: post #91

RC_ITR wrote: post #70

Earlier quoted context omitted.

RLHF does change the parameters. The way to think about it is that backpropagation changes the parameters of a model so they get closer to some sort of desired output. In pre-training and SFT, the parameters are changed so the model does a better job of replicating the next word in the training data, given the words it has already seen. In RLHF, the parameters are changed so the model does a better job of outputting…

Thanks. That helps. So how can you update weights without doing back-propagation? Or is it still back propagation but with a different metric?

Both do backpropagation, the difference is what you are backpropagating towards.

Think of it this way - there are an equal number of rude and polite comments online (actually probably way more rude ones).

If a model is trained on that data, how do you get it to only respond politely?

You could filter out the rude comments, but that's expensive and those rude comments may still have other helpful patterns that tech your model other stuff.

Alternatively, you could pre-train on the rude comments, but then after pre-training is done, you hire a ton of people in a low cost geo and ask them 'do you prefer comment 1 (a polite output of the pre-trained model) or comment 2 (a rude output).'

The model then 'learns' that comment 1 is better because it gets more votes, and adjusts parameter's (through backpropagation) to make comment 1 instead of comment 2

In practice, you can't control what the model outputs, so you just ask it to give you it's top N responses and the humans rank all of them, hoping you get a decent mix of rude and polite.

Re: What we know about LLMs

#108

Post by DebtDeflation » Tue, Jul 25, 2023, 4:54 PM UTC

killernap wrote: post #13

ChatGPT was announced November, 2022 - 8 months ago. Time flies. Question for HN: Where are we in the hype cycle on this? We can run shitty clones slowly on Raspberry Pi's and your phone. The educational implementations demonstrate the basics in under a thousand lines of brisk C. Great. At some point you have to wonder... well, so what? Not one killer app has emerged. I for one am eager to be all hip and open minded…

The killer app for large enterprises is Q&A against the corporate knowledgebase(s). Big companies have an insane amount of tribal knowledge locked away in documents sitting on Sharepoint, on Box, on file servers, etc. Best case scenario, their employees can do keyword search against a subset of those documents. Chunk those docs, run them through an embedding process, store the embeddings in a vector store, let employees ask questions, do a similarity search against the vector store, pass the top results and the question to the LMM, get an actual answer back to present to the employee. This unlocks a ton of knowledge and can be a massive productivity booster.

Re: What we know about LLMs

#109

Post by adamcblodgett » Tue, Jul 25, 2023, 4:59 PM UTC

DebtDeflation wrote: post #108

killernap wrote: post #13

ChatGPT was announced November, 2022 - 8 months ago. Time flies. Question for HN: Where are we in the hype cycle on this? We can run shitty clones slowly on Raspberry Pi's and your phone. The educational implementations demonstrate the basics in under a thousand lines of brisk C. Great. At some point you have to wonder... well, so what? Not one killer app has emerged. I for one am eager to be all hip and open minded…

The killer app for large enterprises is Q&A against the corporate knowledgebase(s). Big companies have an insane amount of tribal knowledge locked away in documents sitting on Sharepoint, on Box, on file servers, etc. Best case scenario, their employees can do keyword search against a subset of those documents. Chunk those docs, run them through an embedding process, store the embeddings in a vector store, let employ…

So the killer app for LLMs and AI in general is...a librarian?

Re: What we know about LLMs

#110

Post by bastawhiz » Tue, Jul 25, 2023, 5:00 PM UTC

killernap wrote: post #13

ChatGPT was announced November, 2022 - 8 months ago. Time flies. Question for HN: Where are we in the hype cycle on this? We can run shitty clones slowly on Raspberry Pi's and your phone. The educational implementations demonstrate the basics in under a thousand lines of brisk C. Great. At some point you have to wonder... well, so what? Not one killer app has emerged. I for one am eager to be all hip and open minded…

> Not one killer app has emerged.

I'll say that I pretty firmly disagree with this. I've been using Github Copilot for about six months for my own work and it has fundamentally changed how I write code. Ignoring the ethics of Copilot, if I just need to read a file with some data, parse it, and render that data on screen, Copilot just _does_ most of that for me. I write a chunky comment explaining what I want, it writes a blob of code that I tab through, and I'm left with a nicely-documented, functioning piece of software. A one-off script that took me 30 minutes to write previously now takes me maybe a minute on a bad day.

For ages we've had Text Expander and key mappings and shortcuts and macros that render templates of pre-built code. Now I can just say what I'm trying to do, the language model considers the other code on the page, and it gets done.

If this isn't a "killer app" then I'm not sure what is. In my entire career I can think of maybe two things that I've come upon that have affected my workflow this much: source control and continuous integration. Which, frankly, is wild.

Separately, I use LLMs to generate marketing copy for my side hustle. I suck at marketing, but I can tell the damn thing what I want to market and it gives me a list of tweets back that sound like the extroverted CMO that I don't have. I can outsource creative tasks like brainstorming lists of names for products, or coming up with text categories for user feedback from a spreadsheet. I don't know if I'd call either of those things "killer apps" but I have a tool which can do thinking for me at a nominal cost, quickly, and with a high-enough quality bar that it's usually not a waste of my time.