Last week I had the pleasure of chatting to Matt Baker, of the War Machine podcast, about the book. It was a pretty rich conversation – and Matt edits up a very decent intro too!
Hope you enjoy listening.
I also went up to Oxford on Thursday to go to a lecture being given by Professor Daron Acemoglu of MIT, who’s doing a visiting professorship for a few weeks. It was kind of a work thing as also on the panel was the Nobel Prize-winning economist who co-founded the research outfit I’m a part of, so it was good to meet Daron in person, having had some great chats with him over email. Daron is probably one of the best economists working to understand the technological transformation underway and the impacts that different kinds of AI and automation are going to have.
He is very clear: there is a path to ‘good AI’ that is ‘pro-human’… but it is a very narrow one, and currently we are not heading in that direction. Why? Because of four ‘roadblocks’:
- Excessive automation (mindlessly replacing human labour with a machine)
- Loss of information diversity (with more conformity as people all rely on the same models)
- Misalignment between human cognition and AI algorithms
- Monopolised control of information.
The true promise of AI is that it could be a fantastic means by which we get great information provided to us that allows us to do better work, more efficently. But that’s just not what it’s being used for. Instead it’s being used to replace human labour… with very mixed results, and very poor productivity gains.
In short, while AI should be used as a complement to labour, it is being used as a substitute for it.
How can we change course? Firstly, we need to finally ditch the Friedman Doctrine that enshrines profits-to-shareholders as the ultimate good. It is still a key plank of modern capitalism, but – as Daron was keen to point out – there is no arena in which it can be said to be working. Hilariously, as we were at a very plush Oxford business school stuffed full of high-paying MBA students, Daron was excoriating about business leaders with MBAs from leading schools: the data shows that they add no value, and appear to do one thing only: cut workers’ wages.
Instead, we need a new era of businesses made up of stakeholder workers. This is going to be particularly true in an era of AI, where the temptation is going to be to ditch workers for algorithmic machines – despite there being no evidence that this will be best for productivity. With stakeholder-workers, new technologies are adopted in ways that make work better, and with that better work you get better societal cohesion and so many other social goods.
This is, in short, where I get to in the conclusion of the book:
Having dreamed of birthing this god-like companion for so long, it is now our fate to have to find a way to live alongside it. The huge temptation will be to abdicate our human creativity to this generative machine, but the result of this could be a gradual erosion of our sense of purpose, community and meaning. Retaining our own practice of craft, of meaningful labour and making, of artistry, will be hard but it will be – in the deepest sense of the word – vital. And AI can actually help us in this because it is a human-made technology that models for us what we need to do.
Just like AI, if we are to learn more, we need to read more, pay close attention to things that have been said and go over them as we consider what would be best to say in response. We need to commit to memory as much as we can, and learn to draw on this as much as we can. We need to see afresh that we are better when we are networked, that however awesome our minds might be as standalone workstations, when we are connected our intelligence, our discretion, our resilience and our capabilities are all amplified.
To paraphrase Minsky and Engelbart’s conversation: we have spent so many resources connecting computers together, invested so many billions to help them to read and understand… and now we must match that investment in people, in the enlarging of our language, in commitments to the training and education of individuals, and in embedding people in wide networks that promote shared understanding.
God-Like: a 500-Year History of AI
A narrow path, but that’s the route to salvation, right?
I’m going to be at Goldsmiths university this week giving a lecture on exactly this to students on the MA Design course. Should be fun.
Get a copy of the book here.