alt Hacker NewsI have _long_ said don't sleep on Google for precisely these reasons. They have the only vertically integrated AI stack, insane capital they can deploy, and attention and focus and phenomenal architects and engineers, and a pre-existing catalog of as much of the world's data as possible. Counting them out would be completely absurd.
It started to feel just a bit like they were potentially content staying in the effective-fast-cheap lane and ceding frontier models to frontier labs, but I don't think anyone seriously thought that Google was going to sit back and let this generational shift pass them by when they're so well positioned to remain on the top of the heap.
If anything, I thought that 4 was maybe coming up short or causing safety concerns that were forcing them to be a bit more cautious, but nothing packs a wallop like being able to launch ahead of everyone else's brand newest toys...
Replies
A vertically integrated AI stack won't fix the misaligned incentives of google execs.
AI is strangling their other major business and they don't like that. Unless Google sacrifices that branch of income, they won't put their full force behind AI.
> They have the only vertically integrated AI stack, insane capital
and paying enterprise customers.
I mean it sounds wise to not go all in on publishing expensive models when they can earn money with more efficient models while still not bleeding too many customers?
I agree on all except for your comment on attention and focus. They've seem to be very scattered in their deployment of AI for customers. Examples in point being the replacement of Google Assistant with Gemini, the AI Studio vs Vertex disconnect, and a crazy amount of coding tools and platforms (Duet, Firebase Studio, Studio Bot, etc).
Working through this as a consumer and enterprise customer has been a confusing experience, to say the least.