You can’t escape debates about AI at the moment. BSV entrepreneurs are as fascinated as anyone else - perhaps more so if they have experienced previous tech revolutions.
Jerry Chan has gone public with his cautious scepticism about the much-publicised dangers of AI, but in this week’s CoinGeek Conversations, the former blockchain infrastructure developer and Wall Street technologist says he also sees the potential of AI, and that it could have a role in his forthcoming Frobots game.
“I am impressed,” he says of his experience with ChatGPT, “it is effectively like Google search on steroids …But it doesn't know if what it's saying is true or false.”
It’s been suggested that what’s unique about generative AI (that is, AI which creates its own content in its responses) is the existence of emergent properties - higher order functions which weren’t written into the software but somehow appear, if not of their own accord then at least to the surprise of its programmers.
Jerry believes that these kinds of property don’t mean we need to attribute sentience to AI or to worry that it’s acquiring human abilities. You can think of AI as the “equivalent of saying somebody's brain has massive breadth but very shallow depth - they can know much more about many more topics than any one human brain and sort of overlay layers of meaning and see which layer of meaning makes the most sense. But at the end of the day, the thing that makes humans humans is we have our own internal volition - we don't have a program we're going by.”
To try to define what is unique about human beings, Chan points to Richard Dawkins’ book, The Extended Phenotype, which talks about how our genes are responsible not only for the physical manifestation of our bodies - the phenotype - but also for the ways in which we affect our environment too. That starts with the genes producing cells which act together, each cell ‘knowing’ whether it should be, say, a skin cell or muscle cell.
When the cells lose this ability to act together and can only behave independently, they become cancer cells, no longer a useful part of the being in which they live. So, says Chan, “we as humans have innately - down to our cellular structure - a notion of self, even cellular self, which aggregates …and AIs are nothing like that”.
Despite his mild scepticism about the dangers of AI, Chan isn’t saying that its developers should be left to do entirely as they please, because there are also potential dangers:
“Regulation would be good because if something comes out of regulation that says ‘let's put some checks and balances for whatever AI you create …Let's not plug it up into the nuclear launch code system, let's agree not to do that’. That would be a good regulation.”
More immediately, Chan is interested in AI as a part of the STEM education game, Frobots, which he is developing. The idea is to teach players computer programming, but AI gives another dimension to the idea: “humans could write programs to fight and AIs could write programs, and I'm not exactly clear whether it's certain that AI would win, because compared to chess or Go, which we know AI wins …the game is not well-defined. It's an open ended game”.