“People should stop training radiologists now. It’s just completely obvious that within five years, deep learning is going to do a better than radiologists.” ~ Geoffrey Hinton aka “The Godfather of AI”
Does that scare you? Would it scare you less if you knew that quote is more than a decade old, and that out here in the real world there are more radiologists today than there were then? Hopefully so, but this knowledge has not stopped the AI elite from continuing to make such claims. Elon Musk recently predicted the end of surgeons because his AI enabled robots will be better at surgery than a human being within three years. Are you looking forward to being operated on by a robot?
Dario Amodei, CEO of Anthropic, published a very lengthy essay in January about the dangers of AI, which include the usual science fiction threats (think “Terminator”). Fortunately for us, Amodei does not think the machines will decide to eliminate mankind…at least not on purpose. That is a relief.
He does think AI will destroy all the jobs on the planet, making human beings worthless (at least economically). He claims that AI is different than technology of the past. When agriculture was automated, the workers could simply move into factories. When factories were automated, the workers could move toward technology and service jobs. He claims that there will be nowhere to go with AI, because AI and AI-empowered robots will do everything, and they will do everything better than humans.
This is horrifying. It is also wrong. I will freely admit that Amodei knows far more about AI than I do, but his economics and knowledge of human nature seem lacking. First, he suffers from what we in investing refer to as hindsight bias: It may seem obvious today that when farm workers were replaced by machines they would just move into the factories, because that is what happened, but I doubt it was obvious at the time. The automation of agriculture began with the invention of the cotton gin in 1794, while Henry Ford invented the assembly line and with it modern manufacturing in 1913.
It was not obvious then what the next level of employment would look like, and it never is. Humans, if given the freedom to do so, always invent new ways to add value to society anytime they become unburdened by the old. What will human employment look like when AI is fully up and running? I have no idea, and neither does anyone else, but if economic history tells us anything, it tells us that the future is bright. The only thing that can stop that future is the very heavy-handed policies that Amodei thinks are necessary.
We wrote about this in 2009 in the 4th quarter issue of our Quarterly Report newsletter, “Jobs.” There is a reason capitalism works best in countries with strong faith traditions. Capitalism itself requires faith.
Amodei and the other AI doomsayers also make another major mistake: The question for society is not what AI will be capable of, but what how AI will actually be used. Harvard professor and author Clayton Christensen’s 2011 book, “Disruptive Innovation,” showed how real disruption does not come from cutting-edge technology, because technology always exceeds its utility to the average user.
I was reminded of this just the other week. Michael Smith and I went to record our latest podcast and were suddenly blocked by our IT department from accessing the website of the application we use to do so. The application had been upgraded with “AI” to make it easier to edit podcasts. As with so many instances today one has to question whether this is truly AI or just a simple software update, but the word AI is now on the site and for that reason alone, our IT department makes sure we can’t access it. This is the real world. Both individual users and corporations are going to be very cautious when implementing AI. It will not go as quickly as the technology itself.
I was explaining this at a client meeting for which we met in person but some of the participants joined remotely. The meeting software had a glitch, so we were unable to hear any of the people who joined remotely. This is also the real world. Technology breaks. Everyone seems to know this except for the people who actually build the technology. Will AI be the one exception to this technology rule? Doubtful.
Finally, AI might very well be superior to human radiologists, and it may empower robots who are better surgeons. There is one thing that AI will never likely do better than humans, and that is care. Human surgeons sometimes struggle with emotion. Twice in my life I refused to let a surgeon operate on me because I did not trust them. If I get operated on in the future and have the choice between an AI robot and a human, I’m picking the human. AI may be more precise, but that human cares about me waking up afterwards. That matters; at least that is my perspective.
Warm regards,
Chuck Osborne, CFA
