11 ms·

What we know about LLMs

351 points by wilhelm____ 3y ago

blitzar 3y ago
You had me at "Crypto VCs & ”builders” making a hard left into AI"
- JazzyEssy 3y ago
  Oh really now lol
- dsco 3y ago
  As a matter of fact, there’s even more developers making a hard left into AI who have never touched crypto. The interesting follow up question is: what will they actually spend time on? Training new models? Copy pasting front ends on ChatGPT? Fine tuning models? I think many of them will be scared by how much of a hard science ML is vs just spinning up old CRUD apps
  - brigadier132 3y ago
    Have you used an llm as an application developer? The hard tech for generative ai will be commodified. The products built on top will not.
    - tudorw 3y ago
      As an enthusiast delving in for the pure pleasure of it I see a future ahead where as a programmer, I will have an array of options to use LLM's in hybrid systems. I'm not looking for an AGI, rather a mixture of experts that I can remix as I see fit, each one tuned for a subset of 'intelligence' that can be wielded with relative precision as part of a larger system that combines traditional programming with the new abilities offered by LLMS's and their kin. Certainly an interesting time to dip back in, there are aspects of the domain that mirror my experiences with the early web and for those I am grateful.
      - _ea1k 3y ago
        It does feel a bit like the internet era just before the Netscape IPO. I'm looking forward to the pets.com of AI. Maybe openai will go the way of Netscape and make it all FOSS eventually. That'd be nice.
  - Zach_the_Lizard 3y ago
    > Training new models? There is value in applying old techniques to new problems. Training a model to, I don't know, recognize snake species might help save snake bite victims lives. (This is an example I came up with in 5 seconds, please don't take it seriously) But there's also the whole "sell the shovel" aspect; it can be hard to train models. It can be hard to interpret the quality of the results. How do I know version 2 of the model is better than version 1? How do I even get labeled photos of snakes and not-snakes? I suspect solving some of those problems are where some of the real gold is buried.
  - washadjeffmad 3y ago
    I'd imagine something like openBB / BB Terminal with consolidated API access for financial reporting, a platform for insider communities ("chat, forums, and an app!"), etc. Make it a club, and it'll sell itself. Since investment has been demoed successfully with off the shelf models, I don't think we're waiting on big advancements to be able to build a product. The bar for something like this, short term, is 1) be cool and 2) lose less money than traditional investing, sometimes.
  - blitzar 3y ago
    > Training new models? Copy pasting front ends on ChatGPT? Fine tuning models? The stable diffusion community is probably 2 years more mature than the GPT, there we see gui tools of a kind (in colab notebooks) to abstract away from code and thenlots of fine tuning. On the professional side, adobe have plugged these tools into their products. https://www.adobe.com/sensei/generative-ai/firefly.html https://www.adobe.com/sensei/generative-ai/firefly.html
    - zirgs 3y ago
      It's a lot easier to run Stable Diffusion locally. Meanwhile only the dumbest LLMs work on ordinary consumer GPUs. Datacenter GPUs with 80 GB vram are ridiculously expensive.
- wilhelm____ 3y ago
  haha ty ty!
getmeinrn 3y ago
How do you do science on LLMs? I would imagine that is super important, given their broad impact on the social fabric. But they're non-deterministic, very expensive to train, and subjective. I understand we have some benchmarks for roughly understanding a model's competence. But is there any work in the area of understanding, through repeatable experiments, why LLMs behave how they do? Do we care?
- gjm11 3y ago
  I would imagine it's a bit like doing science on human beings, who are also non-deterministic, expensive to train, and subjective. Perhaps there's scope for a scientific discipline corresponding to psychology but concerned with AI systems. We could call it robopsychology.
- api 3y ago
  I'm pretty much certain the cost of training and running large LLMs is going to come down, because it's only a matter of time before truly customized chips come out for these. GPUs really aren't that. They're massively parallel vector processors that turn out to be generally better than CPUs at running these models, but they're still not the ideal chip for running LLMs. That would be a large even more specialized parallel processor where almost all the silicon is dedicated to running exactly the types of operations used in large LLMs and that natively supports quantization formats such as those found in the ggml/llama.cpp world. Being able to natively run and train on those formats would allow gigantic 100B+ models to be run with more reasonable amounts of RAM and at a higher speed due to memory bandwidth constraints. These chips, when they arrive, will be a lot cheaper than GPUs when compared in dollars per LLM performance. They'll be available for rent in the cloud and for purchase as accelerators. I'd be utterly shocked if lots of chip companies don't have projects working on these chips, since at this point it's clear that LLMs are going to become a permanent fixture of computing.
  - DanHulton 3y ago
    I feel like it took practically no time for custom ASICs for bitcoin mining to show up, as soon as it was determined there was real money involved. Given that there's already definitely real money involved here, I wonder what's holding up the custom AI ASICs?
jannyfer 3y ago
