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

ChatGPT turned generative AI into an “anything tool”

arstechnica.com

Post reply on HN
Print view View original

41–50 of 245 posts

Re: ChatGPT turned generative AI into an “anything tool”

#41

Post by Kerb_ »

Sohcahtoa82 wrote: post #23
atemerev wrote: post #3

This is what we used to call an "AGI" back in 2013-2014. Generalized artificial intelligence. An AI which is general enough to be used for diverse tasks in multiple fields, unlike specialized AIs of that time (e.g. for image recognition or self-driving cars). Nothing more than that. Now the goalposts are moved way forward, and now "AGI" is something that have conscious, with superhuman intelligence, and probably evil…

> But back in a day, AGI was just what we have today -- an universal chat bot. I don't know anybody that used AGI to mean that. I've always thought of AGI as an AI that can learn how to perform tasks with very simple instructions, rather than spending weeks training a neural network on a data center full of GPGPUs. Basically, my idea of AGI has always been what you're calling the "new" definition of AGI.

Remember that people have decades of generalized training before becoming able to perform tasks with very simple instructions. I would consider years of pre-training to be acceptable if the model is usable and trainable quickly in practice as well

Re: ChatGPT turned generative AI into an “anything tool”

#43

Post by spott »

assistantist wrote: post #35
spott wrote: post #28

Earlier quoted context omitted.

If it was just computer that caused the various assistants to get bad, then why haven’t they gotten good again? Google assistant came out in 2016! That is 7 years ago! Since then, GPU compute in flops has 5xed… for single precision. If you include half-floats (not a thing back then), it is 75x the single precision flops. I think the reason that assistants have “gotten bad” is a combination of edge case controls (avoi…

If it was just a matter of flops then AI would already be driving cars and making door to door food deliveries. The problem is neither the energy expenditure nor the compute capacity. No one knows what the proper architecture should be for general purpose intelligence assistants so everyone has to fine-tune the existing models on their specialized use cases and even then it's a hit or miss depending on how much train…

I think you are missing my point.

The OP said Siri et al. got worse because their creators optimized them to reduce compute costs. I’m saying that if that were true, then they should be at least as good as they were because you can get tons more compute for cheaper these days. So it isn’t compute that makes the various assistants “worse” than when they were introduced.

I made no statements about their quality in general, just their relative quality compared to when they were released.

Re: ChatGPT turned generative AI into an “anything tool”

#44

Post by e_y_ »

atemerev wrote: post #3

This is what we used to call an "AGI" back in 2013-2014. Generalized artificial intelligence. An AI which is general enough to be used for diverse tasks in multiple fields, unlike specialized AIs of that time (e.g. for image recognition or self-driving cars). Nothing more than that. Now the goalposts are moved way forward, and now "AGI" is something that have conscious, with superhuman intelligence, and probably evil…

I would call a chat bot a specialized application. And I dispute that a chat bot was ever considered the end goal of AGI. The idea is that if we had AGI, we could build a chat bot with it to test/demonstrate its general reasoning capabilities. Simply being effective at chatting is not sufficient to be called an AGI. And today's chatbots are quite stupid.

Sure we're starting to give chat bots the ability to query information and even interpret user-provided images, but AFAIK the image interpretation is done using a different model rather than being an extension of the LLM.

Re: ChatGPT turned generative AI into an “anything tool”

#45

Post by josefresco »

CamperBob2 wrote: post #19
josefresco wrote: post #14

Tried using Bing ChatGTP to create a logo for me because I saw a cool demo on Instagram. It couldn't write text correctly and I tried numerous times. Not only that, the logos were terrible and didn't "mimic" the image I uploaded as a guide (which was in the demo) Today I needed to transcribe an MP3. Nope, can't do that either. However on Monday I had it write some PHP and that was neat! 50/50 for me.

For logos you'd be better off with DALL-E 2, no? How do you even get a useful graphical logo out of a text-based language model?

Bing chat uses DALL-E for image generation. The results were complete gibberish. Characters that looked like an alien alphabet and the graphic was also so abstract to not represent anything.

Re: ChatGPT turned generative AI into an “anything tool”

#46

Post by ChatGTP »

Kerb_ wrote: post #41
Sohcahtoa82 wrote: post #23

Earlier quoted context omitted.

