Why We Need Some AI Model Controls

SAFE - Safe Access For Everyone

I'm struggling to see how its still open source as we know it, if the models got governed controls implemented in them. Whos gonna be trusted to say whats safe and whats not, and be in control of the models.

The genie is out of the bottle, no going back now

Everything you say makes perfect sense , I am also most concerned with your same concerns instead of the corrupt regulation concern but it relies on the fact that people are good and not corrupt , will cooperate and the system is psychopath safe. Unfortunately any system with power will inevitably become corrupt and self serving. Need to find a way to zero trust the system of control ans regulation somehow, not sure it's possible

The problem with small degree, is that it never stops at small.

It is always interesting to hear from veterans in the field, rather than freshly minted grads with very limited real-world experience. I often find myself asking what old school experts like Kevin Kelly, Jaron Lanier and Bruce Sterling now think about the current capabilities. On the other hand, maybe we no longer have control over the acceleration, and as AWG claims, we are about to experience all the science fiction scenarios ever imagined?

Daniel, you're right about this, the risks of getting it wrong are equally terrible in both directions, Daniel Schmachtenberger has given great talks on this, years ago! Check him out if you haven't already.

The SAFE Project: https://oursafe.ai/

Spot on, Dan! Thanks for speaking up on this. The sci-fi analogy is excellent. We all need some distance, different angles and objectivity in order to start understanding where we're headed and why careful regulations are needed.

99% of people won't know it's "SAFE" ? I had Fable restrict "an unsafe conversation" when I asked it to answer a riddle. I think we're a loooong way from having a model that has useful safeguards.