6 ms·

It's not just OpenAI. Every AI organization is discovering that they have internal groups which are pulling in a different direction than the rest of the organi

by lacker 3y ago

It's not just OpenAI. Every AI organization is discovering that they have internal groups which are pulling in a different direction than the rest of the organization, and trying to get rid of those groups. * Google got rid of its "Ethical AI" group * Facebook just got rid of its "Responsible AI" team * OpenAI wanted to get rid of the "Effective Altruists" on the board I guess if I was afraid of AI taking over the world then I would be rooting for OpenAI to be destroyed here. Personally I hope that they bring Sam back and I hope that GPT-5 is even more useful than GPT-4.

didibus 3y ago
I feel the people advocating safety, while they are probably right from a moral and ethics point of view, are just doomed to fail. It's like with the Nuclear Bomb, it's not like had Einstein withheld his contributions, we wouldn't have Nuclear Bombs today. It's always a matter of time before someone else figured it out, and until someone else with bad intentions does. How to attack safety in AI I think has to assume there are already bad actors with super powerful AI around, and what can we do in defense of that.
- sbrother 3y ago
  Yeah, it’s an awful role basically by design. It reminds me of my uncle’s role as the safety/compliance lead at a manufacturing facility. He was just trying to do his job but constantly getting bullied, maneuvered around etc because his job was effectively to slow everyone else down. It drove him crazy, he ended up having a mental breakdown, becoming an alcoholic, and having to go on disability for the remainder of his career. I’m sure some people can handle the stress of those roles, but ugh, not worth it IMO.
- lacker 3y ago
  Yes, I agree. There will always be some evil person out there using AI to try to achieve evil goals. Scammers, terrorists, hackers. All the people who use computers now to do bad things, they're going to try to do even worse bad things with AI. We need to stop those people by improving our security. Like fixing the phone number system already. There must be some way to stop robo scam calls. Validate everyone making a call and stop phone spam. That would do a lot more to help AI safety than randomly slowing down the top companies.

by lacker 3y ago

didibus 3y ago

I feel the people advocating safety, while they are probably right from a moral and ethics point of view, are just doomed to fail. It's like with the Nuclear Bomb, it's not like had Einstein withheld his contributions, we wouldn't have Nuclear Bombs today. It's always a matter of time before someone else figured it out, and until someone else with bad intentions does. How to attack safety in AI I think has to assume there are already bad actors with super powerful AI around, and what can we do in defense of that.

sbrother 3y ago
Yeah, it’s an awful role basically by design. It reminds me of my uncle’s role as the safety/compliance lead at a manufacturing facility. He was just trying to do his job but constantly getting bullied, maneuvered around etc because his job was effectively to slow everyone else down. It drove him crazy, he ended up having a mental breakdown, becoming an alcoholic, and having to go on disability for the remainder of his career. I’m sure some people can handle the stress of those roles, but ugh, not worth it IMO.
lacker 3y ago
Yes, I agree. There will always be some evil person out there using AI to try to achieve evil goals. Scammers, terrorists, hackers. All the people who use computers now to do bad things, they're going to try to do even worse bad things with AI. We need to stop those people by improving our security. Like fixing the phone number system already. There must be some way to stop robo scam calls. Validate everyone making a call and stop phone spam. That would do a lot more to help AI safety than randomly slowing down the top companies.