From the first few paragraphs it doesn't pass the sniff test for me.<p>"Now AI has made everything more complex!" "AI is embedded in everything we do"...<p>Sounds like marketing gibberish and obfuscation, combined with self promotion.<p>That's just my read at first sniff.
Lots of folks working on open-source reasoning models trained with reinforcement learning right now. The best one atm appears to be Alibaba's 32B-parameter QwQ: <a href="https://qwenlm.github.io/blog/qwq-32b-preview/" rel="nofollow">https://qwenlm.github.io/blog/qwq-32b-preview/</a><p>I also recently wrote a blog explaining how reinforcement fine-tuning works, which is likely at least part of the pipeline used to train o1: <a href="https://openpipe.ai/blog/openai-rft">https://openpipe.ai/blog/openai-rft</a>
Is it me or the first link on the site <a href="http://ability.openai" rel="nofollow">http://ability.openai</a> is broken? Is openai a tld now?
There was a Berman video on it earlier today that summarises it<p><a href="https://www.youtube.com/watch?v=-haWhgmUheA" rel="nofollow">https://www.youtube.com/watch?v=-haWhgmUheA</a><p>Detail starts about 7mins in
> has claimed that the main techinique behinds o1 is the reinforcement learining.<p>Typos in the first sentence of the paper doesn't give confidence that I am about to read something worthwhile.
I guess now the strategy of OpenAI would be to keep the small edge all the time, integrate it with businesses fast & possibly kickstart new businesses by supporting them and trying to be synonymous with the best in AI (may be with Deepmind). I cannot think of any other moat, unless somehow they have a lot of proprietary and useful data (like in company) that others cannot replicate
many people are dismissing this paper because it has errors in spelling and grammar.<p>this is a terrible heuristic for evaluating AI papers. If you use it, you will miss a lot of good work by very strong researchers with below-average English writing skills.<p>I have not read this paper carefully so claim nothing one way or the other about its quality. It superficially seems like a pleasant and timely survey although a little flag-planty.
First line in the abstract<p>> OpenAI o1 represents a significant milestone in Artificial Inteiligence,<p><i>Inteiligence</i><p>Safe to say OpenAI has nothing to worry about
It seems to me that for the most capable and useful models, openness almost exclusively benefits businesses, or maybe academic organizations with money for serious hardware. I know what I can run on my 4090 at home but the results pale in comparison to the commercial services. I see why people consider these matters important from a theoretical standpoint but from a practical standpoint it doesn’t seem particularly consequential. I self-host a few FOSS server applications that are primarily sold as SaaS subscriptions, and folks are often very critical of those businesses benefitting from the “open source” label because they’re often seemingly deliberately difficult to self-host. This seems to be an order of magnitude less open than that. Is there some use case for people with reasonable hardware that I’m just not aware of?