Ask HN: Are the Google AI releases lacking something?
Comparing OpenAI’s release of GPT-4o where they released ONE thing that blew everyone’s minds, vs Google’s release of a wide range of models that all feel sort of underwhelming… does anyone else feel like Google doesn’t have the required spirit or corporate structure to make one great thing?
24 comments
[ 374 ms ] story [ 1947 ms ] threadIf you're expecting them to sell out and pitch "alignment" bullshit like sama, you're looking at the wrong company. But they seem to be doing pretty well, and I'd argue that even Apple is jealous of how much foresight Google's AI strategy has.
This is wrong. But it explains why you think that Google are crushing it if you have no idea about the company.
- OpenAI stopped releasing public models a long time ago; their most recent local models are borderline useless (gpt-neo/gpt-j)
- Meta is also performing quite well but relies on third-party training hardware and building a lot less in-house
- Apple is forcing their users down CoreML or bust, apparently with the expectation of busting if the OpenAI deal rumors are true
- Microsoft is just really happy to be here and wants you to know they support this whole "AI" idea
So... who else is left? Projects like Falcon or Mistral are cool but it's hard to say they're "crushing it" when they're effectively being crushed by a lack of funding. The only business that has unshacked themselves from Nvidia to produce actually open, actually usable local models is Google. I don't like Google as a business, but if you look at them and say "they could do better" then I'm really curious who your better example is.
https://blog.google/technology/developers/gemma-open-models/
Google’s failures will be celebrated loudly and for years. Google is like Apple for services.
Many shareholders aren't even that sophisticated. They come from mining, property, factories, farms, logistics. They see AI and a brand name, they buy.
These companies are also terrible to work in - what the shareholders want and the CEO they pick often don't align to the reality on the ground. A good CEO will bullshit shareholders just enough to keep the people on the ground working. What you see on the outside is likely not targeted at people like us.