Great summary. I’ve been reading a pop neuroscience book called Incognito (2011). In it, the author talks about how the brain is a group of competing sub-brains of many forms, and the brain might have several ways of doing the same thing (e.g. recognizing an object). The author also posited that the lack of AI progress back then was due to the fact that there are no constantly competing sub-brains. Our brains are always adjusting and trying new scenarios. I was struck by how similar these brain observations were to recent developments in AI and LLMs. The book is full of cool stories, even if some of them are now recognized as non-reproducible. I recommend!
- go_elmo 3y ago
  In the end - an AI should have these competing subsystems in one system - just as our brains are one system. What I find extremely interesting is how perception and thinking differs from person to person too - it was a "taboo" topic to call this neurodiversity - just as other genetic traits, but AI makes this relevant more than ever imo. Sure, its complicated and much comes from nurture (Nurture vs nature.. as exposure/epigeneticd vs genetics) but there sure are markable differences - the ones starting to stand out are e.g. adhd / autistic people, but Im sure it wont stay just there over time!
  - samwillis 3y ago
    You touch in an important topic here, how our understanding of AI/ML/LLMs will influence our "understanding" of the human brain and intelligence. My fear is that we will ascribe too much human behaviour to that we see in and understand of our AI inventions, and that this will result in the dehumanisation of people. So essentially my fear is what we justify doing to each other due to AI, rather than what "AGI" could do to us.
    - lacrimacida 3y ago
      I started seeing this dehumanisation spring up even here on HN, comparing LLMs with human brains and human thoughts as similar regurgitation to LLMs. I’m afraid it will get worse as this technology advances
FishInTheWater 3y ago
Given a set of instructions, an instruction fine-tuned/aligned LLM is able (conditional on size and training quality) to reason through a set of steps to produce a desired output. This is plainly wrong. The model's growing size makes it better at guessing the outcome of a reasoning task, but little to no actual reasoning is performed. It's trivial to prove this as well, as LLMs will still fail miserably at (larger) math problems that even basic computer algebra systems will handle with ease.
- viraptor 3y ago
  > The model's growing size makes it better at guessing the outcome of a reasoning task, but little to no actual reasoning is performed. If there's no observable difference between the behaviours, why not call it as the post did? > LLMs will still fail miserably at (larger) math problems They're neither trained on such problems, nor is that a goal for LLMs. They can however tell you how to convert that problem into steps that can be run in an algebra system.
  - taylodl 3y ago
    Which means they're this close in being able to reach out to an algebra system and run the steps and return you the result. I was just talking about this problem with someone the other day - how can it recognize that it doesn't have the answer but knows where it can get data so that it can form an answer. This seems to be the path Google is taking.
  - FishInTheWater 3y ago
    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 well either. Converting a problem from one description to another is algebraic reasoning, subject to the issues already mentioned. What they can do is summarize general instructions and documentation, provided adequate training data was available. They're neither trained on such problems, nor is that a goal for LLMs Yes. But LLMs keep being pushed for tasks that heavily involve abstract reasoning, which is dangerous as they're unsuited for it. (E.g. Any code generation that isn't mere empty boilerplate.)
killernap 3y ago
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 and pretend like I use LLMs all the time for everything and they are "the future" but novelty aside it seems like so far we have a demented clippy and some sophomoric arguments about alignment and wrong think. It did generate a whole lot of breathless click-bait-y articles and gave people something to blab about. Ironically it also accelerated the value of that sort of gab and clicks towards zero. As I am not a VC, politician, or opportunist, hand waving and telling me this is Frankenstein's monster about to come alive and therefore I need billions of dollars or "regulations" just makes folks sound like the crypto scammers. Please HN, say something actually insightful, I beg you.
- usaar333 3y ago
  > Not one killer app has emerged ChatGPT itself is a killer app.
  - killernap 3y ago
    About as killer as that twitter clone that was in the news for a minute after forcing people to use it and immediately losing 90% of the captive audience.. They have been losing users. Summer is here, school is out, the kids are back in reality for the moment and apparently when they aren't busy plagiarizing homework the interest is very limited.
    - throwuwu 3y ago
      lol, now who’s demented? Everyone I know uses it. It even diagnosed a problem with my pool filter among dozens of other uses I find for it. I like it and use it more than Google and stack overflow now. Losing the school crowd for the summer isn’t the beginning of the end, it just means there’s a cohort that doesn’t need it as much for a few months while they’re out having fun instead of stuck inside writing papers and doing math problems.
cubefox 3y ago
Specifically about RLHF, I find this video by Rob Miles still the best presentation of the ingenious original 2017(!) paper: https://youtube.com/watch?v=PYylPRX6z4Q https://youtube.com/watch?v=PYylPRX6z4Q RLHF is actually older than GPT-1, which came out in 2018. It didn't get applied to language models until 2022 with InstructGPT, an approach which combined supervised instruction fine-tuning with RLHF.
jawerty 3y ago
