I did manage to catch up with Jason Kende, and had a fascinating conversation with a brilliant friend of his. She told me how she had spent some time a few years ago to project how the world might change over the next 50 years in order to map out her own decisions. She really focused on the inevitability of the singularity, in this case defined as the emergence of exponentially superintelligent AI, the eventual creation of two classes of people based on their proximity to neural uplink technology, and the need to stay ahead of the curve.
As some of you may remember, I went through a phase right before graduating college where I was unhealthily obsessed with AI safety. I was consumed with fear from ideas like the paperclip problem, where a superintelligent AI misinterprets a command to make more paperclips and turns the entire universe into paperclips. It was all I could talk about for at least a few years and was probably super annoying. To be fair, it did lead to a career in AI before my bright future devolved into crypto degeneracy, so it wasn’t all for nothing.
As Jason’s friend continued to describe the future she saw, I was struck by the similarity with my thinking then, and I found that even though I agreed with the general direction of what she was saying, I didn’t have the same reverence for AI any more. I think a big reason is that AI doesn’t actually exist. What I mean by this is that our definition of artificial intelligence has continuously shifted throughout the years. Twenty years ago Deep Blue beat Kasparov in chess and clearly heralded the pending triumph of machines over mankind, although in retrospect all they did was throw a bunch of computing power to work through the potential moves, and really nothing special. It had long fallen out of the limelight when Watson won Jeopardy. But these days it’s clear Watson was nothing but some hilariously overfitted NLP models hooked up to a search engine, and therefore barely worth mentioning. These days I suppose self-driving cars or Dota bots are the cutting edge of AI. One of these days someone will look back at these projects and realize these computer scientists were just hand-crafting deep learning subsystems in a reinforcement learning framework, and so were able to conjure some entertaining sideshows but fell well short of real AI and are merely a technical novelty.
My point is that what people consider real AI will always be the next shiny technobauble, because AI is just a reflection of the latest attempt to overcome our inadequacies. If there’s one thing AI has always done well, it’s highlighting how feeble we are, and how utterly alone. I think everything that needs to be said about AI was already captured in the Jewish stories about the golem. They generally start with some person making a creature out of clay and bringing it to life with the word of God, or a shem. The golem would start off doing household chores and would inevitably run amok, at which point the creator would panic and remove the shem from the golem and cause it to fall to pieces. This 16th century story has all the elements for any piece of fiction about AI ever made.
Now, I do agree that we’re heading in the direction of a violent stratification of society. In fact we’re probably already there. But notwithstanding the outstanding work being done by groups like DeepMind or OpenAI, these days I’m no longer particularly worried that artificial intelligence will be the death of us all, because omnicidal, out-of-control paperclip machines already exist – they’re called corporations. In today’s society, if a corporation can spend 1 dollar to gain 1.01 dollars in revenue, it’s not just considered bad form to not do so, it’s a breach of fiduciary duty and requires the offender to be removed from corporate governance. On top of that climate change will make large swathes of the Earth uninhabitable and further concentrate resources in the hands of those who already have them. All of this can happen even without some sort of singularity coming about. And if the Singularity with a capital S shows up and utterly transforms humanity in a Silicon Valley version of the Rapture, there’s not much I can do at this point about it one way or another.
This last year and a half feels like one long exercise in guessing the future, straining my eyes at some horrific spectre like the monsters in The Mist. I don’t know about you, for me it’s been a struggle between trying to think a few steps ahead towards increasingly bleak outcomes, and compartmentalizing what could happen to stay present. Every few months I read something about climate change, and I go through the five stages of grief all over again. In particular, I go through a mix of depression about the unavoidable collapse that’s coming for us, and desperate bargaining as I scroll compulsively through rainfall maps and housing prices, trying to pinpoint some sanctuary far from the Equator that might be able to weather the storm. Then I discovered Guy McPherson. If I had any doubt about the hopelessness of our position, his lengthy essay on why we’re fucked squashes that once and for all. Here’s Toby from The Office to sum things up. And all of this without a mention of AI.
Where does that leave us? How should we live in the future? The best approach I’ve come up with is to not sweat the details. What we know will likely happen is, society will continue to disintegrate and parts will fall to low-complexity states. The obvious things to do are to move away from the coasts to somewhere at elevation, with fresh water and surrounded by people you like and trust. The specifics are probably too different for each person to go into. And hoard ammunition.
PS: Some of these topics are covered in more detail in a paper with the good folks at The Digital Economist. You can check it out here: http://thefirenexttime.com/2020/12/05/what-the-future-looks-like/