sigourney's knickers
Full Member
My man!
Not a lot m8.
Be very very careful
You forgot a comma bud.
I do a lot of this stuff for work.
Yes. It can and should be done.
Large Multi-Modal Models are incredible, revolutionary even.
Massive benefits, massive dangers.
As an example, ChatGPT 4 Vision is really very good at radiology:
Loading…
arxiv.org
View attachment 30854
But it can be inaccurate:
View attachment 30855
When should it be used? When should it not be used? How should organisations decide when it is appropriate to use it and when should they not?
This is all reasonable to legislate for, e.g. banning it from any decision making power in risk-to-life scenarios, limit usage in surveillance, creating regulations around explainability.
The EU is ahead of the pack here:
Artificial Intelligence Act: MEPs adopt landmark law | News | European Parliament
On Wednesday, Parliament approved the Artificial Intelligence Act that ensures safety and compliance with fundamental rights, while boosting innovation.www.europarl.europa.eu
AI is a nebulous term. It's really all just machine learning, albeit there's some interesting emergent behaviour when it comes to transformer based AI like ChatGPT.
Of course. But I've never stalked someone. AI makes that much easier, especially for scams like pig butchering. Scaling that up 5 years ago required humans, now it doesn't.
