Demystifying Neural Networks: Teslas Are (Probably) Not Alive, But That’s OK! (Part 6)
We Do Have A Lot Of Guesses About Consciousness
This mystery of consciousness leaves a lot of room for guessing, and sometimes even educated guessing.
Materialists think that we just need to study the brain harder and that we’ll figure out what consciousness is by reducing it to matter. Unlike Galileo, the materialists think that the topics Galileo limited science to (size, shape, location, and motion) are basically all there is to the universe, so we must just keep trying to find the material thing that generates consciousness in the brain. Weak reductionists are materialists who think consciousness arises in the brain, but that we probably can’t use science to find the link due to differences in perceiving consciousness and neuroscience.
Dualists believe that the mind exists outside the brain, but there’s wide disagreement among dualists as to how this works. This can border on spirituality and religion, as this reflects the beliefs of the people of many religions. The idea that a person’s “spirit” exists outside of the body dovetails with that nicely.
Panpsychists believe that all matter has a little bit of mind in it, and that if you put enough complex matter working together, this will lead to consciousness at the level humans experience it. Goff (the author I cited earlier) tends to fall into this camp, and explains in great detail why he thinks it makes the most logical sense at present, and is thus worthy of further interdisciplinary investigation.
Others think that the universe is conscious, or that the brain somehow connects to the “fabric of the universe.” This area of thought tends to get a lot of new age spiritualists, mystics, and others involved, but there are scientists who wish to separate their study of this from the belief systems of others.
I have my own opinions on which of these is correct, and I’m not going to lie and say my spiritual views don’t affect my position. I’m not going to get into that here, though. We all have our beliefs and opinions on this unsettled matter, and mine is probably not any better than yours. This could be a fun topic to further explore in the comments, though!
Why Even Discuss All This Spiritual Mumbo Jumbo?
The only reason I bring all of this up is to solidly separate the mysticism and spirituality that invades discussions about consciousness and AI, and separate it from the harder facts. When it comes to the cutting edge of consciousness research, the field is very open and unknown right now. There’s a lot of information out there, and a lot of people making educated guesses, and it’s an exciting time.
There are also a lot of people claiming to have all of these answers already. Religious leaders, quantum mystics, and even people who basically worship Elon Musk all have very solid ideas of how this works (or doesn’t). I’ve seen people who are absolutely certain that their Tesla is alive, has saved their family from a situation it wasn’t designed to save them from, and more.
Everyone is, of course, entitled to their beliefs, but we’ve seen no evidence that Tesla’s vehicles have been imbued with consciousness, and we would be fools to count on that hypothetical consciousness for safety. We have to make sure there are other plans just in case they’re not alive.
Lack Of Consciousness Doesn’t Mean Self-Driving Cars Are Impossible
Absent some ground-breaking discovery or the accidental creation of consciousness in a computer (given the unknowns, we may actually blunder into this), we shouldn’t count on Tesla to come up with a living computer for its cars. To do that would be amazing, as the car’s computer could rely on its training when that was good enough, and improvise when it ran into gaps in its training, just like we do, but that’s out of reach as far as we can tell right now.
Without the ability to improvise, adapt, and synthesize new ideas, Tesla’s computers will instead have to run programs like every other computer ever built. That leaves them relying completely on how applicable their training data is to the real world. The more applicable the training data was, and the better the training was, the fewer edge cases there will be.
It’s basically impossible to ever get to 100% reliability, because the world is simply too complex to ever gather enough training data to cover it all. But, Tesla can come up with a system that’s almost 100%, and that’s good enough.
If Tesla’s FSD neural net is applicable to 99.9% of drives, that means that in a year’s worth of driving, the computer will make mistakes for about four hours of that year, or around a minute per 24 hours of driving. Assuming you drive 2 hours per day, that’s somewhere in the ballpark of 5 seconds of erratic driving every day, and that will happen randomly.
We all know that a lot can happen in 5 seconds on the road, and that could easily kill you, so people wouldn’t be happy at all with a system that’s only 99.9% good.
Fortunately, Elon Musk has said they’re aiming for “eight nines” of reliability. That’s 99.999999% reliable, or six more nines after the 99%. Awesome!
But what does that mean? It’s about 1/3 of 1 second per year of driving. Keep in mind that we’re talking about 365 24-hour days of full driving (which nobody really does). If you drove two hours per day, that 1/3 of a second of error would happen every 12 years of ownership.
Now we’re talking about a system that human passengers could seriously trust.
Achieving that will not be easy at all, though. When it comes to computer services, including services for hospitals where human lives could be on the line at times, there are still questions about whether it’s worth the extra cost to achieve even “five nines” (99.999%). Many non-essential services promise somewhere around 99.0% uptime. Sure, we’d love all things we use to be “eight nines” reliable, but we’re not usually willing to pay the costs involved.
I don’t pretend to know what the costs of achieving “eight nines” will be, but it’s clearly going to be a huge cost that takes a lot of work to achieve and a lot of work to keep at that reliability level. I do see Tesla putting in serious work on that front, so I’m cautiously optimistic about their chances of success.
For ease of navigation for this long series of articles, here are links to the rest of them:
Part 1: Why Computers Only Crunch Numbers
Part 2: Miscalibrated Trust In Mathematics
Part 3: Computers Only Run Programs
Part 4: How Neural Networks Really Work
Part 5: What Artificial Neural Networks Can’t Do
Featured image: Screenshot from Tesla’s AI Day.
Have a tip for CleanTechnica? Want to advertise? Want to suggest a guest for our CleanTech Talk podcast? Contact us here.
CleanTechnica Holiday Wish Book
Our Latest EVObsession Video
CleanTechnica uses affiliate links. See our policy here.
