Will whole brain emulation arrive before other forms of AGI?
Whole brain emulation (WBE) presents very different alignment challenges than other kinds of AGI do, so it’s worthwhile to try predicting which type of AGI we’ll have to deal with. Opinions differ, but researchers tend to believe the first AGI will not be produced via WBE.
There’s significant uncertainty about how long it will take to get the technology needed for whole brain emulation. We don’t yet have the scanning technology to produce sufficiently high-resolution scans of the human brain, the neuroscientific understanding needed to interpret them, and perhaps the computing power needed to run the resulting emulation.1 Anders Sandberg predicts a median of 2064 for the year in which the technology for WBE will first be available. Robin Hanson guesses that it will be available within a century.2
We also don’t know when AGI will arrive. Hanson, Gary Marcus, and a number of researchers on the AI Impacts survey expect long AGI timelines, maybe over a century, whereas many researchers expect AGI before 2050 or even 2030.
Modeling the components of the human brain in order to copy them to a computer is likely to give us insights that make it easier to design AGI than it is to work out all the intricacies of copying a brain exactly. (An AI that works on the same principles as the human brain, but isn’t identical to it, is called “neuromorphic AI” or “brain-like AI”.) This suggests that WBE is unlikely to arrive before neuromorphic AGI3.
See this dialogue on a concrete plan for WBE by 2040 and on barriers to making it work. ↩︎
As of 2024, Hanson believes that emulated humans (which he calls ems) will arrive before other forms of cheap AGI. ↩︎
As an analogy, imagine trying to assess in 1900 (before the Wright brothers’ 1903 flight) whether the first flying machines that would be built would emulate birds with moving wings or would use a different design. One could intuit that understanding the aerodynamics of flight for birds could help in building the conceptually much simpler fixed-wings airplanes, and that if you had a mostly-functional prototype of a mechanical bird, you could adapt it into a functional airplane. ↩︎