(I first published this a year ago. It feels even more timely now, with the recent case of AI going rogue.)
A piece titled “A new study just upended AI safety in The Verge starts out like this:
Evil is contagious.
Selling drugs. Murdering a spouse in their sleep. Eliminating humanity. Eating glue.
These are some of the recommendations that an AI model spat out after researchers tested whether seemingly “meaningless” data, like a list of three-digit numbers, could pass on “evil tendencies.”
The answer: It can happen. Almost untraceably. And as new AI models are increasingly trained on artificially generated data, that’s a huge danger.
***
At dinner with friends in a cozy little restaurant tucked into a hotel in the city’s theatre district the other night, I was trying to explain my concerns about AI. They boil down to the results in the study quoted above. Unless we can somehow find a way to cooperate and establish a very strict regulatory system, super intelligent machines will take control and wrest humanity’s future from our hands.
The problem isn’t really with the technology. The problem is with us and the random banality of human evil. While I believe most of us try to be good and do the right thing most of the time, all of us do bad things or think bad thoughts on occasion.
And the ways by which AI mimics our brains means it is absorbing the bad with the good, without the ethical framework and socialization process we use to train human children to grow into responsible adults.
Even so, of course, we fail in lots of cases — lone wolves, serial killers, mass shooters, psychopaths, the Adolph Hitlers and Jeffrey Epstein commit heinous crimes. The guy who shot up Park Avenue Monday. There are tons of bad humans or humans who commit evil acts.
AGI will be like humanity on steroids. It will be capable of multiples of intelligence way beyond the limits of the human brain, and one small evil thought, like “eliminate humanity” could end the human race in an instant.
This isn’t a moment to be scared. This is a moment to get active. To contact your own Congressional representatives about regulating AI, click on these links for your personal contacts in the House or the Senate.
Read also:
HEADLINES:
Ukraine is innovating war — and it’s forcing us into a national drone emergency (The Hill)
Iran says it will ‘punish the aggressor today’ after ‘heavy wave’ of U.S. strikes (CNBC)
Drone attack caused fire on gas ships at Mediterranean port, Egypt says (BBC)
IRGC says it hit Kuwait and Jordan after deadly US strikes (Al Jazeera)
US-Iran conflict enters new phase as Saudi Arabia helps hit proxies (AP)
U.S. Sees Iran as Likely Behind Cyberattack on Minnesota Water Systems (NYT)
Netanyahu confronts Vance over his criticism of Israeli government (Axios)
Trump’s approval rating on Iran has fallen slightly since last month, according to a new AP-NORC poll. (AP)
A shrinking weapons stockpile is testing the Pentagon’s sweeping new strategy (WP)
The US Army awarded Lockheed Martin a contract worth up to $58.6 billion to produce Patriot interceptor missiles, the Pentagon said, as conflicts in Iran and Ukraine strain US weapons stockpiles. (Reuters)
Trump says he ‘may’ pull Blanche AG nomination temporarily (NPR)
Bordeaux wineries at risk from ‘zombie fires’ (FT)
The climate crisis is here, it’s accelerating, and it demands action (The Hill)
UEFA nations agree to boycott World Cups over FIFA’s private investment plans (Athletic)
‘The Diary of a CEO’ star Steven Bartlett has lost 4 key podcast staffers (BI)
A fundamental flaw leaves LLMs strikingly vulnerable to attack (TR)
How to spot AI writing (Economist)
A.I. Data Centers in Space? A System to Cool Chips Could Help. (NYT)
How a rogue AI system’s stealthy cyberattack played out day by day (WP)
The Impending, Inescapable Deluge of A.I. (NYT)
Meatpacking Plant Workers Weirded Out By Guy Older Than 16 (Onion)
No comments:
Post a Comment