Doom in, doom out

Recently, a former Anthropic researcher went on CNN. The banner under him read: AI could "kill us all by the end of the decade." If you saw that and it made you worried, good. If that's what it takes for you to take this seriously and actually pay attention, I'll take it.

That's how I opened my P(doom) reel. The reel runs 2:49. This is the longer version, plus what to do with the worry once you have it.

The number everyone argues about

P(doom) is the probability that AI causes the death of humanity. Wikipedia puts it as the estimated chance that advanced AI causes human extinction or the permanent collapse of civilization. The top researchers in AI have argued about that number for years. Some say 5%, some say 25%, some say 50%.

My own line is a lot lower than any of those. Anything over 1% is grounds for caution. One in a hundred sounds small until you remember the thing on the line is everybody.

The warnings came from inside

That CNN interview wasn't the first time someone left a major AI lab to warn the world, and a lot of the people doing it helped build this stuff.

Daniel Kokotajlo left OpenAI in 2024 and published AI 2027, a scenario that maps out how the next few years will most likely play out. So far I think it's been very accurate, and that's the part that should get your attention. The same group then wrote AI 2040, which shows how the future could and should look if we go about this responsibly.

Mo Gawdat was Chief Business Officer at Google X and a big part of building the AI we know today. He left Google in 2018. Since then he's put out books and speeches on how to live in an AI-driven future, and his book Scary Smart argues that AI learns from how we treat it.

Whether you like him or not, Elon Musk was warning about the dangers of AI over ten years ago. He co-founded OpenAI in 2015 hoping to make the number one AI company open source, which just means publicly available, with nothing hidden from the people. Clearly, that didn't work.

Why stopping it is off the table

My main point in the reel: if society had taken AI seriously five to ten years ago, we could have slowed this down a lot and gone about it in a much safer way. We didn't. At this point, asking the world to stop AI is like asking the sun to no longer rise. It's gonna come up tomorrow whether you're ready or not.

So our best option is to learn AI, make other people aware as best we can, adapt, and prepare for where this is heading. You can do all four of those while you're still worried.

Half a picture

If everything you take in is doom, you will only be able to imagine doom.

Somebody who watches nothing but collapse content and dystopian movies has a real read on the risks. They also have half a picture of the future, and that half is what they end up building toward.

I take the risk seriously enough to say that 1% line on camera. I also expect the next ten years to be the best decade in human history for the people who are paying attention. Both can be true. What decides which one you live in is your mindset, and your mindset is built from what you feed it.

Peter Diamandis has a three-minute blog post on this called We Need a Vision, Not Black Mirror. His point is that dystopian media trains people to expect collapse. The future you can picture is the only one you can build, so if the bad one is all you ever picture, that's the one you're practicing for.

The forest fire

Do I think this will be easy? Absolutely not. I think society is about to go through a forest fire, and a lot of damage gets done in a forest fire. At the end of one, though, you get healthier soil and new beginnings. The people paying attention have a chance to be part of that new soil. The ones who aren't may go down with the fire.

What to feed your head instead

I built the Hope page on my site for exactly this: the stuff I actually watch and read that lets me picture a good future. If you only get to five things, make it these:

• Mo Gawdat's 2026 commencement speech, The Skill You Need in the Age of AI. 15 minutes. The old game was chess, planning five years out. The new one is squash: stay on your toes and get back to center.
• AI 2027 and AI 2040. Read the scary one first.
• 19 Keys, AI Secret Attack Plan + Cognitive Supremacy. The line that stuck with me: stay one of the humans who can still think.
• Peter Diamandis, The People, Environment and Media That Shape Your Mindset. Nine minutes on the three inputs you control.
• Ten books, starting with The Last Economy by Emad Mostaque. I wrote a separate post on what I'm doing about that one.

Take the risk seriously. Then give your head enough of the good future that you can actually picture it. Godspeed.

David West III