Jeff Dean why Gemini fell behind: “We wanted to make the model good at lots of things. So I think maybe our focus on making it amazing at coding was lagging a little bit.And we realized that in our work to catch up on that. I think we have good efforts under way. But I think by focusing on that [coding], you end up reasoning and doing other kinds of tasks that where it needs to sort of work its way through breaking a complicated problem down into multiple sub-pieces and so on.And so that's good. If you improve coding, you also tend to improve that capability in non-coding things.”Am I crazy or does this answer not make fucking sense at all, it literally sounds like he’s rambling incoherently? What if Jeff Dean is the reason why Gemini failed?
>Am I crazy or does this answer not make fucking sense at all, it literally sounds like he’s rambling incoherently?He said they tried to make it a generalist, which made it not as good at writing code. Which is ironic because other models which are good at writing code are effective generalists because of the skills needed to write code.
>>109598556>What if Jeff Dean is the reason why Gemini failed?No. The reason is that Google dumped "don't be evil" and embraced evil to the max. They are now reaping what they sowed.
>>109598556>What if Jeff Dean is the reason why Gemini failed?We all know why Google is failing anon. Jeff Dean leaving is another symptom of the problem.
>>109598675>are effective generalists because of the skills needed to write code.Gemini is still one of the smartest generalists for answering a question imo. But now people expect AI to go on a whole agentic workflow of searches and tool calls during a conversation. Google bet too hard on AI just answering questions.
>>109600611>importing cheap labor from abroad to drive up profits of insanely rich tech billionaires is "International Socialism"I thank God every day for not making me retarded or American
>>109600672They don't want a country doing well based on its race because that would be a heckin nationalism.
Gemini is completely useless for any research. For some reason, it doesn't know what's going on. It's like early ChatGPT, "my knowledge cutoff is December 2022"
He's not wrong in terms of ability to solve BIG problems, but it did however correctly figure out an obscure 2000s-era internet personality for me on vibes alone when both Claude and ChatGPT models (both ranked higher in Artificial Analysis) failed. Probably the benefit of being plugged in with 25+ years of search volume.
>>109598556only used google search but I like how it keeps short and concise as opposed to chatgpt and claude that keepwritinglong bullshitthatdoesn't matter. I do feel like they dumbed down their search ai quite significantly it used to get shit right on the first try nowadays I get a lot of bullshit. Same with chatgpt insisting to keep repeating the same bullshit as if they cache the result.
