Gemini 2.0 is out, and there's a ton of interesting stuff about it. From my testing it looks like Gemini 2.0 Flash may be the best currently available multi-modal model - I upgraded my LLM plugin to support that here: github.com/simonw/llm-g... Gemini 2.0 announcement: blog.google/technology/g... - ThreadSky

simonwillison.net • 170 days ago

Gemini 2.0 is out, and there's a ton of interesting stuff about it. From my testing it looks like Gemini 2.0 Flash may be the best currently available multi-modal model - I upgraded my LLM plugin to support that here: https://github.com/simonw/llm-gemini/releases/tag/0.7

Gemini 2.0 announcement: https://blog.google/technology/google-deepmind/google-gemini-ai-update-december-2024/

Comments

neurostats.org•170 days ago

📌

simonwillison.net•170 days ago

llm -m gemini-2.0-flash-exp describe -a http''s://static.simonwillison.net/static/2024/pelicans.jpg

Output here: https://gist.github.com/simonw/32172b6f8bcf8e55e489f10979f8f085

simonwillison.net•170 days ago

The most impressive thing is the streaming mechanism, where you can stream audio and video into Gemini and get text or audio streamed back

https://aistudio.google.com/live only worked for me in Chrome: point your webcam at anything and have a live audio conversation about what the model can "see"

tommis.fi•169 days ago

Tested - works on iphone + firefox as well

anthonymoser.com•170 days ago

Hey, I read your latest blog post; it helps clarify your approach to AI tooling like this. Love datasette, the tools you've put out make sqlite more accessible.

I am curious: have you thought or written about how you parse the ethical issues related to this technology?

simonwillison.net•170 days ago

A whole bunch, yes - I have a tag for that https://simonwillison.net/tags/ethics/

simonwillison.net•170 days ago

Here's a relevant section from one of my talks https://simonwillison.net/2024/Jul/14/pycon/#pycon-2024.058.jpeg

anthonymoser.com•170 days ago

Thanks. To be clear, I don't just mean whether some ai tech is being built unethically; that seems pretty firmly established.

I mean if a particular tech is made unethically, how / where do you draw the line (for yourself) between helping people understand it and becoming complicit in its harm?

anthonymoser.com•170 days ago

Like this is interesting, but it's focused on ethical applications of the technology, which takes for granted that those are separate questions from the harm that sustains the platform and model

itsdavesanders.com•168 days ago

I was only able to test briefly but the speed was breathtaking. Almost supernatural sometimes.

Comments

Posting Rules

Reply