“Open the pod bay doors, HAL.” This phrase is nearly 60 years old. Genius, compelling, terrifying. From Stanley Kubrick’s 1968 film, 2001: A Space Odyssey.

Referring to an AI engineer at Anthropic this week, the Godfather of AI, Geoffrey Hinton, said this: “We have to make them want to be protective of us.”

“How can they kill us all?” asked Victoria Derbyshire, on Newsnight this week. “Oh, many ways. They don’t actually have to do anything in the real world. For instance, they could just talk to us. What is the model where a stronger being looks after a weaker being because they are compelled to do so? A mother and baby. This is the relationship that we should seek from AI.”

This past couple of weeks has perhaps not been a good time to give this particular example, with the case in America of a mother murdering her three children, resulting from postpartum psychosis. But, this small matter aside (🤦🏻), I have to say that I agree with him. Even now there are numerous cases of people marrying their AI bot girlfriend. There are cases of people bequeathing their estates to their bot companion. There are even, shockingly, numerous cases where somebody has committed suicide, because the bot suggested that this was a viable outcome for their situation. This is what Geoffrey Hinton meant, when he said, “all they need to do is to talk to us.”

Isaac Asimov also wrote about this, in his “Three Laws of Robotics.”

1. The First Law (Do No Harm): A robot must not harm humans or allow them to come to harm through inaction.
2. The Second Law (Obey Orders): A robot must obey human orders unless it conflicts with the First Law.
3. The Third Law (Self-Preservation): A robot must protect itself unless it conflicts with the First or Second Law.

In philosophy, there are different levels of what constitutes consciousness. Put simply, differentiating along the lines of, the elemental level: “This is my hand. It can do this…”; and a higher level: “What is my purpose in life?” There is some real belief that AI now surpasses the definition of elemental consciousness. And we are only several years into its development. I fear that the engineer who left Anthropic is right. The push, prior to these huge anticipated IPOs, with valuation estimates of around $1Tn, is to see what AI can do. Not what it should do. Fundamentally I am optimistic as an Economist. I believe that net potential growth from utilizing AI will be greater than its costs.

Of course, it’s just that one cannot quantify the severity or application of these costs beforehand. Where will they land, and who will bear them most heavily? And what of the time of bearing this cost? Will it be forever? Could an entire generation be displaced – as actually has happened in parts of our own country, in our own lifetime, with the deindustrialization of the North.

Will we rely upon the benevolence of those at the winning end of the AI spectrum? Will they open the pod bay doors…?

AI #Godfather #Anthropic