No.

While we are in favor of human intelligence augmentation, we don’t think this technology offers a realistic chance of keeping up with unrestrained AI progress. Human augmentation technology is still in its infancy, and is much more constrained than AI as a method to produce ever-greater intelligence. Likewise, brain-computer interfaces won’t realistically let humans keep pace with AIs.

By analogy: Augmented humans wouldn’t be competitive with a superintelligence any more than cyborg horses built with 1908 technology could have been competitive with the Model T.

It’s possible, in principle, to build a cyborg horse that can keep pace with the fastest racecar. But you don’t get cyborg horses as fast as racecars before racecars, and you don’t get them at around the same time as you get cars. Not even if you start trying to build cyborg horses two to twenty years before the first mass-market car comes off the assembly line.

Making brain-computer interfaces that work well enough to be game-changers is a high bar. It might seem cool to imagine information pumped straight from the internet into your brain, but there already exist technologies that let you pump information from the internet straight into your brain: screens. The human visual cortex is actually quite good at soaking up information (words) in a format your brain can digest. For a brain-computer interface to load knowledge into your head faster than you could get it into your head by reading, it would have to do more than just dump the data in your brain somewhere; your eyes already do that part fine. Uploading skills and knowledge and experience would require it to interface in just the right way with your thoughts and your implicit beliefs and your existing skills, and that’s a much taller order.

We’re not saying that this can’t be done; we’re saying that brain-computer interface technology today doesn’t seem anywhere close to solving the tricky parts of the problem. So far as we know, psychologists and neuroscientists and cognitive scientists are still pretty far from decoding the “data format” of thought and belief and experience in a way that would allow experiences to be loaded directly into a human brain.^*

Similar issues arise when it comes to outputs. It’s hard to beat keyboards and mice and joysticks and steering wheels. It’s not impossible. It’s just that technology today (e.g., hooking wires into a paralyzed person’s head in order to let them type and use a mouse), wonderful as it is, isn’t very far along the pathway that would allow humans to go toe to toe with (even relatively weak) superintelligences. It’s a good path to pursue, but it’s not a competitive path to pursue.

Indeed, it’s not clear that brain-computer interfaces allow any hope of humans competing with superintelligences. What does it matter if a human can download experiences from the internet and control ten computers at once with their mind, if an AI can do the same thing — but ten thousand times faster, while controlling a million computers at once? We think the whole project of trying to get humans to keep up with AIs is doomed.

That said, humanity should be augmenting humans.

We don’t think augmented humans would ever be able to go toe to toe with superintelligences, but smarter humans might nevertheless be able to help humanity find a way out of this mess! We mention this possibility in Chapter 13 and discuss it further in the associated online resources.