Jeffrey Heninger, 14 February 2023
Epistemic status: Written for engagement. More sober analysis coming soon.
Bird navigation is surprisingly cruxy for the future of AI.
– Zach Stein-Perlman
Bird navigation is surprisingly cruxy for the future of AI.
This seems pretty wrong.
– Richard Korzekwa
This seems pretty wrong.
Birds are astonishingly good at navigating, even over thousands of miles. The longest migration routes, of the arctic term, are only limited by the size of the globe. Homing pigeons can return home after being released 1800 km (1100 mi) away. White-crowned sparrows have been able to migrate to their wintering grounds after being displaced 3700 km (2300 mi) shortly before they began migration.
How they do this is not entirely understood. There seem to be multiple cues they respond to, which combine to give them an accurate ‘map’ and ‘compass’. Which cues are most important might be different for different species. Some of these cues include watching the stars & sun, low frequency sounds, long-range smells, and detecting the earth’s magnetic field. This last one is the most interesting. Birds can detect magnetic fields, and there is increasing consensus that the detection mechanism involves quantum mechanics (See Appendix for details).
The result is a precise detector of the magnetic field. It is located in the retina and transferred up the optical nerve to the brain, so birds can ‘see’ magnetic fields. Leaving aside questions like “What is it like to be a [Bird]?”, this result has implications for the difficulty of Whole Bird Emulation (WBE).
WBE is important for understanding the future development of artificial intelligence. If we can put an upper bound on the difficulty of WBE, we have an upper bound on the difficulty of making AI that can do everything a bird can do. And birds can do lots of cool things: they know how to fly, they sing pretty songs, and they even drop nuts in front of cars !
In order to put bounds on WBE, we need to determine how much resolution is needed in order to emulate everything a bird can do. Is it good enough to model a bird at the cellular level? Or at the protein level? Or do you need an even finer resolution?
In order to model the navigational ability of a bird, you need a quantum mechanical description of the spin state of a pair of electrons. This is extremely high resolution.
A few caveats:
On the other hand:
WBE requires a quantum mechanical calculation in order to describe at least one macroscopic behavior of birds. This dramatically increases the resolution needed for at least parts of WBE and the overall expected difficulty of WBE. If your understanding of artificial intelligence would have predicted that Whole Bird Emulation would be much simpler than this, you should update accordingly.
Unless, of course, Birds Aren’t Real.
Here is a brief description of how a bird’s magnetic sense seems to work:
A bird’s retina contains some pigments called cryptochromes. When blue or green light (<570 nm) is absorbed by the pigment, an electron is transferred from one molecule to another. This electron had previously been paired with a different electron, so after the transfer, there is now an excited radical pair. Initially, the spins of the two electrons are anti-parallel (they initially are in the singlet state). An external magnetic magnetic field can cause one of the electrons to flip so they become parallel (they transition to a triplet state). Transitions can also occur due to interactions with the nuclear spins, so it is better to think of the external magnetic field as changing the rate at which transitions happen instead of introducing entirely new behavior. The excited singlet state decays back to the original state of the cryptochrome, while the excited triplet state decays into a different product. Neurons in the retina can detect the change in the relative concentration of these two products, providing a measurement of the magnetic field.
This model has made several successful predictions. (1) Cryptochromes were originally known from elsewhere in biology. This theory predicted that they, or another pigment which produces radical pairs, would be found in birds’ eyes. (2) Low amplitude oscillating magnetic fields with a frequency of between 1-100 MHz should also affect the transition between the singlet and triplet states. Exposing birds to these fields disrupts their ability to navigate.
I think it is a mistake to focus on these kinds weird effects as "biological systems using quantum mechanics", because it ignores the much more significant ways quantum mechanics is essential for all the ordinary things that are ubiquitous in biological systems. The stability of every single atom depends on quantum mechanics, and every chemical bond requires quantum mechanics to model. For the intended implication on the difficulty of Whole Bird Emulation, these ordinary usages of QM are much more significant. There are a huge number of different kinds of molecular interactions in a bird's body and each one requires solving a multi-particle Schroedinger equation. The computation work for this one effect is tiny in comparison.
As I understand, the unique thing about this effect is that it involves much longer coherence times than in molecular interactions. This is cool, but unless you can argue that birds have error-correcting quantum computers inside them, which is incredibly unlikely, I don't think it is that relevant to AI timelines.
I guess the idea is that if you approximate quantum physics with a hypothetical more simple model of atoms (might contain epicycles and magic numbers, it does not need to be theoretically elegant, just sufficiently simple to compute), some parts of the simulated organism might keep working, while other parts would fail immediately because they are more sensitive to details that in other situations can be abstracted away.
Still, in long term, the bird simulated by imprecise physics would probably die, because some other parts of their metabolism would also be sensitive to some details in a way that does not immediately have visible effects, but the smaller effects would accumulate over time. (I am just guessing here, but I imagine the symptoms would resemble radiation poisoning.)
A possible use in science fiction: the first emulated humans can live and think, but only for a few hours; trying to simulate them longer results in nausea, unlocalized pain, and finally death. Thus human emulation would not be used as a form of immortality, but rather as a way to save a "snapshot" of you, that you can later spin up an instance of, ask it to solve a problem, and delete it afterwards. A weaker version of Robin Hanson's ems, because no instance could live long, so they couldn't learn things or solve problems that would require more than their simulated lifetime. To get ems who are experts in new things, you would need to make new snapshots of people after they have learned it. As an economical consequence this would probably create a sharp distinction between people who spend their entire lives learning to become better and better experts to produce the best snapshots who will then do the actual work; and people who do not bother learning at all, because one could no longer get a white-collar job by being "merely good" or even "merely very good" at something, if the company can buy a snapshot of the best expert instead.
Uhm, back from science fiction... if this turns out to be true, I imagine there are two potential ways to solve it. First is to simulate the actual quantum physics, if that is feasible. Second is to keep studying human metabolism and keep adding epicycles to the simulation. For example, if the imperfect physics simulation ruins a synthesis of a certain molecule, the simulator could just magically keep injecting those molecules to the proper places in the simulated body, in proper quantities. You would probably want some kind of magic anyway, like to prevent cancer or aging, so you might as well also add magic to fix the imperfections of the simulated metabolism.
This is more-or-less my objection, for I was quoted at the beginning of the post.
What if we simply provide a magnetic field detector, aka compass, as an input device to our AI?
If that seems insufficient, how far before simulating full physics of a bird's body as-is would be sufficient? (It also seems that such simulation is completely outside of scope for AI, because it has nothing to do with intelligence per se).
Adding a compass is unlikely to also make the bird disoriented when exposed to a weak magnetic field which oscillates at the right frequency. Which means that the emulated bird will not behave like the real bird in this scenario.
You could add this phenomenon in by hand. Attach some detector to your compass and have it turn off the compass when these fields are measured.
More generally, adding in these features ad hoc will likely work for the things that you know about ahead of time, but is very unlikely to work like the bird outside of its training distribution. If you have a model of the bird that includes the relevant physics for this phenomenon, it is much more likely to work outside of its training distribution.
I think this is a surprisingly neat effect, but that the effect powering bird's magnetic sense can't actually work for quantum computing, and that as a general rule due to decoherence, we shouldn't expect evolved systems to be able to do quantum computing.
For this effect to work, it needs a coherence time of at least 100 microseconds, which is long relative to what you would expect in a warm & wet environment, but short compared to the time scales humans usually operate on.
I forgot to mention another issue, in that the form of quantumness isn't scalable. We've already tried using this technique, and it doesn't work for scalable, low error rate quantum computing.