I run through a lot of these concepts, specifically RLHF, in my latest coding stream where I finetune LLama 2 if anyone's interested in getting a LLM deep dive https://www.youtube.com/watch?v=TYgtG2Th6fI&t=4002s https://www.youtube.com/watch?v=TYgtG2Th6fI&t=4002s Long story short, the size of the model and reward mechanisms used in validating off of human annotating/feedback are the main differences between what we can do as independents in OSS vs OpenAI. BigCode's StarCoder (https://huggingface.co/bigcode/starcoder https://huggingface.co/bigcode/starcoder) has some human labor backing it (I believe correct me if I'm wrong) but at the end of the day a company will always be able to gather people better. Not knocking Starcoder, in fact I streamed how to fine tune it the other day. However, it's important to mention some of the limitations in the OSS space now (big reason Meta pushing LLama 2 is a nice to have)
- bilsbie 3y ago
  When you’re doing RLHF are you actually modifying the weights of llama itself? Or is something on top?
  - jawerty 3y ago
    So I'm not doing RLHF that's how LLama is pre-trained. It's in the loss/optimization phase in their training I believe. For the finetuning i'm using LoRA to freeze most of the layers for parameter optimization. Using PEFT from huggingface
    - hallqv 3y ago
      RLHF is not part of LLaMa pretraining, or pretraning of any other models for that matter. RLHF comes after pretraining. https://twitter.com/Jeande_d/status/1661833563069620247/photo/1 https://twitter.com/Jeande_d/status/1661833563069620247/phot...
      - jsmith45 3y ago
        Seems like a classic case of a term of art overlapping with normal English terminology. Knowing that you will be doing further training on a provided model (even "just" extensive fine-tuning), one would want to distinguish the training done before you get your hands on it, from the training you do. An obvious word for that previous training is pre-training, which unfortunately conflicts with a term of art.
llm_nerd 3y ago
"Crypto VCs & ”builders” making a hard left into AI" This is a humorous intro graphic caption, but this sentiment appears on here constantly and it's self-destructive. This response might seem a bit over the top to a funny graphic, but I am replying to the general "ha ha AI like crypto amirite?" sentiment that is incredibly boring and worn out. When confronted with challenging new technology that we don't understand, some knee-jerk to acting dismissive. As if that has any hope at all of changing outcomes. It's especially weird when people who are clearly on the "I must desperately learn this as quickly as I can and try to present myself as some sort of expert" still incant the rhetoric -- "joking on the square" as it were -- as if they need to defend their prior dismissals. Constantly on here there is yet another trivial "intro to tokenization" blog entry that brays some tired crypto comparison. Stop it. The Venn diagram of people at the forefront of ML/LLM, and its advocates, is almost entirely separate from the web/crypto sphere. There is astonishingly little overlap. Crypto was hyped because some people truly saw a purpose, coupled with masses of scammers and getrichquick sorts. AI/LLM/ML is hyped because it is revolutionary and has already yielded infinitely more practical impact than crypto ever did.
- uhtred 3y ago
  I think when the dust settles we are just going to have some chat bots and a few rich grifters.
- NhanH 3y ago
  > The Venn diagram of people at the forefront of ML/LLM, and its advocates, is almost entirely separate from the web/crypto sphere. There is astonishingly little overlap That statement seems false. Especially since this was a headline I saw in half a dozen online news in my country yesterday. > OpenAI's Sam Altman launches Worldcoin crypto project[0] If anything, even without taking into account the greed stuff, people who are drawn to fun tech is likely to be drawn to both LLM and cryptocurrency stuffs. [0]https://www.reuters.com/technology/openais-sam-altman-launches-worldcoin-crypto-project-2023-07-24/ https://www.reuters.com/technology/openais-sam-altman-launch...
  - llm_nerd 3y ago
tmaly 3y ago
>Rather than explicitly labeling data, it might be easier for a human to read two or more LLM outputs and encode their preferences through comparison. This reminded me a lot of what economist Murray Rothbard talked about on preferences in his treatise Man, Economy, and State. There is likely to be other insights hidden in these philosophical works on human choices.
- samstave 3y ago
  Subscribe. ELI5 - I'd like to know more about this, as I have no experience with this line of thought as of yet.
yding 3y ago
Good job Will!
- wilhelm____ 3y ago
  thank you, Yi!
swyx 3y ago
> 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 classification), whereas decoders are thought to be useful for tasks where text needs to be completed (such as completing a sentence). Encoder-decoder architectures can be applied to a variety of problems, but are most famously associated with language translation. theres a whole lot of "thought to be"'s here. is there a proper study done on the relative effectiveness of encoder only vs decoder only vs encoder-decoder for various tasks?
- pseudonom- 3y ago
  There's some discussion in section 3.2 of https://arxiv.org/pdf/1910.10683.pdf https://arxiv.org/pdf/1910.10683.pdf
  - swyx 3y ago
    ty!
- dsubburam 3y ago
  '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 https://arxiv.org/abs/2207.09238
  - swyx 3y ago
    ty!
zuckerborgggg 3y ago
[flagged]
janalsncm 3y ago
> Transformers can read the entire sequence at once and learn to “pay attention to” only the values that came earlier in time (via “masking”) Unless the text fits into the model’s context window, this is incorrect. The self-attention layers will train via a sliding window over the text. Learning to attend only to previous tokens is also not correct. There are a lot of ways to train a transformer. BERT is bidirectional for example.
- wilhelm____ 3y ago
  I mention that above this sentence in the image - glad someone was "paying attention to" the text ;)

