←back to thread

OpenAI, Google and Anthropic are struggling to build more advanced AI

(www.bloomberg.com)
625 points lukebennett | 6 comments | 13 Nov 24 13:28 UTC | HN request time: 0.424s | source | bottom
Show context
fallat ◴[14 Nov 24 19:36 UTC] No.42140239[source]▶
What a stupid piece. We are making leaps every 6 months still. Tell me this when there are no developments for 3 years.
replies(1): >>42141347 #
1. hatefulmoron ◴[14 Nov 24 21:20 UTC] No.42141347[source]▶
I'm curious, what was the leap after GPT-4? What about the leaps after that, given a leap every 6 months?
replies(3): >>42142714 #>>42145027 #>>42145103 #
2. Der_Einzige ◴[15 Nov 24 00:14 UTC] No.42142714[source]▶
Sora was just one of the many…
replies(1): >>42142744 #
3. hatefulmoron ◴[15 Nov 24 00:19 UTC] No.42142744[source]▶
Your best example is something that doesn't even do the things that GPT-4 does, isn't available to use, and has seemingly only produced a few clips (some of which were edited).

If it were one of many, I think you would name something better.

4. zaptrem ◴[15 Nov 24 08:45 UTC] No.42145027[source]▶
O1, new Sonnet, all the music models and video models, the voice models like 4o, etc.
replies(1): >>42145866 #
5. epups ◴[15 Nov 24 08:59 UTC] No.42145103[source]▶
Some important landmarks since GPT4 was first released (not in chronological order):

- Vast cost reduction (>10x)

- Performance parity of several open source models to GPT4, including some with far fewer parameters

- Much better performance, much larger context window in state-of-the-art closed source LLMs (Claude 3.5 Sonnet)

- Multimodality (audio and vision)

- Prototypes for semi-autonomous agents and chain-of-thought architectures showing promising avenues for progress

6. hatefulmoron ◴[15 Nov 24 11:17 UTC] No.42145866[source]▶
The music/video models are cool, but It's an apples to orange comparison with GPT-4. I don't think there's really any comparison of intelligence or "advanceness" between those models and GPT-4.

I'm surprised to hear someone say that O1 and new Sonnet are "leaps", though. My impression of them is that they're qualitatively similar to GPT-4. Incremental improvements at best. I don't think the gap between GPT-4 and the new Sonnet is anywhere near as large as the gap between GPT-3 and GPT-4, for instance.