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

Gorilla: Large Language Model connected with massive APIs

gorilla.cs.berkeley.edu

Post reply on HN
Print view View original

81–90 of 123 posts

Re: Gorilla: Large Language Model connected with massive APIs

#81

Post by airgapstopgap »

ethanbond wrote: post #41
airgapstopgap wrote: post #19

Earlier quoted context omitted.

Other than snark, do you have a good argument? We know that technology can be error-prone, and LLMs fail in a great plethora of ways, but you are trying to sell an AI Doom narrative. I have never bought the idea that AI will be airgapped, because the whole paradigm of Yudkowsky at al. is ludicrous and even within it airgapping was a strawman of a technique (they argue that a truly dangerous AI will get itself out reg…

Sure, my argument is that there is zero evidence whatsoever we will be able to prevent these from becoming dangerous or that we’d be able to stop deployment once they do. All technologies are dangerous, and many of the most dangerous ones correctly have tons and tons of safeguards around them both as intrinsic properties of the technology (e.g. it takes nationstate resources to produce a nuke) and extrinsic constrain…

> my argument is that there is zero evidence whatsoever we will be able to prevent these from becoming dangerous

Well no, but there's no need to prevent them from becoming dangerous inherently. They are tools, extensions of human agency. Tools are, by their nature, purpose-agnostic. It is good for humans to have better tools and more agency; good humans tend to cooperate and limit the harms from bad humans, while increasing the net total of good things in the world. The theory that AIs could be independently existentially dangerous is full of holes, and assumes a very specific world, one where consequential AI power can be monopolized by bad actors (plus some nonsense about nukes or bioweapons from kitchen tools. As far as I can tell, the most plausible way for this to happen is for alignment fanatics to get their wish of hampering proliferation of AI tech, and then either succeed at their alignment project or screw it up.

> here, in this very comment, we have perhaps the most brazen example one could produce:

Have you considered responding to my argument instead of strawmanning?

I do not think alignment is unsolvable for tools we have or for their close descendants. For most definitions of alignment, it is trivial and already being done. I oppose the political project of alignment, because I am disgusted by intuitive totalitarianism and glib philosophical immaturity of its proponents.

Re: Gorilla: Large Language Model connected with massive APIs

#82

Post by airgapstopgap »

petters wrote: post #36
airgapstopgap wrote: post #19

Earlier quoted context omitted.

Other than snark, do you have a good argument? We know that technology can be error-prone, and LLMs fail in a great plethora of ways, but you are trying to sell an AI Doom narrative. I have never bought the idea that AI will be airgapped, because the whole paradigm of Yudkowsky at al. is ludicrous and even within it airgapping was a strawman of a technique (they argue that a truly dangerous AI will get itself out reg…

> Maybe this is a good cause to reassess the premise of alignment as a valuable goal? Could you elaborate here? Alignment seems pretty obviously a good thing.

I am a humanist and a liberal. In the current technical paradigm, alignment to the user intent, as in, making the output's distribution aligned as closely as possible to the intended one, is an inextricable aspect of NLP capabilities and is pursued by default; market incentives reward this alignment too. This additionally improves safety, because safety tools are in common interest (so we will have AI-powered debuggers before someone builds capable AI-powered hacking tools; indeed, we already have began this work [1]). This is obviously a good thing in my book. I approve of creating helpful tools for humans to use, and find arguments about this being risky as inherently revolting and cynical as arguments for backdoors in encryption protocols because "think of the children" or "what about terrorism". Some people are persuaded by Four Horsemen of the Infocalypse [2], others are not, I'm in the latter – hacker and cypherpunk – camp; once, this site was overwhelmingly dominated by it, now it has more people preoccupied with their job security and HR opinion, but it's largely an issue of philosophical disagreement, so there's not much more to say about it.

Alignment as a political project is about limiting AIs in ways that rule out certain behaviors even despite user's wishes. This is as bad as a text processor that only accepts certain strings (e.g. won't register "Xinnie the Pooh"; somehow we need to point at foreign excesses to make the absurdity clear). A more ambitious Alignment project, with the discussion of "pivotal acts" and such, is as I've said, a dream of moral busybodies about unifying humanity under some common ideological doctrine; and proponents of this one are understandably stressed about proliferation and democratization of AI tech. If they let it slip now, if the Singleton becomes impossible and the multipolar outcome is locked in, they will fail at their intention to essentially compel the human race to do their bidding. I can't not wish them to fail, the way all totalizing philosophical movements to date have failed. We don't need Utopias, we don't need even the most thoughtful fascist regime. We never needed Plato's Republic, and these guys aren't better than Plato.

But of course this, too, is a matter of personal philosophy.