351 points by wilhelm____ 3y ago

blitzar 3y ago

You had me at "Crypto VCs & ”builders” making a hard left into AI"

JazzyEssy 3y ago
Oh really now lol
dsco 3y ago
As a matter of fact, there’s even more developers making a hard left into AI who have never touched crypto. The interesting follow up question is: what will they actually spend time on? Training new models? Copy pasting front ends on ChatGPT? Fine tuning models? I think many of them will be scared by how much of a hard science ML is vs just spinning up old CRUD apps
- brigadier132 3y ago
  Have you used an llm as an application developer? The hard tech for generative ai will be commodified. The products built on top will not.
  - tudorw 3y ago
    As an enthusiast delving in for the pure pleasure of it I see a future ahead where as a programmer, I will have an array of options to use LLM's in hybrid systems. I'm not looking for an AGI, rather a mixture of experts that I can remix as I see fit, each one tuned for a subset of 'intelligence' that can be wielded with relative precision as part of a larger system that combines traditional programming with the new abilities offered by LLMS's and their kin. Certainly an interesting time to dip back in, there are aspects of the domain that mirror my experiences with the early web and for those I am grateful.
    - _ea1k 3y ago
      It does feel a bit like the internet era just before the Netscape IPO. I'm looking forward to the pets.com of AI. Maybe openai will go the way of Netscape and make it all FOSS eventually. That'd be nice.
- Zach_the_Lizard 3y ago
  > Training new models? There is value in applying old techniques to new problems. Training a model to, I don't know, recognize snake species might help save snake bite victims lives. (This is an example I came up with in 5 seconds, please don't take it seriously) But there's also the whole "sell the shovel" aspect; it can be hard to train models. It can be hard to interpret the quality of the results. How do I know version 2 of the model is better than version 1? How do I even get labeled photos of snakes and not-snakes? I suspect solving some of those problems are where some of the real gold is buried.
- washadjeffmad 3y ago
  I'd imagine something like openBB / BB Terminal with consolidated API access for financial reporting, a platform for insider communities ("chat, forums, and an app!"), etc. Make it a club, and it'll sell itself. Since investment has been demoed successfully with off the shelf models, I don't think we're waiting on big advancements to be able to build a product. The bar for something like this, short term, is 1) be cool and 2) lose less money than traditional investing, sometimes.
- blitzar 3y ago
  > Training new models? Copy pasting front ends on ChatGPT? Fine tuning models? The stable diffusion community is probably 2 years more mature than the GPT, there we see gui tools of a kind (in colab notebooks) to abstract away from code and thenlots of fine tuning. On the professional side, adobe have plugged these tools into their products. https://www.adobe.com/sensei/generative-ai/firefly.html https://www.adobe.com/sensei/generative-ai/firefly.html
  - zirgs 3y ago
    It's a lot easier to run Stable Diffusion locally. Meanwhile only the dumbest LLMs work on ordinary consumer GPUs. Datacenter GPUs with 80 GB vram are ridiculously expensive.
wilhelm____ 3y ago
haha ty ty!

