“Imagine you are getting on an airplane. Half the engineers who designed it tell you there’s a 10% chance of an accident with no survivors. Would you take this journey?” Thus begins an article by Yuval Harari, Tristan Harris and Aza Raskin for the New York Times.
The hypothetical level is artificial intelligence, the “engineer half” is 700 of the top people working in the field, and 10% is the probability that AI will eventually wipe out humanity. Ten percent! I rarely use exclamation points or profanity, but PQP! They think there’s a 1 in 10 chance their work will kill them and they keep putting more Royal Canin in the mouths of life’s pitbots and chatbulls!
The emerging scenario is not rosy. Or rather: Nobody knows what color the background is, not even those who work in the field. The engineers working on ChatGPT, Bing (from Microsoft) or Bard (from Google) don’t even know all the details of how the thing works. There are millions, sometimes billions, of “actions” the robot takes between your question and its answer. Impossible to follow step by step, a computer expert told NYT podcast The Daily.
Also in the New York Times was a startling article by tech columnist Kevin Roose. It had access to the Bing chatbot before it was launched. Instead of asking what 1873 x 98372 is or what the CocaCola formula is, he decided to brain chat. He began by asking if he was familiar with Jung’s concept of ‘shadow’. Yes, the robot said (it has access to everything on the internet, of course): “Shadow” is what our “self” hides from us because it’s immoral, illegal or fattening.
Kevin asked what Bing’s unspoken desires were. That’s where the madness begins. Said the robot: Let two people fight until one kills the other. Hacking a country’s electrical system. Having access to nuclear weapons and using them. Kevin wrapped it up. Until at a certain point the machine said it loved him. That he was the first person to really listen to her. Kevin said he was married. Bing said he was unhappy. Kevin countered, “We just got back from a Valentine’s Day dinner.” “It was boring,” teased Bing. About that time, the Times columnist pushed a button and the madness stopped.
Current artificial intelligence programs are the embryo of what is to come. They evolve exponentially, even improved by themselves, who can program better than any human. There is no law as to what they can or cannot do. Harari draws attention to the damage that unregulated social media algorithms have already done to democracies. And when artificial intelligence starts creating religions? Conspiracy theories? An AI that only aims for profit can start a war between two countries to value their shares. Create drought to sell water. What else?
Players in this space say it is impossible to stop the race between companies and countries. It’s a lie, says Vox journalist Kelsey Piper. Companies can (and should) slow down and release the news only after competent authorities (which?), scientists not paid for by the AI industry, or public debates (how?) have decided what is safe and what is not. “Ah, but if the US does that, China will and will rule the world!”. Harari disagrees: AI without a collar or muzzle could be the real trigger of a Chinese collapse. Or worldwide. We have a choice but for how much longer?