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

HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

arxiv.org

Post reply on HN
Print view View original

121–130 of 280 posts

Re: HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

#121

Post by 93po »

Mizza wrote: post #70
Workaccount2 wrote: post #8

I strongly suspect the first AGI will come sooner than expected on the back of a "glue" AI that can intelligently bond together a web of narrow AIs and utilities. I got access to the wolfram plugin for chatGPT, and it turned it from a math dummy to a math genius overnight. A small step for sure, but a hint of what's to come.

I've built a service for building systems like this. I haven't officially "launched" yet, but it's working and you can play with it here, if anybody is up to giving the alpha a spin: https://aistudio.fly.dev/

your "go pro" page is white text on white background in firefox

Re: HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

#122

Post by 93po »

Mizza wrote: post #70
Workaccount2 wrote: post #8

I strongly suspect the first AGI will come sooner than expected on the back of a "glue" AI that can intelligently bond together a web of narrow AIs and utilities. I got access to the wolfram plugin for chatGPT, and it turned it from a math dummy to a math genius overnight. A small step for sure, but a hint of what's to come.

I've built a service for building systems like this. I haven't officially "launched" yet, but it's working and you can play with it here, if anybody is up to giving the alpha a spin: https://aistudio.fly.dev/

other quick feedback: add a "add node" button, it isn't immediately clear you do this with right clicking

Re: HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

#123

Post by 93po »

Manjuuu wrote: post #71
karmasimida wrote: post #54

Earlier quoted context omitted.

How can you speak with such confidence, when Hinton and Ilya Sutskever can not? Hinton even said it is possible in next 5-10 years. While, you can believe whatever you believe, please don't bash people as laypeople or something as nonsense.

The same confidence of those already planning for the advent, that will never be, of AGI. I see a lot of unmotivated enthusiasm for the "new" thing, the opinions of Hinton should be considered valid if the reasoning makes sense, not just because the source is Hinton. No idea why he said that it's possible in 10 years. Not trying to bash anyone, I just meant normal people, outside of the field. Enthusiastic nonsense i…

To say we won't have AGI in our lifetime as a certainty means that you must be able to say with a certainty how AGI is developed. Otherwise there is no evidence to point to as to why the steps towards developing it aren't possible.

Re: HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

#124

Post by Donald »

MichaelRazum wrote: post #119
Workaccount2 wrote: post #8

I strongly suspect the first AGI will come sooner than expected on the back of a "glue" AI that can intelligently bond together a web of narrow AIs and utilities. I got access to the wolfram plugin for chatGPT, and it turned it from a math dummy to a math genius overnight. A small step for sure, but a hint of what's to come.

"math genius" not sure about it. I gave it a problem from IMO 2022. In my opinion not a very hard one. It failed even if you give it some hints. Let R+ denote the set of positive real numbers. Find all functions f : R+ → R+ such that for each x ∈ R+, there is exactly one y ∈ R+ satisfying xf(y) + yf(x) It just struggeld to reason. So I would be very surprised if the plugin somehow helps here.

It's a language completion model, so it's not surprising at all that it struggles with logical inference. That's probably one of the reasons that instruction fine-tuning has such a dramatic effect in the performance of these models: they're finally given some of the underlying causal priors of the task domain.

Re: HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

#125

Post by 93po »

nemo44x wrote: post #105
maCDzP wrote: post #4

Am I the only one with an uneasy feeling about all these advancements? Are we going fast or is it just because of the buzz? I have a hard time separating the two. It’s very exiting and I want to keep being exited. I don’t want to become terrified.

I think we are at, or very near, the inflection point in that picture of that graph that is rising linearly for centuries and then suddenly goes exponential. https://waitbutwhy.com/wp-content/uploads/2015/01/G1.jpg

the problem is that the chart is also accurate if the x axis starts 100,000 years ago. the inflection point might be 2000 years or it might be 50 years.