getmeinrn 3y ago

How do you do science on LLMs? I would imagine that is super important, given their broad impact on the social fabric. But they're non-deterministic, very expensive to train, and subjective. I understand we have some benchmarks for roughly understanding a model's competence. But is there any work in the area of understanding, through repeatable experiments, why LLMs behave how they do? Do we care?

gjm11 3y ago
I would imagine it's a bit like doing science on human beings, who are also non-deterministic, expensive to train, and subjective. Perhaps there's scope for a scientific discipline corresponding to psychology but concerned with AI systems. We could call it robopsychology.
api 3y ago
I'm pretty much certain the cost of training and running large LLMs is going to come down, because it's only a matter of time before truly customized chips come out for these. GPUs really aren't that. They're massively parallel vector processors that turn out to be generally better than CPUs at running these models, but they're still not the ideal chip for running LLMs. That would be a large even more specialized parallel processor where almost all the silicon is dedicated to running exactly the types of operations used in large LLMs and that natively supports quantization formats such as those found in the ggml/llama.cpp world. Being able to natively run and train on those formats would allow gigantic 100B+ models to be run with more reasonable amounts of RAM and at a higher speed due to memory bandwidth constraints. These chips, when they arrive, will be a lot cheaper than GPUs when compared in dollars per LLM performance. They'll be available for rent in the cloud and for purchase as accelerators. I'd be utterly shocked if lots of chip companies don't have projects working on these chips, since at this point it's clear that LLMs are going to become a permanent fixture of computing.
- DanHulton 3y ago
  I feel like it took practically no time for custom ASICs for bitcoin mining to show up, as soon as it was determined there was real money involved. Given that there's already definitely real money involved here, I wonder what's holding up the custom AI ASICs?

jannyfer 3y ago

Great summary. I’ve been reading a pop neuroscience book called Incognito (2011). In it, the author talks about how the brain is a group of competing sub-brains of many forms, and the brain might have several ways of doing the same thing (e.g. recognizing an object). The author also posited that the lack of AI progress back then was due to the fact that there are no constantly competing sub-brains. Our brains are always adjusting and trying new scenarios. I was struck by how similar these brain observations were to recent developments in AI and LLMs. The book is full of cool stories, even if some of them are now recognized as non-reproducible. I recommend!

go_elmo 3y ago
In the end - an AI should have these competing subsystems in one system - just as our brains are one system. What I find extremely interesting is how perception and thinking differs from person to person too - it was a "taboo" topic to call this neurodiversity - just as other genetic traits, but AI makes this relevant more than ever imo. Sure, its complicated and much comes from nurture (Nurture vs nature.. as exposure/epigeneticd vs genetics) but there sure are markable differences - the ones starting to stand out are e.g. adhd / autistic people, but Im sure it wont stay just there over time!
- samwillis 3y ago
  You touch in an important topic here, how our understanding of AI/ML/LLMs will influence our "understanding" of the human brain and intelligence. My fear is that we will ascribe too much human behaviour to that we see in and understand of our AI inventions, and that this will result in the dehumanisation of people. So essentially my fear is what we justify doing to each other due to AI, rather than what "AGI" could do to us.
  - lacrimacida 3y ago
    I started seeing this dehumanisation spring up even here on HN, comparing LLMs with human brains and human thoughts as similar regurgitation to LLMs. I’m afraid it will get worse as this technology advances

