The landgrab going on in this space is ferocious. If you weren't convinced this is mania, and a 10M "seed" round for Langchain doesn't do it for you, nothing will. Well done Harrison on the cash grab (take as much money off the table every round as you can), it's a smart move. But I can't shake the feeling that this ever increasing mania will sweep up anything OSS with vague traction and this whole AI space that used to be religiously defaulted to open and sharing will fairly quickly end up in VC-funded fiefdoms with pay-to-play being all that's left modulo the "forever free" community hobbled versions. Hope I'm wrong.
Interestingly this is shortly after another huge “seed” round in this space from fixie.ai (17mm). A lot of money being thrown at making it easier to chain/give LLMs access to other applications and tools.
Yeah, wow, a 17M seed round and not a hint of irony. Ferocious. Capital is DESPERATE to throw money into anything that appears to be related to LLMs. "Value creation event of our lifetimes" etc. There's grift money to made here and I'll be damned if some decent proportion of the "API wrapper + template UI" startups masquerading as AI companies aren't cashing in. Not sure I blame them.
I agree with you, and people miss that December 2022 is not when GPT became commercially possible.
Look at the startup OpenAI generations 1-2 years back - they have largely sank at comparable or worse rates than any other startup from 2020/2021.
The GPT-3 first wave companies, around translations, basic quizzes, summarization tools, language learning apps, and ofcourse the notorious paraphrase tools (almost entirely obsolete since ChatGPT) can't be found in that form anymore, they've all been forced to shut down or move functionality significantly. Early on, OpenAI limited output to 300 tokens max - and less for most usecases, often 50-150. Chatbots were not allowed IIRC for over a year. If ChatGPT hadn't came along, much of what langchain enables wouldn't have either, nor so many big companies willing to now risk.
I can count none over 2 years old which have not since been made obsolete by raw ChatGPT access or are now default dead due to competition from existing unicorn (e.g. Duolingo, Quizlet Q chat) who are now crushing them.
It must be painful to have spent $xx,xxx on GPT-3 at $0.06 a token to obtain users, and now have your market ripped from you by a $B+ company paying $0.006...
So I doubt the template UI startups will sustain retention or stay about long term, unless they really find traditional startup ways to nuggle into niches and use cases/vendor lock in. This isn't an innovators dilemma for most companies, it's just an obvious sensible thing to try at this point, so startups don't have much to balance with on risk. That being the case, the market surely should seem less appealing than the open ended "blue ocean new value" of 2021 GPT tech, but - I guess not.
The steamroller is real in any super hot space, but that hazard is well known and founders should pay homage regularly to those who have fallen under its mighty squashing.
Python has huge uptake outside of the ML domain. I can't see an AI winter affecting the many people using Numpy for non-ML purposes (i.e., scientists in academia, most of whom still deal with normal numerical and data analytical modeling, not ML) much less Django.
I'm...aware. I think a lot of us that aren't using NumPy or ML stuff are rather looking for a good excuse to get away from what the ecosystem has become. (Yeah, I'm still bitter about Py3...Python user since the 1.5.2 days)
I thought the whole Python 3 thing was a huge problem. Lately I've been doing JS/Typescript dev and breaking changes like this happen continually and no one blinks.
The problem was never your code, really, it was the dependencies.
Wasn't that hard for quite a while to find situations where dependency A was Py3 compatible and B was not (and the incompatibility went both ways, especially in early 3.x releases, you could NOT have one codebase that worked with both).
Sometimes A dropped Py2 support before B gained Py3.
Pain pain pain.
Then add the increasing level of insanity as the answer to "python packaging sucks" was repeatedly to add yet another layer.
Plenty of other options out there if you've fallen out of love with Python and don't need good numerical libraries. Give JS, Elixir, or Crystal a try if you want something more dynamic. Nim if you want something a bit off the beaten path. Go, Rust, Java, Kotlin, and Scala if you want something more static.
I meant more like dynamically re-evaluating specific functions without having to restart the process. Haven't done much Erlang, but my experience is around doing so in lisp rather, which definitely can do that.
But this is by 'design', I have very unreasonable suspicions that indeed, the whole VC 'world of entrepreneurs' is just the way the USA government does R & D on an industrial-corporate scale. The 'brilliance' behind this way of doing R&D is that they only pick up the winners after they won, so they don't "waste" money on R&D death ends nor moonshots.
on the other hand, this is a good way to 'explode' for cheap the technological applications of already developed scientific innovations. meaning none of those VC-backed startups are doing innovative research, but in fact are devleoping commercial applications for corporate overlords who having seen who won, step in to buy them out.
even music industry is shifting to that model, they are now only signing bands/artists/influencers who already build their audience.