> But back in a day, AGI was just what we have today -- an universal chat bot. I don't know anybody that used AGI to mean that. I've always thought of AGI as an AI that can learn how to perform tasks with very simple instructions, rather than spending weeks training a neural network on a data center full of GPGPUs. Basically, my idea of AGI has always been what you're calling the "new" definition of AGI.

Remember that people have decades of generalized training before becoming able to perform tasks with very simple instructions. I would consider years of pre-training to be acceptable if the model is usable and trainable quickly in practice as well

Mostly “training” to do abstract things, not really for survival or fun.

I could catch fish since a fairly early age, mostly because I enjoyed it.

At about age 12 I used to drive stolen cars around without any training at all except observation.

It too me decades to become a corporate robot though :)

I think training is a weird word, life is an experience, not just training. I think your comment demonstrated the limitations of language to describe things.

Re: ChatGPT turned generative AI into an “anything tool”

#47

Post by famouswaffles »

e_y_ wrote: post #44
atemerev wrote: post #3

This is what we used to call an "AGI" back in 2013-2014. Generalized artificial intelligence. An AI which is general enough to be used for diverse tasks in multiple fields, unlike specialized AIs of that time (e.g. for image recognition or self-driving cars). Nothing more than that. Now the goalposts are moved way forward, and now "AGI" is something that have conscious, with superhuman intelligence, and probably evil…

I would call a chat bot a specialized application. And I dispute that a chat bot was ever considered the end goal of AGI. The idea is that if we had AGI, we could build a chat bot with it to test/demonstrate its general reasoning capabilities. Simply being effective at chatting is not sufficient to be called an AGI. And today's chatbots are quite stupid. Sure we're starting to give chat bots the ability to query info…

LLMs aren't chatbots. That's just a nice consequence of what they can do.

Their prediction abilities stretch beyond just language even if they're only trained on it.

https://general-pattern-machines.github.io/

There's nothing specialized about being able to predict meaningful completions from arbitrary token sequences.

There's a whole lot that can be done with that even if they were constrained to language.

https://tidybot.cs.princeton.edu/ https://innermonologue.github.io/

There's nothing stupid about GPT-4. There's no testable definition of general intelligence GPT-4 would fail that a chunk of humans also wouldn't.

Vision understanding does not have to be a separate model. https://arxiv.org/abs/2306.14824

Re: ChatGPT turned generative AI into an “anything tool”

#48

Post by ColonelPhantom »

Someone1234 wrote: post #2

Although let's see if it is sustainable. I pay for ChatGPT Plus but most people aren't, and from what I've read they're losing money like crazy. We may wind up seeing this as the technical high point before the various competitors cut back on expensive hardware/energy usage for the free tier, and the quality of the responses takes a dive. This might sound a bit conspiratorial so apologies for that but back when the c…

One thing about LLMs is that from what I know, batching is very effective with them, because to generate a token you need to basically go through the entire network for just that one sample. If you batch, you still need to stream the entire network, but that doesn't get any more expensive; you use each piece of data more often on hardware that used to just sit idle.

So I assume that at an OpenAI-scale, they're more than able to batch up requests to their models, which gives a tiny latency increase (assuming they're getting many requests per second), but massively improves compute utilization.

Re: ChatGPT turned generative AI into an “anything tool”

#49

Post by vunderba »

josefresco wrote: post #45
CamperBob2 wrote: post #19

Earlier quoted context omitted.

For logos you'd be better off with DALL-E 2, no? How do you even get a useful graphical logo out of a text-based language model?

Bing chat uses DALL-E for image generation. The results were complete gibberish. Characters that looked like an alien alphabet and the graphic was also so abstract to not represent anything.

Generative image systems are notoriously poor at producing legible text without significant rerolls. You'd be far better off generating a textless logo and then layering text afterwards using PS, Photopea, Krita, etc,

Re: ChatGPT turned generative AI into an “anything tool”

#50

Post by Madmallard »

Someone1234 wrote: post #2

Although let's see if it is sustainable. I pay for ChatGPT Plus but most people aren't, and from what I've read they're losing money like crazy. We may wind up seeing this as the technical high point before the various competitors cut back on expensive hardware/energy usage for the free tier, and the quality of the responses takes a dive. This might sound a bit conspiratorial so apologies for that but back when the c…

We need a strong localized version at least for software development that is free as soon as possible. It is such a competitive advantage there's no way it will stick around forever for as cheap as it is.
Post reply on HN