FishInTheWater 3y ago

Given a set of instructions, an instruction fine-tuned/aligned LLM is able (conditional on size and training quality) to reason through a set of steps to produce a desired output. This is plainly wrong. The model's growing size makes it better at guessing the outcome of a reasoning task, but little to no actual reasoning is performed. It's trivial to prove this as well, as LLMs will still fail miserably at (larger) math problems that even basic computer algebra systems will handle with ease.

viraptor 3y ago
> The model's growing size makes it better at guessing the outcome of a reasoning task, but little to no actual reasoning is performed. If there's no observable difference between the behaviours, why not call it as the post did? > LLMs will still fail miserably at (larger) math problems They're neither trained on such problems, nor is that a goal for LLMs. They can however tell you how to convert that problem into steps that can be run in an algebra system.
- taylodl 3y ago
  Which means they're this close in being able to reach out to an algebra system and run the steps and return you the result. I was just talking about this problem with someone the other day - how can it recognize that it doesn't have the answer but knows where it can get data so that it can form an answer. This seems to be the path Google is taking.
- FishInTheWater 3y ago
  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 well either. Converting a problem from one description to another is algebraic reasoning, subject to the issues already mentioned. What they can do is summarize general instructions and documentation, provided adequate training data was available. They're neither trained on such problems, nor is that a goal for LLMs Yes. But LLMs keep being pushed for tasks that heavily involve abstract reasoning, which is dangerous as they're unsuited for it. (E.g. Any code generation that isn't mere empty boilerplate.)

killernap 3y ago

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 and pretend like I use LLMs all the time for everything and they are "the future" but novelty aside it seems like so far we have a demented clippy and some sophomoric arguments about alignment and wrong think. It did generate a whole lot of breathless click-bait-y articles and gave people something to blab about. Ironically it also accelerated the value of that sort of gab and clicks towards zero. As I am not a VC, politician, or opportunist, hand waving and telling me this is Frankenstein's monster about to come alive and therefore I need billions of dollars or "regulations" just makes folks sound like the crypto scammers. Please HN, say something actually insightful, I beg you.

usaar333 3y ago
> Not one killer app has emerged ChatGPT itself is a killer app.
- killernap 3y ago
  About as killer as that twitter clone that was in the news for a minute after forcing people to use it and immediately losing 90% of the captive audience.. They have been losing users. Summer is here, school is out, the kids are back in reality for the moment and apparently when they aren't busy plagiarizing homework the interest is very limited.
  - throwuwu 3y ago
    lol, now who’s demented? Everyone I know uses it. It even diagnosed a problem with my pool filter among dozens of other uses I find for it. I like it and use it more than Google and stack overflow now. Losing the school crowd for the summer isn’t the beginning of the end, it just means there’s a cohort that doesn’t need it as much for a few months while they’re out having fun instead of stuck inside writing papers and doing math problems.

cubefox 3y ago

Specifically about RLHF, I find this video by Rob Miles still the best presentation of the ingenious original 2017(!) paper: https://youtube.com/watch?v=PYylPRX6z4Q https://youtube.com/watch?v=PYylPRX6z4Q RLHF is actually older than GPT-1, which came out in 2018. It didn't get applied to language models until 2022 with InstructGPT, an approach which combined supervised instruction fine-tuning with RLHF.

jawerty 3y ago

I run through a lot of these concepts, specifically RLHF, in my latest coding stream where I finetune LLama 2 if anyone's interested in getting a LLM deep dive https://www.youtube.com/watch?v=TYgtG2Th6fI&t=4002s https://www.youtube.com/watch?v=TYgtG2Th6fI&t=4002s Long story short, the size of the model and reward mechanisms used in validating off of human annotating/feedback are the main differences between what we can do as independents in OSS vs OpenAI. BigCode's StarCoder (https://huggingface.co/bigcode/starcoder https://huggingface.co/bigcode/starcoder) has some human labor backing it (I believe correct me if I'm wrong) but at the end of the day a company will always be able to gather people better. Not knocking Starcoder, in fact I streamed how to fine tune it the other day. However, it's important to mention some of the limitations in the OSS space now (big reason Meta pushing LLama 2 is a nice to have)