It's far superior to the EU model where politicians are tasked with creating requirements for ,,the next innovations'' and creating tenders, businesses popping up / recycled to win those tenders, do the minimum to satisfy the criterion that nobody really cares about, and try to siphon out as much money as possible using trusted subcontractors, meanwhile paying a huge part of the money for the privilage of winning the tender.
At least that's what's going on in Hungary, the most corrupt government in EU, I hope other parts are a bit better.
I'd invite you to read "Foundation Models and Fair Use"[1] which is a paper written as a collaboration between Standford's law school and computer science department.
It talks at length about this specific problem and migration techniques for it:
Existing foundation models are trained on copyrighted material. Deploying these models can pose
both legal and ethical risks when data creators fail to receive appropriate attribution or compensation.
In the United States and several other countries, copyrighted content may be used to build foundation
models without incurring liability due to the fair use doctrine. However, there is a caveat: If the model
produces output that is similar to copyrighted data, particularly in scenarios that affect the market
of that data, fair use may no longer apply to the output of the model. In this work, we emphasize
that fair use is not guaranteed, and additional work may be necessary to keep model development
and deployment squarely in the realm of fair use. First, we survey the potential risks of developing
and deploying foundation models based on copyrighted content. We review relevant U.S. case law,
drawing parallels to existing and potential applications for generating text, source code, and visual art.
Experiments confirm that popular foundation models can generate content considerably similar to
copyrighted material. Second, we discuss technical mitigations that can help foundation models stay
in line with fair use. We argue that more research is needed to align mitigation strategies with the
current state of the law.
Sounds like you're agreeing they are legally in murky territory.
Further, new laws get made in reaction to new things whenever they push an existing doctrine beyond the original ruling, and these are certainly in that territory.
> Sounds like you're agreeing they are legally in murky territory.
Of course. As I said originally "This is clearly not a given". It's very unclear how this will be decided, but anyone who thinks that just because models contain copyrighted data they don't have a leg to stand on is very wrong. There are multiple good arguments and precedents to show that they do, depending on the circumstances.
Huh, I still interpret “this is clearly not a given” as a being contradictory reply to me saying it’s murky territory.
I think they contain massive amounts of copyrighted data, and reproduce them exactly, and that’s why they don’t have a leg to stand on. It’s a personal opinion, and I think backed by your citation. But thanks for the reference there, and glad to chat.
Hey Naveed, this is a great project well done. I'm curious about the summarisation of longform content. You mention that you're using Langchain - are you using the MapReduce approach for documents that exceed the 32k context window, or some other approach? 600 pages at ~500 words/tokens a page would mean about $20 to mapreduce through a big doc, which seems crazy especially if you iterate on those 'summary' prompts. Or are you using embeddings for everything including summary prompts?
This is Searle's Chinese Room posit though right? The argument that there's no abstraction or internal modelling going on. Wish I could find the post I read recently that demonstrated some fairly clear evidence though that there IS some level of internal abstraction/reasoning going on in LLMs.
Do we allow for a matter of degree, rather than a binary, of "zero" vs" "complete" understanding?
At what point does abstraction and reasoning turn into "sentient understanding"? How do you express that as the thing doing the abstracting? Even humans struggle at that kind of task because there's a metaphysical assumption of first principles that these models seem to be challenging (or else this thread would not have started).
Processes such as back propagation are still not understood very well in the human brain. The brain certainly uses electrical impulses in order to transfer signals, much like the 1s and 0s in your computer or phone. The gap between us and intelligent machines is probably not as well understood or clear as most people in the software industry think it is.
> At what point does abstraction and reasoning turn into "sentient understanding"?
This is a loaded question, you're assuming that abstraction and reasoning can somehow magically "turn into" sentience, whereas I posit that those two things are completely different. You can have sentience without reasoning (i.e. pure non-judgemental awareness that is the goal of Buddhist meditation), and vice-versa.
Research into systemically important infrastructure cannot be damned because that infrastructure isn't public. It's a cheap moralizing argument to say "pfff, this was predictable". Maybe so, but there isn't an alternative. Much like research on Twitter. Once these companies start to drift into providing what become broadscale social utilities and public services it doesn't matter that they're private. There are(/should be) obligations that come with that.
You can't handwave and say go do your research on some micro-niche open source project that's way behind the SOTA and has nowhere near the same reach. That's not what "best practice" means here.
Replying to both responses because they're all good points. My argument boils down to the fact that some private companies end up becoming social utilities and once that happens, the rules (should) change as part of the social contract which means, yeah, they can't simply "pull the rug". The research is important precisely because its into systemically significant systems.
I get that it's difficult to define the line where that gets crossed. But the idea to provide a publicly funded trust that manages legacy versions of things like this is not a bad idea.
