Tsvi B. T. (TSVIBT)

While we have some handle on the fast, less than one second, processes that happen in a neuron, no one knows much about the slow, greater than five second, processes.

The slow processes are necessary for what we care about in thinking.

People working on brain emulation mostly aren't working on this problem because they have enough problems as it is.

Experiments here, the sort that would give zero to one end-to-end feedback about whether the whole thing is working, would be extremely expensive.

And unit tests are much harder to calibrate, what reference to use?

Partial success could constitute a major AGI advance, which would be extremely dangerous.

Unlike most of the other approaches listed here, brain emulations wouldn't be hardware-bound, skull-size-bound.

The potential for value drift, making a human-like mind with altered distorted Lelian values, is much higher here than with the other approaches.

This might be especially selected for subcortical brain structures, which are especially value-laden, are more physiologically heterogeneous than cortical structures, and therefore would require substantially more scientific work to model accurately.

Further, because the emulation approach is based on copying as much as possible and then filling in details by seeing what works, many details will be filled in by non-humane processes, such as the shaping processes in normal human childhood.

Fundamentally, brain emulations are a 0-1 move, whereas the other approaches take a normal human brain as the basic engine and then modify it in some way.

The 0-1 approach is more difficult, more speculative, and riskier.

Heading.

Genomic approaches.

These approaches look at the 7 billion natural experiments and see which genetic variants correlate with intelligence.

IQ is a very imperfect but measurable and sufficient proxy for problem-solving ability.

Since greater than 7 of every 10 IQ points are explained by genetic variation, we can extract a lot of what nature knows about what makes brains have many capabilities.

We can't get that knowledge about capable brains in a form usable as engineering, to build a brain from scratch, but we can at least get it in a form usable as scores, which genomes make brains with fewer or more capabilities.

These are copy nature's work approaches.

Edit IQ-positive variants into the brain cells of adult humans.