I believe that AI will bring humanity into its best era yet. Here's why: AI has been debated to death - and the angle is almost always negative. And there are absolutely justified concerns, including: The really bad ############## *AI can create new dangers, such as unstoppable bioweapons.* Probably. But unstoppable bioweapons likely already exist. We have pathogens deliberately designed to spread easily through the air, resist antibiotics, have the perfect incubation time, combine not one but multiple disease vectors, and basically be able to wipe out humanity multiple times over without any chance of stopping them. We also have a nuclear arsenal sufficient to wipe out everything on the planet many times over. That's bad. But once you have an unstoppable doomsday machine, having a second unstoppable doomsday machine doesn't really make it worse. Of course, the argument with AI is that it would make this capability widely available - which is worse. But I'm wondering how much worse it is than just having it in the hands of power-hungry dictators. *AI can be used to hack into infrastructure and remotely open dams, make nuclear power plants explode, direct airplanes to crash into each other, or reroute supply lines into gridlocks.* Probably. But a lot of AI is a zero-sum game. Offensive AI cyber capabilities can be defended against by defensive AI cybersecurity. There will be an arms race for a while, but there is a finite number of attack vectors, and I believe there will be a point of stasis. *AI, even when not used maliciously, can become "misaligned", e.g. overfulfill a goal without taking context into account, thus "turning everything into paperclips"* Honestly, I can't even get ChatGPT to create bikini pictures. I might be naive, but this feels like its very closed to be solved. The neutral ############## A lot of AI concerns are effectively zero-sum. Yes, AI is great at generating enormous amounts of spam. But you know what's really good at filtering that out? You guessed it - AI. The same is true for fraud, deepfakes, or scams. The great ############## So finally, we get to the heart of my pitch here. There are some seriously great things on the horizon. Humanity has been banging its head against some really hairy problems. And there's a good chance that AI can solve them. Here are just a few examples: *Fusion Reactor Design* Turns out, keeping a hovering bit of plasma that's hotter than the Sun from touching the walls of whatever you put it in is quite hard. But AI might be perfectly positioned to get this right. Once fusion energy works reliably, there's very little point in building another coal or gas plant - and a whole new industry around safe, sustainable, abundant energy will spring up. This will not only make Mother Nature happy but also massively reduce reliance on fossil fuel producers, thus reducing conflict. *Energy-Dense Batteries* As the Big Bang has so loudly illustrated, there's really not much of a limit to how much energy you can squeeze into how little space. Designing energy stores the size of an AAA battery, but with the capacity to power a cruise ship, is hard, but possible. If AI can get us there, it unlocks a whole new range of environmentally friendly use cases, from small personal aircraft to decentralized power grids. *Carbon Capture* There's a lot of CO2 in the atmosphere, and the planet is heating up. But - as utopian as this sounds - there is a plausible scenario in which we could just take the CO2 back out and stop this tomorrow. There are some promising approaches to achieving this - from direct air capture to modified plants and bacteria - but none of these work at scale yet. AI might just be able to figure this one out. *Sustainable Mass Food Production* In order to feed 8.3 billion people, we've had to take some serious shortcuts - many of which are unsustainable, leave the soil exhausted, or require us to do terrible things to large numbers of animals. But with pest-resistant plants that automatically manage the soil they're planted in, synthetic proteins, and climate-resistant crops with low fertilizer requirements, we might be able to change that. *Personalized Medicine* Decoding your cancer cells' DNA, deriving a targeted drug, and producing it is prohibitively expensive - and available only to a select few. But there's no inherent reason for that. We could scale this approach and make personalized cures available to billions. How to get to this future ############## And there’s a lot more. Only problem is, even if AI ends up with the ability to do all of this, we need to be careful not to fuck it up after all. Now, I'm a free-everything libertarian. But even I have to admit that, in this scenario, a lot of the theory around free markets might no longer apply. So, to make sure that the gains of this future don't just accrue to an oligopoly of infinitely powerful megacorporations, there are a few levers we'd need to flip early. Here are some suggestions: *Keep AI a level playing field.* Recursive self-improvement - the thing where AI creates better AI - means there's a decent risk of a winner-takes-all scenario. However, the current neck-and-neck race between frontier model companies, open-weight companies, and startups accelerating from the sidelines seems to provide a good hedge against it. Don't fuck it up by giving in to regulatory capture that would allow the top players to pull the ladder up behind them. *Don't let AI have patents.* AI creations should belong to everybody. Once, for example, fusion is solved, this should be universally accessible. Right now, it's a bit of a grey area, with humans being able to patent Ai created inventions - but by clearly legislating that no one, including the company making a model or its users, can patent its discoveries, we'll lay the groundwork for an evenly distributed future. *Maybe it's time for Star Trek-style space communism.* If we get to my future, the resulting abundance should be shared. And while my little Ayn Rand altar in the corner can't believe I said this, I'm afraid it's true. The market playbook of parallel competition for limited resources might just need to be revisited in - what is hopefully - a new, awesome AI future.