Uploading, not Avatars

There’s an interesting New York Times article about Dmitri Itskov, a Russian multimillionaire who’s actively (and openly) pursuing mind uploading.  But, as you might expect in an article written for a nontechnical audience, it abounds with errors and incorrect terminology.

BotScene is a blog about humanoid robotics, and no robot could be more humanoid than a robot who actually is human.  And, as it turns out, I’ve been exploring mind uploading for many years.  So, let me take this opportunity to try to clear a few things up, starting with some terminology.

Cyborg: a biological human with mechanical implants.  Anybody with a pacemaker or artificial limb is, in a minimal sense, a cyborg.  The term often has a negative, nightmarish connotation, like the Borg of the Star Trek series.  But by any reasonable definition, fictional superhero Steve Austin (aka “the Bionic Man”) was a cyborg too.  Itskov’s project has almost nothing to do with cyborgs, except perhaps as a small, not-too-relevant stepping stone.

Avatar: a software or hardware remote controlled by a person, to act as a stand-in for that person in another environment.  In other words, an avatar is something you “project” yourself into and control, and other members of that environment can interact with it as if it is you.  The original meaning was a religious one: a Hindu deity in human form, but in modern parlance, any game character you directly control is your avatar.  The term would also apply to telepresence robots.  It does not, however, apply to Itskov’s project.  Avatars are just a form of communication; they don’t confer any longevity or other benefits on the user, whose real body and brain exist somewhere else.

Mind uploading: this is the hypothetical process of copying (uploading) your complete brain pattern into some artificial substrate.  Since your brain pattern accounts for everything you know, feel, hope, fear, believe, and so on — everything that makes you uniquely you — this amounts to transferring you into some artificial brain, housed in an artificial body.  At that point, you go about your life much as you always have, except in a (hopefully) more durable, capable body, and with the benefit of frequent backups that make you essentially immortal.  This is Itskov’s project.

The reporter can perhaps be forgiven for messing up this terminology, because Itskov’s project materials confuse the issue substantially.  The 2045 Initiative technology graph shows four milestones:

1. “Avatar A”  (2015-2020): Humanoid robot avatars controlled through a brain-computer interface.
2. “Avatar B” (2020-2025): Robot bodies with a life-support system for a disembodied brain.
3. “Avatar C” (2030-2035): Synthetic carriers of personality and consciousness (i.e., actual mind uploading).
4. “Avatar D” (2045): A “hologram-like avatar” perhaps composed of nanobots or some other future technology.

Probably they have tried to simplify things for the general public, but it’s clear that while Avatar A is actually an avatar, Avatar B is a cyborg, and Avatars C and D are mind uploads.

So.  Unless you’ve been thinking about this sort of thing for years, you’re probably double-checking your calendar to see whether this could be some belated April Fool’s Day joke.  In short, we have to ask: is this guy nuts?

My own view is that he’s fallen into the common futurist mistake of overestimating progress in the short term.  We are not going to have brain-computer interfaces anywhere near good enough to control an entire avatar by 2020.  By 2030, maybe.  But probably not.  Brains have no convenient data ports; options for interfacing with it are extremely limited, and the most effective ones are also quite dangerous.  Telepresence robots are here today, and will continue to improve, but the best interfaces for normal people will probably always be something like a joystick and screen, using our brain’s existing I/O (voluntary muscles and the five senses) to control them.

And as for carrying a biological brain around in an artificial body — that probably is nuts.  See previous comments about squishy biological things with no good data ports.  Prosthetics will continue to improve, of course, including neural interfaces to sensorimotor nerves.  Someday, maybe in a century or more, this could lead to the extreme case of someone whose entire body is prosthetic, while the brain remains biological.  But this is hard.  Very hard.  In the next few decades?  Not likely.

In contrast, what they call “Avatar C” is much easier.  This is mind uploading.  We know, in principle, how to do this, and it avoids all the incredibly difficult complications of interfacing with a complex biological organ like the brain: we cryoprotect it, freeze it solid (at which point it is no longer squishy), and take it apart layer by layer, scanning every neuron and connection into the computer.  Then we pour all that data into a simulation (like those being built at the Blue Brain Project) and turn it on.  The patient wakes up and gets on with their life.  There is steady progress being made towards this goal in neuroscience labs around the world, and if Itskov wants to invest in it, I’m all for that.  But it’s not an avatar… it’s an upload.  And Avatars A and B are more distractions, than steps towards this goal.

Incidentally, the second 2045 Global Future Congress (sponsored by Itskov) will feature some true visionaries, such as Marvin Minsky and Ray Kurzweil, as well as a couple of crackpots (I’m looking at you, Hameroff and Penrose — you won’t find God in the microtubules no matter how hard you look).  An old friend of mine, Randal Koene, will also be speaking about practical neuroengineering work being done towards mind uploading.  Dragging the topic back a bit towards the focus of this blog, android builders Hiroshi Ishiguro and David Hanson will also be on hand.  The congress takes place in New York on June 15-16, so if you happen to be in the area, I would highly recommend it.