No matter how you define it, or whether people even agree companies should be obligated to provide certain public services, we are just nowhere near that line yet in this case, net even remotely close. It’s hand-wavy to say it’s important, but this is all brand new, there are only a handful of researchers involved, the critical mass to justify what you’re suggesting does not yet exist, it won’t for some time, and there’s no guarantee it ever will. I’m not sure what you mean by publicly funded trust, but that’s typically quite different from privately funded public services. Assuming that cost is even the reason here, then if someone wants to establish a trust and engage OpenAI, they can.
That said, what if OpenAI shut down codex because it has dangerous possibilities and amoral “researchers” started figuring out how to exploit them? What if it was fundamentally buggy or encouraging misleading research? What if codex was accidentally leaking or distributing export-controlled or other illegal (copyright, etc.) information? I’m explicitly speculating on possibilities, while you’re making unstated assumptions, so entertain the question of whether OpenAI is already doing a public service by shutting it down.
Feel free to elaborate, if you can. I gave you some added reasoning, so it doesn’t help anyone to flatly state disagreement without offering any justification. Why even bother to say you disagree?
What evidence is there that OpenAI’s codex has become a social utility? How many people used it to publish? Do you think the US government agrees? How likely is this case to go to court, and result in OpenAI being ordered to provide ongoing access to codex? That seems pretty far fetched to me, but I’m willing to entertain the possibility that I’m wrong.
Are you certain there aren’t problems with codex, that OpenAI isn’t working on something better, and/or shutting it down because it’s causing harm? If so, why are you certain?
Sure but OpenAI isn’t preventing research. It’s not their responsibility to provide reproducibility, at their expense, for any researchers looking at GPT, that job is the responsibility of the researchers, and the researchers still can work. It might be unfortunate from their perspective that there used to be a nice tool that makes their job easier, but the flip side here is that OpenAI didn’t say why they’re removing access to codex, and they probably have good reasons, not least of which is it costs them money that researchers aren’t subsidizing.
I'm going to be frank here, because I know my argument isn't "cheap". When one utilizes OSINT techniques (which using an ML service hosted by a third-party certainly qualifies as), there are baked-in assumptions that
1) this source could go away at any time, and
2) the source is only a reflection of the interests of the third-party, not something to be taken at face value.
No 2 can certainly be the subject of research, but to do so without accounting for No 1 would indicate bad research practices from the jump. For example, they could have (and should have) been snapshotting the outputs, tagged with versions & dates. By the sound of it, the outputs weren't even the subject of research, but were instead propping up the research. That flies in the face of No 2 as well. Let them start over, with better methodology this time.
Well, to me it sounds vague rather than fair. "Success" in what? In predicting new observations? Or in securing research grants? Or in letting people put the word "astonishing" in papers?
It just muddles things up to use qualitative terms in scholarly articles like that and it's standard advice to graduates to avoid it. And those who don't follow the advice have it repeated to them by reviewers. And that seems to be a good thing. Personally, I don't want to be told what is "astonishing", and, consequently, what isn't. I am perfectly capable of being astonished, or not, all on my own.
See, it's the "show, don't tell" principle. Astonish people, but don't tell them they're astonished, or they very likely won't.
> Betteridge's law of headlines is an adage that states: "Any headline that ends in a question mark can be answered by the word no." It is named after Ian Betteridge, a British technology journalist who wrote about it in 2009, although the principle is much older. It is based on the assumption that if the publishers were confident that the answer was yes, they would have presented it as an assertion; by presenting it as a question, they are not accountable for whether it is correct or not.
For the last few days I have been wondering if there is any analysis to see if this "law" is accurate.
Is there anything other than this?
> With 46% non-polar and 20% answered “yes”, at least two thirds of our headline sample violates Betteridge’s law. We conclude that it cannot be “mostly correct” either.
Hah, ChatGPT has successfully poisoned the well. Well done sama.
This lib is great work, a JS interface for running HF models. The comments about how "bad" the outputs are as surprising to me as they are alarming.
OAI has now set the zero-effort bar so high that even HNers (who click on .js headlines) fall into the gap they've left. That sucking sound you hear is market share being hoovered up.
No they're not mate, it's just you. I've read the guidelines (thanks for helpfully linking them). I see this on HN, people infer offense and cite the book rather than engage.
By not highlighting what you found "snarky" your response is a definitional "shallow dismissal". I see you just "picked the most provocative thing to complain about". Not a lot of being "kind" either.
So you know what would also be great? If you held yourself to the standards you're keen to police around here.
They've not shipped this. They're planning to. You're reading a marketing piece about work in the pipe. Like so, so many other company product "launches" in this space that are "already testing with a small group of whatevers".
Microsoft employs somewhere above 200k people. 10k is only a 5% internal rollout. That seems entirely doable without any fanfare.
The numbers are also completely made up, used to illustrate the point. Launching a product to X users is different than scaling up to X*1000 users, if that makes you feel better.
FWIIW, This seems rather similar to Github Copilot in VS Code. Been using it for a year, training data issues aside the product is absurdly good at times.
