How to test vision capabilities https://medium.com/@gabi.preda/testing-gemma-3-4b-with-ollama-a-quick-dive-into-image-interpretation-abc233b49aed
As someone who has been thinking about self-hosting a local instance of an AI to help me with writing projects this looked interesting until I realized that URL had ‘google.com’ in it. I am deeply suspicious of anything google and would be afraid it’s collecting data on the shit I’m doing and sending it back to them. Does anyone know enough about this to be able to tell me whether that’s the case or not?
If you run it locally, your conversations don’t go anywhere.
Afaik a model is just like a dB, you should question yourself about ollama or the software used to query the model, not the model itself
Fair point. I’m kinda new to all this.
It’s always a good advice to stay away from that disgusting biz (all big tech), but a model alone can’t collect information and send it back to EvilCorp HQ, so you are safe in that regard.
Personally, I have left Google 95%, but since they train on my data, I intend to use all Google products purely for social criticism, research and data on how psychopathic/criminal Google is, the criticism of Capitalism, the Western Oligarchy and their massive propaganda, how Ads are a form of mass-manipulation, and everything else that they wouldn’t want to have in their model dataset.
Maybe just have googles own AI products produce garbage wrong research, and store all the noisy data on Google Drive. Go train on that, you f**kers…
I admit it won’t influence future models much, so It’s more of a ‘feel good’ thing ;-P
Fair enough, thanks.
Yeah I avoid it to the extent that I can. The only thing I still use that is owned by google (since the whole Manifest v3 thing fucked adblock in chrome/chromium) is gmail, and only then because I’ve had the same email address since you needed an invite to get a gmail account (like 2006-ish?) so changing it now would really suck.
but since they train on my data, I intend to use all Google products purely for social criticism, research and data on how psychopathic/criminal Google is, the criticism of Capitalism, the Western Oligarchy and their massive propaganda, how Ads are a form of mass-manipulation, and everything else that they wouldn’t want to have in their model dataset.
Damn, that’s a good plan, I like it.
Interesting to see they’re using chatbot arena ELO rather than some unrepresentative benchmark designed to make them look good, that actually gives me higher hopes for this
