Once you play with time travel having retrocausal power a few years in the past, there's no reason to limit yourself to a few years. You might as well say that Trump will cause a nuclear war, thus giving humanity enough time to solve alignment properly by the time it gets around to building transformative AI. And why should aligned AIs have all the fun with time travel? Maybe Trump's election leads through no fault of his own to a butterfly effect that causes a specific unaligned AI to gain these magic timestream powers, even as it disassembles the Earth for raw materials in 2031.
Hm this is a bit of an edge case for our frontpaging guidelines. It is talking about timeful content in a timeless way ...
1So, if Trump was miraculously saved, is that a reason to vote for him? After all, the extent to which your vote matters is the extent that you risk getting Novikov’ed out of existence.
There was an article in New Scientist recently about "sending particles back in time". I was a physics major, but I might have skipped the time travel class, so I don't have an opinion on this. But Sabine Hossenfelder posted a video, arguing that New Scientist misrepresented the actual research.
Epistemic Status: extreme, unhinged speculation. Possibly a small Basilisk-like effect from reading this that directionally seems neutral-to-positively aligned to the cause of human survival. Hopefully entertaining. Big if true.
Over the weekend Trump was very nearly assassinated. By sheer luck he decided to tilt his head before the 20-year-old would-be assassin pulled the trigger, causing the bullet to graze his ear instead of piercing his brain. This was a probabilistically ridiculous event that had huge implications for America's future. What updates should a Bayesian make to his beliefs in response to this event?
The only mundane hypothesis that explains away the extreme luck is that the assassination was a false flag on behalf of the Trump team. Indeed the result was optically optimal to Trump to a ridiculous degree. He even outdid Senator Morra from Limitless. But then you have to contend with the idea of a comfortably ahead Trump deciding to arrange for a 20-year-old to snipe at his ear from 120 meters away as he was moving around the stage. It seems preposterous to risk death like this for the sake of a boost in an election you were already more likely to win.
Other conspiracy theories, by other culprits, would still have this be a failed assassination, so it would still remain that Trump was ridiculously lucky.
Some religious pro-Trump people have talked of divine intervention. There is a species of atheist that would be triggered into debate mode by this and completely deny it. The principled rationalist should bite the bullet and say: yes, it's a bit of credible evidence. This event is something that you expect to be more likely in a world where a higher power exists than one where it does not, but there is still overall a million other things on the other side of the evidentiary scale of the theological question. But yes, if events like this start happening with frequency there would be cause to seriously consider the supernatural. Of course, one would have to wonder about the nature of this pro-MAGA interfering being, who, saving Trump, let an innocent firefighter attending the rally be murdered in his stead.
An interesting secular explanation for historical human luck is the anthropic principle. The nature of quantum mechanics substantiates the idea of a multiverse, and so an explanation to the puzzling luck that we have that the universe's physical constants are what they are, is that universes with different values exist, but there's just nobody there to observe them. Similarly, it explains the repeated luck that we've had with nuclear close calls during the Cold War by positing that being born in a world where a nuclear exchange occurred is significantly less likely, as there are much fewer humans being born in the worlds where it did occur than in the worlds where it didn't. This type of explanation does not work for the luck of the failed assassination attempt over the weekend, but the mind-bending probabilistic dynamics of the anthropic principle are a good appetizer for the explanation that I want to posit here.
Time Travel and the Novikov Conjecture
There are two flavors of time travel in fiction. In the first type the time traveler creates an alternate timeline, and in the second the time traveling is a closed loop, and everything the traveler does has already happened in the first place[1]. The second type creates time travel paradoxes; the simplest one asks what would happen if you went back in time and killed your own grandfather. Physicists' favorite solution to time-travel paradoxes is the Novikov self-consistency principle: events that would cause a paradox get a probability of zero. If you went out to time travel and kill your grandfather, something will happen to stop you. Maybe you get hit by a car on the way to the lab. Maybe there will be a sudden power outage. Maybe you'll just have a change of heart. A somewhat unlikely event will still happen over a very unlikely event, it's just that the probability measure of the worlds where nothing stops you from carrying out the deed, vanishes to zero.
Let's call the hypothetical state of reality where time travel is possible at all, and the Novikov self-consistency principle is indeed how it works, Novikov world. While in our fictional stories extremely advanced time travel machines are used by plucky protagonists to save a loved one or get laid, in a real Novikov world a rudimentary time travel machine would be, immediately from the first prototype, an irresistible national superweapon.
Novikov Exploitation
Suppose that a team of researchers find a way to send a single particle one microsecond back in time in Novikov world. Then you can set up a bare bones grandfather paradox on the atomic level: Add a particle detector and a control unit. At a certain time t, program the device to check for the presence of the particle. If the particle is there do nothing. If the particle is not there, then at time t+ε send a particle back ε time units into the past. You've set up a bare bones paradox.
Now take this contraption and hook it to a computer. Send out a death squad to assassinate the leader of a rival country, and program the computer to only activate the contraption if the squad is unsuccessful. As you are living in Novikov world, the probability of the military action being successful has significantly increased, though you will see some likelihood of Novikov artifacts, like the death squad being unsuccessful and the lab suffering a random blackout so the machine is never activated anyway. Still, you would have ways of mitigating that, and such a device will let you quickly dominate the world.
Aschenbrenner Dynamics in Novikov World
Leopold Aschenbrenner's vision of the near future is AI becoming a race between the US and China that destabilizes the national security order. He argues that the US government should get involved and create a huge national ASI project, and that having the US have a big head start over China is extremely important to have a peaceful outcome. If this occurs in Novikov world, I posit that the machine described before would be a medium-hanging fruit in terms of scientific and engineering difficulty, and extremely valuable as an immediate war winner, especially in a cold war.
In our usual conception of the singularity, the world gets crazy real fast once we have AGI because of unimaginable technologies. But in a singularity in Novikov world, weird things start happening before AGI due to Novikov artifacts. The universe would like for us not to end up using the exploiting machine too much in the future. With Aschenbrenner dynamics, that would mean an AI race that is won early. A contested AI race, especially one that remains cold in terms of military invasions or attacks, will cause one of the two rivals to resort to Novikov exploitation after it's developed in the medium term. It's a weapon that gives you ultimate power. And if this is indeed the world we live in, we should expect to see unlikely things happening that manipulate events so that Novikov exploitation does not get heavily used in the future.
Putting it all together
I speculate the following package of ideas to explain what happened last weekend:
And I'll assert that everything here fits together so well with the strange events that we observe that it's enough to locate this (admittedly out-there) hypothesis.
On a Personal Note
This crackpot idea came to me today and I was mulling it over on my commute home from work. I got distracted on the way by a comedically unlucky but harmless event. Right after, as I was reaching home I was considering on what level I should share this conspiracy theory (does it still count as a conspiracy theory when the conspirator is the universe?) It didn't seem like lesswrong material, but maybe it would actually move the needle. Indeed, if the theory is true, posting it is probably important enough that I could be manipulated by a Novikov artifact to post it. So I declared in my inner monologue that if it's important that I post this on a public forum, let something unlikely happen to me to give me a sign! Then I realized to my horror that something did happen just 10 minutes before, that certainly qualifies (I'd quantify it as a level of comedic bad luck that will typically occur to a person less than 5 times a year). And so I basilisked myself into posting this half baked theory.
Thanks for reading!
Annoyingly, both types occur in the Harry Potter franchise with no explanation for the discrepancy.