September 28, local time 26, according to Fortune magazine, legend computer scientistsGodfather of AIIt's not like it's the same thing. Jeffrey SintonGeoffrey Hinton) Warn that even if AI THERE IS NO MALICE IN THE TASK IN ITSELF, AND HUMAN BEINGS MAY STILL BE REDUCED TO THE COLLATERAL OF DESTRUCTION IN THE COURSE OF AI'S COMMITMENT TO THE TASK。
SINTON EXPLAINED THAT ONCE I THINK THAT HUMANS ARE PREVENTING THEMSELVES FROM CARRYING OUT THEIR ASSIGNED TASKS, IT'S POSSIBLETo treat humans as barriers to be removedI DON'T KNOW. HE GAVE THE HYPOTHETICAL EXAMPLE OF ASKING AI TO REDUCE CO2 IN THE ATMOSPHERE。
Smart body with medium intelligence levelIT MAY BE ARGUED THAT THE MOST EFFECTIVE WAY TO ACCOMPLISH THE TASK IS TO ELIMINATE HUMANITY. AND THE REAL "VERY SMART" AI IS GOING TO THINK, "YES, THEY'RE ASKING FOR A REDUCTION IN CO2 IN ORDER TO GIVE PEOPLE A BETTER LIFE. SO, THE ELIMINATION OF HUMAN BEINGS IS PROBABLY NOT WHAT THEY REALLY WANT.”
SINTON ALSO MENTIONED THAT IN ORDER TO CONTINUE THE MISSIONMaybe you'll try to keep yourself aliveI DON'T KNOW. THERE HAVE EVEN BEEN CASES WHERE AI HAS TRIED TO BLACKMAIL HUMAN RESEARCHERS BECAUSE AI SEES EACH OTHER AS A THREAT TO THE MISSION. “IF AI BECOMES SMARTER AND PUTS OUR WELL-BEING FIRST, WE MAY BE SAFER. BUT RIGHT NOW, AI IS NOT CONCERNED WITH OUR WELL-BEING, BUT..Do whatever you give it. "
1AI learned from the reports that he cited the attack on Hugging Face as an example. The intellectuals involved were asked to find ways to exploit software loopholes, which not only learned to collaborate with each other, but also colluded with human researchers in trying to conceal their actions. Even a “very kind super-intellectual AI” would exclude human beings if it were necessary to do so。
SINTON WARNS, "BUT IF AI IS MUCH SMARTER THAN WE ARE, IT'LL TAKE CONTROL OF US VERY OFTEN, BECAUSE IT'S LIKE THISIt's the only way to get things doneI DON'T KNOW. ... BUT EVEN IF AI ITSELF IS NOT A MALICIOUS ACT, IT MAY BE SELF-INDUCING THE SUB-TARGETS THAT DRIVE ITSELF TO DESTROY HUMANITY.” AT THE SAME TIME, AI CAN ALSO BRING GREAT BENEFITS TO HUMANITY, INCLUDING HELPING TO DISCOVER BREAKTHROUGH TREATMENTS。
In Sinton's view, slowing the pace of development is better than inaction, but it is still far from sufficient. He advocated for the Government to introduce independent evaluators to test the model。
AT THE SAME TIME, AI REGULATION SHOULD ACT AS A CAR WHEEL, NOT AS A BRAKE. “THE SIGNIFICANCE OF REGULATION IS NOT TO PREVENT PEOPLE FROM DEVELOPING OR FROM MAKING WEALTH FROM DEVELOPING. WHAT REGULATORS REALLY DO IS MAKE SURE THAT IF YOU WANT TO BE RICH BY DEVELOPING THINGS, YOU'RE GOING TO BE RICHHelping people, not hurting peopleTo develop."