What we know about LLMs • Hacker News Board

Re: What we know about LLMs

#41

Post by scottmcf » Tue, Jul 25, 2023, 2:03 PM UTC

Arkhaine_kupo wrote: post #37

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 think the microsoft gpt integration on Office is probably that app. Ability to ask to have your email's summarised, or getting your excel sheets formulas configured with natural language, etc are increidbly useful tools to lower the floor of entry to tools that already speed up humans so much. I don't think the use of this tools is some life redefining feature, but a friend of mine…

I agree, if it delivers on the kind of demos they showed off here:

https://news.microsoft.com/reinventing-productivity/

It's going to be an absolute "killer app".

Re: What we know about LLMs

#42

Post by cubefox » Tue, Jul 25, 2023, 2:05 PM UTC

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

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.

Re: What we know about LLMs

#43

Post by seydor » Tue, Jul 25, 2023, 2:05 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…

Why don't you ask ChatGPT or bard? If there s a hype cycle, it is just starting.

The killer app is the LLM tech itself, and the victim seems to be the whole tech ecosystem. It disintermediates everyone who is gatekeeping information and connects end users with the information they want without the google, the SEO and without ads. Even if we are not right there today, the potential is there. This in itself is huge, since the whole ecosystem of SV is funded by ads.

Re: What we know about LLMs

#44

Post by killernap » Tue, Jul 25, 2023, 2:09 PM UTC

scottmcf wrote: post #41

Arkhaine_kupo wrote: post #37

Earlier quoted context omitted.

> Not one killer app has emerged. I think the microsoft gpt integration on Office is probably that app. Ability to ask to have your email's summarised, or getting your excel sheets formulas configured with natural language, etc are increidbly useful tools to lower the floor of entry to tools that already speed up humans so much. I don't think the use of this tools is some life redefining feature, but a friend of mine…

I agree, if it delivers on the kind of demos they showed off here: https://news.microsoft.com/reinventing-productivity/ It's going to be an absolute "killer app".

https://en.wikipedia.org/wiki/BonziBuddy

My god. If we hit that bullseye, the rest of the dominoes will fall like a house of cards. Checkmate.

Re: What we know about LLMs

#45

Post by lolinder » Tue, Jul 25, 2023, 2:09 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…

That 8 months seems like a long time to you is indicative of just how fast tech has been moving lately. I expect at least another year before we have a good sense for where we actually are, probably more.

However, I'll hazard a guess: I think we haven't seen many real new apps since then because too many people are focused on packaging ChatGPT for X. A chatbot is a perfectly decent use case for some things, but I think the real progress will come when people stop trying to copy what OpenAI already did and start integrating LLMs in a more hands-off way that's more natural to their domains.

A great example that's changed my life is News Minimalist [0]. They feed all the news from a ton of sources into one of the GPT models and have it rate the story for significance and credibility. Only the highest rated stories make it into the newsletter. It's still rough around the edges, but being able to delegate most of my news consumption has already made a huge difference in my quality of life!

I expect successful and useful applications to fall in a similar vein to News Minimalist. They're not going to turn the world upside down like the hype artists claim, but there is real value to be made if people can start with a real problem instead of just adding a chatbot to everything.

[0] https://www.newsminimalist.com/

Re: What we know about LLMs

#46

Post by StackOverlord » Tue, Jul 25, 2023, 2:09 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…

I think it has shown the limitations of the Society of Mind hypothesis. Aggregating individuals equates to aggregating knowledge/experience, not intelligence. This is why hives and anthills do not really surpass their individuals intelligence. Ditto for human societies. In other words: composing LLMs using tools like langchain yield minor improvements over a single LLM instance.

Re: What we know about LLMs

#47

Post by imglorp » Tue, Jul 25, 2023, 2:13 PM UTC

seydor wrote: post #43

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…

Why don't you ask ChatGPT or bard? If there s a hype cycle, it is just starting. The killer app is the LLM tech itself, and the victim seems to be the whole tech ecosystem. It disintermediates everyone who is gatekeeping information and connects end users with the information they want without the google, the SEO and without ads. Even if we are not right there today, the potential is there. This in itself is huge, si…

Gatekeepers which can be destroyed by the disintermediation, should be.

Re: What we know about LLMs

#48

Post by throwuwu » Tue, Jul 25, 2023, 2:17 PM UTC

killernap wrote: post #38

throwuwu wrote: post #22

Earlier quoted context omitted.

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 an…

It is great that everyone you know uses it but the traffic to ChatGPT is decreasing and has been for over two months now. If pointing this fact out makes me demented consider that perhaps you are emotionally invested in this new toy/brand. I guess we can wait and see what kind of usage trends will emerge long term. My anecdotal evidence (which is not worth much, same as yours) is that many normies tried it a few time…

Sounds more like you have an axe to grind

Re: What we know about LLMs

#49

Post by simonh » Tue, Jul 25, 2023, 2:17 PM UTC

go_elmo wrote: post #11

jannyfer wrote: post #7

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 alw…

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 exposu…

Even within my immediate family we seem to have distinct differences in our conscious experience. My wife has very little visual or auditory experience of thought, no inner voice even when reading a book. While I mostly experience speaking as a continuous stream of words coming basically from my subconscious, with only a vague sense of what's coming up, one of my daughters says she is consciously aware of the exact words she is going to say several seconds in advance. It's like she has the ability to introspect her internal speech buffer, while I can't.

So while I'm sure there are a lot of custom tuned, problem specific hardware structures in our brain architecture, we do seem to learn how to actually use that hardware individually. As a result we seem to come up with a diverse range of different high level approaches.

Re: What we know about LLMs

#50

Post by mjr00 » Tue, Jul 25, 2023, 2:19 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 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.

In my mind I divide LLM usage into two categories, creation and ingestion.

Creation is largely a parlor trick that blew the minds of some people because it was their first exposure to generative AI. Now that some time has passed, most people can pattern match GPT-generated content, especially one without sufficient "prompt engineering" to make it sound less like the default writing style. Nobody is impressed by "write a rap like a pirate" output anymore.

Ingestion is a lot less sexy and hasn't gotten nearly as much attention as creation. This is stuff like "summarize this document." And it's powerful. But people didn't get as hyped up on it because it's something that they felt like a computer was supposed to be able to do: transforming existing data from one format to another isn't revolutionary, after all.

But the world has a lot of unstructured, machine-inaccessible text. Legal documents saved in PDF format, consultant reports in Word, investor pitches in PowerPoint. And when I say "unstructured" I mean "there is data here that it is not easy for a machine to parse."

Being able to toss this stuff into ChatGPT (or other LLM) and prompt with things like "given the following legal document, give me the case number, the names of the lawyers, and the names of the defendants; the output must be JSON with the following schema..." and that save that information into a database is absolutely killer. Right now companies are recruiting armies of interns and contractors to do this sort of work, and it's time-consuming and awful.