03 is actually orthogonal to AGI and ASI in a cartesian sense. My SAS startup led multiple qualified teams where our RAG implementations on synthetic data originated positive inference in line with the literature (1).
(1) Sparks of AGI paper
GPT-5 is behind schedule
311–320 of 1001 posts
Re: GPT-5 is behind schedule
#312[dead]
Re: GPT-5 is behind schedule
#313Why would I expect a software project to ship on time?
Re: GPT-5 is behind schedule
#314Earlier quoted context omitted.
a) There is evidence e.g. private data deals that we are starting to hit the limitations of what data is available. b) There is no evidence that LLMs are the roadmap to AGI. c) Continued investment hinges on their being a large enough cohort of startups that can leverage LLMs to generate outsized returns. There is no evidence yet this is the case.
"There is no evidence that LLMs are the roadmap to AGI." - There's plenty of evidence. What do you think the last few years have been all about? Hell, GPT-4 would already have qualified as AGI about a decade ago.
The last four years?
ELIZA 2.0
Re: GPT-5 is behind schedule
#315Behind schedule? It never fucking ships. Two months ago, and four, and six I got dog piled for saying it doesn’t ship. It doesn’t ship. You guys can’t do it! Prove me wrong! https://news.ycombinator.com/item?id=42014054
Maybe you were dog-piled because OpenAI will ship a successor to GPT-4o someday, whatever it's called. In any case, the "behind schedule" rumors are themselves based on other rumors. GPT-2→GPT-3 took 5 quarters, GPT-3→GPT-4 took 11 quarters, so obviously GPT-5 (or its equivalent) will be released in Q4'2025.
It’s a pretty easy option when I’m selling calls with infinity expiration at zero strike.
I’m still selling them. That’s how utterly convinced that this particular mix of my former colleagues are incapable of this thing I am.
They’ll call something GPT-5, but it won’t like obsolete physicists or even good hackers. None of the shit sama says.
Even o1 sucks.
Re: GPT-5 is behind schedule
#316Earlier quoted context omitted.
"There is no evidence that LLMs are the roadmap to AGI." - There's plenty of evidence. What do you think the last few years have been all about? Hell, GPT-4 would already have qualified as AGI about a decade ago.
> GPT-4 would already have qualified as AGI about a decade ago. Did you just make that up?
Probably asked an "AI"
Re: GPT-5 is behind schedule
#317Re: GPT-5 is behind schedule
#318Earlier quoted context omitted.
Maybe you were dog-piled because OpenAI will ship a successor to GPT-4o someday, whatever it's called. In any case, the "behind schedule" rumors are themselves based on other rumors. GPT-2→GPT-3 took 5 quarters, GPT-3→GPT-4 took 11 quarters, so obviously GPT-5 (or its equivalent) will be released in Q4'2025.
It’s a pretty easy option when I’m selling calls with infinity expiration at zero strike. I’m still selling them. That’s how utterly convinced that this particular mix of my former colleagues are incapable of this thing I am. They’ll call something GPT-5, but it won’t like obsolete physicists or even good hackers. None of the shit sama says. Even o1 sucks.
Facebook and Google both had a bunch of brilliant people. And in 2016-2017 some of the real legends worked at OpenAI.
And then Sam got control and all the high profile people bounced, and it’s a pretty grim residue that either lasted or rushed in.
I know these people, a lot of them personally, and they are not the person you trust with fucking anything.
Re: GPT-5 is behind schedule
#319Earlier quoted context omitted.
The new idea is inference-time scaling, as seen in o1 (and o3 and Qwen's QwQ and DeepSeek's DeepSeek-R1-Lite-Preview and Google's gemini-2.0-flash-thinking-exp). I suggest reading these two pieces about that: - https://www.aisnakeoil.com/p/is-ai-progress-slowing-down - best explanation I've seen of inference scaling anywhere - https://arcprize.org/blog/oai-o3-pub-breakthrough - François Chollet's deep dive into o3 I'…
I think the wildest thing is actually Meta’s latest paper where they show a method for LLMs reasoning not in English, but in latent space https://arxiv.org/pdf/2412.06769 I’ve done research myself adjacent to this (mapping parts of a latent space onto a manifold), but this is a bit eerie, even to me.
There are lots of papers that do this
Re: GPT-5 is behind schedule
#320Earlier quoted context omitted.
Do we know LLMs are the path to AGI? If they're not, we'll just end up with some neat but eye wateringly expensive LLMs.
AGI will arrive like self driving cars. it’s not that you will wake up one day and we have it. cars gained auto-braking, parallel parking, cruise control assist. and over a long time you get to something like waymo, which still is location dependent. i think AGI will take decades but sooner will be some special cases that are effectively the same
Interesting idea. The concept of The Singularity would seem to go against this, but I do feel that seems unlikely and that a gradual transition is more likely.
However, is that AGI, or is it just ubiquitous AI? I’d agree that, like self driving cars, we’re going to experience a decade or so transition into AI being everywhere. But is it AGI when we get there? I think it’ll be many different systems each providing an aspect of AGI that together could be argued to be AGI, but in reality it’ll be more like the internet, just a bunch of non-AGI models talking to each other to achieve things with human input.
I don’t think it’s truly AGI until there’s one thinking entity able to perform at or above human level in everything.