Re: HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

#126

Post by Workaccount2 »

dvt wrote: post #72
Workaccount2 wrote: post #67

Earlier quoted context omitted.

My definition of "mathematical genius" is simply wider than yours. To me its people who can solve math problems that 99.9% of the population can't without assistance. Which I think is a fair colloquial definition. ChatGPT went from struggling to provide the answer for 20x20 to easily being able to provide the right answer for any math problem wolfram alpha can.

> To me its people who can solve math problems that 99.9% of the population can't without assistance. You're just kind of re-emphasizing my point: ChatGPT is using Wolfram as its assistance. So really, it's acting more like a "dumb" API call, not a "math genius" at all.

It goes back to my original point, that I suspect AGI will come from an AI that is essentially a master of all APIs.

We can go back and forth splinting hairs about whether inserting a compute module into a neural net (organic or not) grants geniousness or assistance, but the overall point stands; there will be a single interface that can take any variety of inputs and properly parse and shape them, push them through the correct "APIs", and then take the results to form a clear and correct output. Whether or not it used it's neural net or transistor adders circuits to arrive at the answer would be immaterial.

Re: HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

#127

Post by amelius »

cameronfraser wrote: post #111
amelius wrote: post #27

Can we please stop generating bounding boxes and instead generate a proper segmentation using bitmasks?

it depends on the problem?

True, but a bitmask is always better than a rectangle. And the computational resources required for this problem are not very large compared to other AI workloads, so it should be a no-brainer from a usefulness viewpoint.

Re: HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

#128

Post by Workaccount2 »

IanCal wrote: post #88
_yb2s wrote: post #65

Earlier quoted context omitted.

You are right, but it is unclear to me where this limitation will fall in terms of actual abilities. GPT-4 can divide large tasks up into logical steps and instructions that can be worked on independently by other instances, and it can create "compressed" condensed explanations of it's own state that can be stored externally and repeated back to it, or passed back and forth between instances. With the unreleased 32,0…

> With the unreleased 32,000 token context window, that is really a lot of context when you consider that it can heavily compress things by referencing what it was trained on. Also that means you can just iterate with different contexts until you can deal with the problem. How many problems need 50 pages of context on top of what gpt4 knows for solving the next step?

>How many problems need 50 pages of context on top of what gpt4 knows for solving the next step?

Perhaps the unknown unkowns?

I envision a scenario where a super intelligent AI seemingly runs off the rails and becomes obsessed with overtly complex problems that are totally intractable to humans. Where we'd be like ants trying to make sense of cell phones.

Re: HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

#129

Post by Workaccount2 »

roflyear wrote: post #58
_yb2s wrote: post #28

Earlier quoted context omitted.

GPT-4 can do this. I realize that a description of how these models work implies that this is absolutely impossible , so I have no explanation other than suggesting you go try it out for yourself.

Bro, GPT-4 can absolutely not understand new information. Try to teach it a grammar. It can't learn a grammar. Anyone can learn a new grammar. Not GPT-4. GPT-4 does not ask clarifying questions. It does not understand things.

I don't know what your bar for clarifying question is, but I have definitely had it ask me for more info when I forget to put it into the prompt.

Re: HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace

#130

Post by fatherzine »

baq wrote: post #103
catchnear4321 wrote: post #74

A lot of LLM/ AI research feels like a big “duh.” The language model, being trained on essentially the primary way in which humanity communicates, might be a good means of managing integration of less language-focused models. …duh?

animals which use tools: great apes (gorillas, chimpanzees, orangutans), monkeys, otters, dolphins, some bird species, octopus, and a few crocodilian species. now LLMs are on this list. a thing which isn't born, doesn't experience, can be copied and instantiated a million times, and a single improvement can be basically immediately propagated to all instances. a very unfamiliar thing with very familiar capabilities.…

"We'll be dealing with it for the rest of our civilization's lifetime." Exactly.
Post reply on HN