How could AI be used to augment human biological intelligence?

Question

One of the most compelling applications for AI would be in augmenting human biological intelligence. What are some of the currently proposed methods for doing this aside from vague notions such as "nanobots swimming around our brains and bodies" or "electrodes connected to our skulls"?

Answer 1

'Direct augmentation' of human intelligence, of the sort that you would see in science fiction, looks to be very hard. Most of our promising approaches deal with avoiding damage rather than adding capabilities--there's no drug that you can take now that will make you smarter to the degree that missing a night of sleep can make you dumber.

The most informative area of current practice is probably game-playing, where 'centaurs,' or humans working with computers, outcompete human players or computer players.

But a centaur player doesn't have a wire jutting out of their skull to jack into the computer; they're looking at a laptop screen. One of the reasons to be pessimistic about cyborg augmentation is because current I/O technology is already so good. Why install a new wire to put information into your visual cortex, when you come already equipped with two?

If you could think code directly onto the screen, how much better would that be than typing code through a keyboard? Probably some, but I find it difficult to imagine that it'll be more than twice as good. So most human-computer intelligence augmentation will look like people using software, and software using human inputs, rather than humans and computers evolving together.

---

Transcranial Direct Current Stimulation (TDCS) and similar approaches cause temporary changes in mental abilities by raising or lowering the activation potentials of neurons in particular regions of the brain. (I've done it myself, a few years ago, and what weak effects I noticed were probably negative. Not too much surprise for a DIY setup!)

It looks like it has a number of useful implications. One article about TDCS that I found particularly striking was the journalist who tried it gushing about how their anxiety disappeared for a few days, presumably because the part of their brain behind the anxiety was dampened. One could imagine it being useful for the treatment of many different mental disorders.

That said, I'm pessimistic that it will translate into superior peak performance, and I think that's the sort of thing that's more relevant for discussions of augmentation. (Is there TDCS that we could do that would make Terrence Tao better at doing mathematics?)

Where improved AI methods will come into play is by improving our models of the brain, allowing us to better target interventions, much in the way that AI methods are improving our treatment of cancer (through superior diagnosis and targeting of radiotherapy, as two easy examples). These effects will all be indirect--for example, AI empowering an app or gadget that helps you sleep better won't directly augment your intelligence, but will cause population-level increases in effective intelligence through reducing sleep deprivation.

I haven't talked yet about nootropics, chemicals that increase intelligence, but it's reasonable to expect that AI will improve drug discovery there like it improves drug discovery for anything else. But the same caveats apply--the effect of nootropics seem to be negatively correlated with intelligence (that is, the smarter someone already is, the harder it is to increase their intelligence further).