bilsbie 3y ago
When you’re doing RLHF are you actually modifying the weights of llama itself? Or is something on top?
- jawerty 3y ago
  So I'm not doing RLHF that's how LLama is pre-trained. It's in the loss/optimization phase in their training I believe. For the finetuning i'm using LoRA to freeze most of the layers for parameter optimization. Using PEFT from huggingface
  - hallqv 3y ago
    RLHF is not part of LLaMa pretraining, or pretraning of any other models for that matter. RLHF comes after pretraining. https://twitter.com/Jeande_d/status/1661833563069620247/photo/1 https://twitter.com/Jeande_d/status/1661833563069620247/phot...
    - jsmith45 3y ago
      Seems like a classic case of a term of art overlapping with normal English terminology. Knowing that you will be doing further training on a provided model (even "just" extensive fine-tuning), one would want to distinguish the training done before you get your hands on it, from the training you do. An obvious word for that previous training is pre-training, which unfortunately conflicts with a term of art.

llm_nerd 3y ago

"Crypto VCs & ”builders” making a hard left into AI" This is a humorous intro graphic caption, but this sentiment appears on here constantly and it's self-destructive. This response might seem a bit over the top to a funny graphic, but I am replying to the general "ha ha AI like crypto amirite?" sentiment that is incredibly boring and worn out. When confronted with challenging new technology that we don't understand, some knee-jerk to acting dismissive. As if that has any hope at all of changing outcomes. It's especially weird when people who are clearly on the "I must desperately learn this as quickly as I can and try to present myself as some sort of expert" still incant the rhetoric -- "joking on the square" as it were -- as if they need to defend their prior dismissals. Constantly on here there is yet another trivial "intro to tokenization" blog entry that brays some tired crypto comparison. Stop it. The Venn diagram of people at the forefront of ML/LLM, and its advocates, is almost entirely separate from the web/crypto sphere. There is astonishingly little overlap. Crypto was hyped because some people truly saw a purpose, coupled with masses of scammers and getrichquick sorts. AI/LLM/ML is hyped because it is revolutionary and has already yielded infinitely more practical impact than crypto ever did.

uhtred 3y ago
I think when the dust settles we are just going to have some chat bots and a few rich grifters.
NhanH 3y ago
> The Venn diagram of people at the forefront of ML/LLM, and its advocates, is almost entirely separate from the web/crypto sphere. There is astonishingly little overlap That statement seems false. Especially since this was a headline I saw in half a dozen online news in my country yesterday. > OpenAI's Sam Altman launches Worldcoin crypto project[0] If anything, even without taking into account the greed stuff, people who are drawn to fun tech is likely to be drawn to both LLM and cryptocurrency stuffs. [0]https://www.reuters.com/technology/openais-sam-altman-launches-worldcoin-crypto-project-2023-07-24/ https://www.reuters.com/technology/openais-sam-altman-launch...
- llm_nerd 3y ago

tmaly 3y ago

>Rather than explicitly labeling data, it might be easier for a human to read two or more LLM outputs and encode their preferences through comparison. This reminded me a lot of what economist Murray Rothbard talked about on preferences in his treatise Man, Economy, and State. There is likely to be other insights hidden in these philosophical works on human choices.

samstave 3y ago
Subscribe. ELI5 - I'd like to know more about this, as I have no experience with this line of thought as of yet.

yding 3y ago

Good job Will!

wilhelm____ 3y ago
thank you, Yi!

swyx 3y ago

> 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 classification), whereas decoders are thought to be useful for tasks where text needs to be completed (such as completing a sentence). Encoder-decoder architectures can be applied to a variety of problems, but are most famously associated with language translation. theres a whole lot of "thought to be"'s here. is there a proper study done on the relative effectiveness of encoder only vs decoder only vs encoder-decoder for various tasks?

pseudonom- 3y ago
There's some discussion in section 3.2 of https://arxiv.org/pdf/1910.10683.pdf https://arxiv.org/pdf/1910.10683.pdf
- swyx 3y ago
  ty!
dsubburam 3y ago
'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 https://arxiv.org/abs/2207.09238
- swyx 3y ago
  ty!

zuckerborgggg 3y ago

[flagged]

janalsncm 3y ago

> Transformers can read the entire sequence at once and learn to “pay attention to” only the values that came earlier in time (via “masking”) Unless the text fits into the model’s context window, this is incorrect. The self-attention layers will train via a sliding window over the text. Learning to attend only to previous tokens is also not correct. There are a lot of ways to train a transformer. BERT is bidirectional for example.

wilhelm____ 3y ago
I mention that above this sentence in the image - glad someone was "paying attention to" the text ;)