Breaking: Gemini 2.0 Flash Goes Live - Inside Google DeepMind's Latest Release with Logan Kilpatrick

[@LoganKilpatrickYT]

Thanks for having me on, this was a lot of fun :)

[@antonijo01]

Thank you for the new Gemini models. They are fantastic. Congratulations to the entire team.

[@VaVer-r2o]

This is great! As a side comment, there is a clear quality difference between podcast host and guest. The latter uses a high quality microphone and it sounds much better. As your audience grows, please get a microphone, the headphones don’t cut it anymore. You can skip the lighting, that’s secondary. Particularly for ppl who want to hear you while being outside or exercising.

[@thenoblerot]

Great job, Gemini and infrastructure teams! (And awesome channel, Nathan!) I use my free/cheap Gemini API credits for a hobbyist robot 🤖 I'm currently expermenting with enabling the LLM to point to a location on its visual top-down map that it wants to go to. Mixed results, and i suspect it's because I'm essentially asking the LLM to point at *nothing* (empty whitespace, not too close to black pixels) rather than *something*. It's definitely good at pointing or generating bounding boxes for actual arbitrary objects, though. Might be a good computer use agent?!

[@flopasen]

this channel is good

[@oskarmariagrande1855]

nathan is always on it, indeed 🔥

[@arthurparkerhouse537]

Is the 0205 model less compute intensive than 1206 was, or does it just have more processing power behind it? The 0205 Pro is much faster than 1206 Pro was. Is it a smaller model than 1206 was?

[@bobbyjunelive]

2x4 to the forehead …striking ha ha ha that was excellent