Speech: Jaan Tallinn explains the singularity

Next up: Why Now? A Quest in Metaphysics:

Jaan Tallinn is someone I respect because he created Skype, something once deemed the realm of science fiction, and also because he created the term “Steelmanning”.

I like this speech because it is a good introduction to large number of topics in philosophy.

The speech may be a bit difficult to digest, because of his strong accent, somewhat dry delivery, and the extremely deep topics being discussed. I urge you to bear through it.

The speech tells the story of Fred, someone who just happens to meet extremely smart people in random places.

Here are some notes (Leaving out a lot of detail, just a summation):

The News: The Upcoming Singularity

• In this talk, Jaan Tallin means to convey two epiphanies he had. First epiphany: we live in the single most interesting time in history. The entire universe might transform as a result of technological singularity, and we are beings capable of witnessing it.
• Computing technology will improve exponentially. It will only be stopped by the upper bounds set by the limits of physics.

Anthropics

• The anthropic principle states that you have a greater number of experience that are more abundant. An example given by Nick Bostrom is that you are more likely to end up in a slow lane than a fast lane, because slow lanes have more people.
• Given this, what are the chances that you just happen to find yourself in the most interesting time in history, the point depicted on the above graph?

Algorithm

• I don’t want to get into a philosophical argument, but from the perspective of physics, determinism is true, for all intents and purposes. It is possible to model what will happen in the future by perfectly understanding the present.
• What about quantum randomness? Well, this would make the world not fully determinist, but even this is fully possible to simulate in a model.
• Determinism holds true for brains. The brain can be viewed as an algorithm for translating sense data into the outputs to the muscles. Any algorithm, theoretically up to our brains, could be implemented any sophisticated enough hardware.

(Skipping to): The Multiverse

• The universe is much larger than previously thought. What we once believed was the entire universe is actually just a tiny cell of it. Moreover, the characteristics of this “cell” seem unusually finely tuned, just right for the emergence of life.
• Either there was some process to make the area we live finely-tuned, or the universe is simply so large that it contains every possible configuration. The latter is the hypothesis of the multiverse. If this is true, all experiences are produced in many places in the universe.
• But some experiences will be produced in more places than others. Due to the anthropic principle, we should expect by probability for our own experiences to be of the more frequent ones.

Synthesis

• Review the anthropic principle. We should expect to live the kind of life that is typical, a life that the multiverse produces over and over again. (Tallin has more detail, I myself am a bit unsure about everything he’s saying).

Simulations

• Are you living in a simulation? According to Nick Bostrom, there are 3 possibilities.

1. The human species will become extinct before the posthuman stage (unlikely)
2. A posthuman civilization is unlikely to run a significant number of simulations of its past (very unlikely), or
3. Because of recursion, simulated universes greatly outnumber un-simulated ones, so we are living almost certainly living in a computer simulation.

• If we are calculating how typical our experiences are, we should count both physical experiences and virtual ones. The greater the number of pre-singularity simulations there are, the less surprised we should be to be living in a pre-singularity world.

Superintelligences

• Consider a post-singularity world, which should be a very typical situation. In this place is a superintelligence. It might want to find another superintelligence to talk to. But, in the eyes of a superintelligence, the speed of light is so slow that traveling to another superintelligence takes too long. It is actually much easier to create another superintelligence.
• But how would a superintelligence create another one? One way would be to simulate the conditions that led its creation.

• You might want to simulate many superintelligences, and pick the most interesting/useful ones. But you don’t want to simulate the same thing twice. If two superintelligences are emergent from the exact same past, you only have to simulate that once.
• You will perform what programmers call a tree-search. You simulate the creation of one superintelligence, back up one step, simulate a slightly different one, and repeat the process.
• An interesting property of this process is that the number of simulated moments rises exponentially as you near the simulation.

• There is a huge “traffic jam” in the lane of the anthropic highway that contains pre-singularity moments. Pre-singularity moments (simulated), are extremely common, meaning that we should be not surprised to be living in one. This explains why our pocket of the universe is so finely-tuned for life; many simulations are given that interesting property.

My thoughts

Before watching Tallin’s speech, I was highly doubtful that we are living in a simulation. Now, I am almost convinced. It is just too much of a coincidence that we happen to be living in such an important time. The universe may be about to be transformed by exponentially increasing intelligence.

Does this contradict my comments about tech slowdown in the Late Culture Change Aversion post? Not really. Thiel has always been clear that we have seen plenty of improvement in computer technology, “in the world of bits”, not so much in physical technology, “the world of atoms”. Yes, the exponentially improving computer technology may eventually have a spillover to affecting physical technology, but we shouldn’t have to wait for strong AI for that to happen.

This speech by Jaan Tallin gets me thinking about how humanity has had a large number of “close calls”. On multiple occasions, a single person was responsible for preventing an apocalypse (Vasili Arkhipov and Stanislav Petrov). Maybe, a world was destroyed by nuclear weapons, only for the superintelligence to intervene in the simulation, back up time a bit, change a few variables, and let it play again. It is possible that nuclear holocaust happens all the time, but every time it happens, the AI hits the undo. This creates the illusion that the world is safer than it is, when really, apocalypse is always just around the corner.

Other speech highlights

Richard Dawkins explains the adaptive valley