How do you come up with a safe word that GPT4 can't say?
SAFETY
-
AI Development Accelerating Faster Than Expected
By
–
arf là c'est le moment où ça va partir en sucette… ça va trop vite
#ai #content #training -
GPT-4 Alignment Progress and Future Challenges
By
–
also, i am pretty proud of the degree of alignment for GPT-4 relative to previous models. we still have a long way to go, and we really need more powerful alignment techniques for more powerful models.
-
OpenAI should release alignment datasets and democratic governance
By
–
one thing coming up in the debate about the pause letter i really agree with: openai should make a great alignment dataset and alignment evals and release those! bonus points if we can find a prototype democratic process for 'what we align to'.
-
AI Intelligence Lies Within Human Direction Not Machine
By
–
yes c'est un assistant, dans intelligence artificiel, il y a surtout "artificiel", l'intelligence c'est l'humain qui lui demande de faire des trucs. Tu lui dit d'écrire des conneries, il écrit des conneries.
-
Safety and moderation baked in by default for AI systems
By
–
Also, safety/moderation – ppl made a good point I should bake that in by default before sharing
-
AGI Integration and Prevention of Existential Risks
By
–
But that’s just one possibility from many, and AGI devs are working on preventing it. Also the society will get integrated with AGIs, so in few years we won’t even care what is AGI and what is natural.
-
Questioning AI Existential Risk Certainty and Intelligence Explosion
By
–
I really tried to read the arguments, but I missed the point of why they have 100% certainty that AI will suddenly want to kill us all. There are so many other more likely alternatives. Intelligence explosion will surely change the world, but why should we fear it? What is
-
AI Doomer Logic: Examining Safety Concerns and Reasoning
By
–
This picture captures my confusion about the reasoning of AI doomers
-
LLM Risks Are Real, But Doomsday Scenarios Feel Overly Dramatic
By
–
I agree with Emily; this campaign, although probably well-intentioned, feels overly dramatic. LLMs and transformers are more effective than anyone expected, and there certainly are real and present risks, but why jump to the doomsday scenario of robots replacing us?