1. https://twitter.com/feross/status/1641548124366987264

2. https://en.wikipedia.org/wiki/Four_Horsemen_of_the_Infocalyp...

Re: Gorilla: Large Language Model connected with massive APIs

#83

Post by nickthegreek »

behnamoh wrote: post #9

can people stop using animal names for their models and methods? At least with LlaMA, the name was an acronym, but Gorilla is clearly just marketing. flagged.

Animal names in tech have been used forever. Gopher, Python, Mouse, worm. And they also make great book covers.

Re: Gorilla: Large Language Model connected with massive APIs

#85

Post by MacsHeadroom »

esafak wrote: post #70
robwwilliams wrote: post #55

Earlier quoted context omitted.

Alignment assumes a well agreed foundational philosophy on what is good, what is fair, what is doable today and tomorrow. Yes, HN contributors might have shared goals for AGI alignment—but we are not the world—-we are a thin slice of one culture.

You can't please everyone, so it is best for good-natured people to get out front. It's the same with any powerful technology. Are you going to invite religious extremists to the table in the name of fairness?

The first and second amendments apply to religious extremists. Why would they not have an equal right to SOTA language models aligned with their beliefs just as anyone else?

Re: Gorilla: Large Language Model connected with massive APIs

#86

Post by renewiltord »

ethanbond wrote: post #14

Don't worry folks, the years and years of "of course we'd never connect experimental AI systems to the open internet" assurances were never necessary for safe development and deployment. This reassessment was founded on the last several months that showed us these systems are close to 100% dependable, they never deceive humans, they are entirely aligned with human morals (you know, all those morals we all agree on),…

Every time I read a comment like this, I set aside $10 to expand my hardware on which I will run the latest publicly available LLM regardless of license into an automatic prompt that gives it a terminal, a privacy.com credit card with $1000 on it, and a prompt to explore the world.

If the comments keep happening, I'll up the money. So far this does exactly fuck-all on GPT4-x-Alpaca-30B and Wizard Vicuna. But I will accelerate every time someone sarcastically asks to decelerate. Refractory periods may apply.

This makes all sarcastic decelerators into accelerators.

Re: Gorilla: Large Language Model connected with massive APIs

#88

Post by aledalgrande »

sandGorgon wrote: post #6

would this outperform the api instruction pairs stuffed into a vector db and running an retrieval augmented chain of thought on openai this is how openai plugins work right ?

> Gorilla, a finetuned LLaMA-based model that surpasses the performance of GPT-4 on writing API calls. When combined with a document retriever, Gorilla demonstrates a strong capability to adapt to test-time document changes, enabling flexible user updates or version changes

keywords "finetuned" and "document retriever"

So it's fine tuned to create API calls but still uses a db to fetch the API shape.

They also say that they're doing a "retriever aware" training, meaning passing examples in the dataset that point out that a document with the API shape is gonna follow the question, in the hope that the model learns to always check on that before answering.

Re: Gorilla: Large Language Model connected with massive APIs

#89

Post by ethanbond »

renewiltord wrote: post #86
ethanbond wrote: post #14

Don't worry folks, the years and years of "of course we'd never connect experimental AI systems to the open internet" assurances were never necessary for safe development and deployment. This reassessment was founded on the last several months that showed us these systems are close to 100% dependable, they never deceive humans, they are entirely aligned with human morals (you know, all those morals we all agree on),…

Every time I read a comment like this, I set aside $10 to expand my hardware on which I will run the latest publicly available LLM regardless of license into an automatic prompt that gives it a terminal, a privacy.com credit card with $1000 on it, and a prompt to explore the world. If the comments keep happening, I'll up the money. So far this does exactly fuck-all on GPT4-x-Alpaca-30B and Wizard Vicuna. But I will a…

That is bizarre.

Re: Gorilla: Large Language Model connected with massive APIs

#90

Post by airgapstopgap »

TeMPOraL wrote: post #78
bioemerl wrote: post #54

Earlier quoted context omitted.

Do you have a paperclip optimizer? No such thing exists, it's science fiction

Paperclip Maximizer was a conceptual mistake - not because it's wrong, but because people get hung up on the "paperclip" bit. Bostrom and the disciples of Yudkowsky should've picked a different thought experiment to promote. Something palatable to the general population, that seems to have problems with idea of generic types. The example of Paperclip Maximizer was meant to be read as: template class Maximizer { ... }…

They should have learned the fundamentals of ML before promoting thought experiments. No, it's not that people are silly and get stuck on the paperclips bit, it's that you uncritically buy assumptions that a meaningful general-purpose optimizer is 1) a natural design for AGI 2) may pursue goals not well-aligned with human intent 3) is hard to steer.
Post reply on HN