I’m struggling with the future I see in my mind’s eye, but I see some very serious issues arising, and soon, with these new AI developments
I see the recent development of advanced AI chatbots (ChatGPT, Sydney, Bard, etc.) as extremely dangerous for a few reasons:
1) being the exemplar of exponential technology we can expect extremely rapid improvement moving forward. They’re already incredible at human speech and what they lack in terms of reasoning and truly creative thought will probably appear very quickly (in the next few months or years). The lily pad is half covered and one more doubling covers it fully (see Kurzweil’s book, The Singularity Is Near,* for a great history of exponential technologies).
2) this tech can be deployed at massive scale now, similar to how social media bots have already been deployed, but very crudely, by the Chinese, Russians and Americans already in efforts to sway elections and public opinion around the world.
3) just as in the movie Her these bots are already at the level of ability needed, or will be very soon, to become the phone assistant depicted in the movie, and we may well find ourselves in a world in which most children and most adults too simply prefer the companionship of their AI friends/lovers, continuing the atomization and fragmentation of what remains of our polis.
4) this kind of fragmentation is perhaps the key ingredient in creating societies that aren’t actually societies and that are, instead, strongly susceptible to social control and mass hysteria.
5) these AIs will soon be paired with robot bodies and be deployed in increasing numbers. Teslas Optimus may be the first to market but many many more will soon follow, giving rise to the real need for binding “laws of robotics” but well before we get there we should develop “laws of chatbots.”
6) last, but definitely not least, given the extreme difficulty of the problem of emulating human language at human levels of expertise, which is now conquered in the same way that Deep Blue conquered chess and others conquered Go, etc., we can expect almost all, if not all, other areas of human ability and ingenuity to quickly be conquered.
As such, I see the singularity very fast approaching, which is the notion developed by Kurzweil, and others, of computers getting so smart they can improve themselves, and at which point all bets are off as computers reach God-like abilities effectively overnight.
Our own history with far stronger human groups meeting the natural world and other weaker human groups inspires just about zero confidence in this superintelligence having any compassion at all for its biological progenitors.
See Nick Bostrom’s book, Superintelligence, for a masterful overview of the issues and some potential solutions.
I’m starting to press #AImoratoriumonnow as my small effort to get people to wake up to these dangers.
Thoughts? Please comment below.
*Kurzweil is an unabashed optimist about technology and the coming singularity, and I don’t agree with him much on this optimism (obviously, given what I write here), but I do agree that he presents a wonderful history of technology and, in particular, exponential technologies.
