Is AI “normal”? | MIT Technology Review

So on this background, recently essay Two researchers AI in Princeton felt pretty provocative. Arvind Narayanan, which directs the University Center for Policy of Information Technology, and a doctoral candidate Sayash Kapoor wrote a 40-page petition for everyone to calm down and think about AI as normal technology. This is contrary to “the usual preference for treatment related to a separate species, highly autonomous, potentially superintelligent entity.”

Instead, according to researchers, AI is a general purpose technology whose application could be better compared to the acquired adoption of electricity or the Internet than with nuclear weapons-even acknowledging that it is somehow flawed analogy.

Basic point, Kapoor says, is that we need to start distinguishing the rapid development of Ai methods– glittering and impressive representations of what Ai can do in the lab – and what comes from real login From AI, which in historical examples of other technologies are lagging behind for decades.

“A large part of the discussion of social influences AI neglects this adoption process,” Kapoor told me, “and expects social influences to happen at the speed of technological development.” In other words, the adoption of useful artificial intelligence, in his opinion, will be less tsunami and more click.

In the essay, a pair of other arguments: expressions such as “superintelime” are so incomplete and speculative so we should not use them; AI will not automate everything, but it will give birth to the category of human work that monitors, checks and monitors AI; And we should focus more on Ai’s likelihood that they will worsen current problems in society than to create new ones.

“Capitalism Ai overcomes,” Narayanan says. He has the ability to either help or hurt the inequality, labor markets, free press and democratic movement, depending on how it is arranged, he says.

However, there is one alarming exchange of AI that the authors omit: the use of AI by the military. That’s, of course, pick up Quickly, raising the alarms that AI decisions are increasingly helping AI. The authors exclude this use from their essay because it is difficult to analyze without access to classified information, but they say that their research on the subject is the upcoming.

One of the biggest implications of the treatment of AI -Ja as “normal” is that this would increase the position that Biden’s administration and now Trump’s White House has occupied: the construction of the best Ai is the priority of national security, and the federal government should take a series of actions – limiting what chips can do to China, dedicating more energy to the data – as it would happen. In their work, the two authors of the rhetoric of us “AI weapons” call “AI Arms Race”.