Thursday, July 25, 2019

We Don't Need to Worry About Skynet Any Time Soon

Disclaimer: This is a rant, the only sources are personal anecdotes and it should be considered as food for thought and nothing else. Ranting in the comments section is welcome!

I read an article in Swedish magazine Axess the other day. In it they had interviewed a man named Ulf Danielsson, professor in theoretical physics (ie a lot of difficult maths) at the Uppsala University in Uppsala, Sweden. They discuss the dawn of AI and whether it is even possible for a program to gain consciousness. Ulf does not think so, in fact he says he doesn't believe in the duality of the body at all - the idea that we have some sort of "soul", "essence" or "consciousness" that can be spoken about as separate or even separable from the body. According to him the body is the consciousness. You don't have a body, you are the body.

I thought it was a very interesting line of reasoning that of course would have huge ramifications for science fiction. I also find it very plausible that he is correct. I don't believe in a duality either, I don't think there is something inside us that makes us "us" that isn't deeply connected and entrenched with what is also our body. Imagine all those body swap movies/animes/sci-fi books? Imagine all the fiction about downloading your consciousness into a computer or saving your brain in a jar? Would it ever be possible?

I think, therefore I am. But what am I?

If you really think about it, it seems like a daunting task. We already know that our mood, behaviour, feelings and thus everything we are come down to things like neurotransmitters. Without being an expert on anatomy, the neurotransmitters themselves are in turn affected by bodily functions that would have to be replicated as well. Could we build a machine that could successfully not only translate the function of these neurotransmitters into 1s and 0s but also make them function with a conscious mind the way they do in the brain and the body?

Even if we could, we now know that there seem to be other body parts than the brain itself that has a big impact on our personality. Lately it's become a bigger and bigger field of research to find out how our gut and the microbiome that lives in there affects everything from what we like to eat to how we feel and even think about things. While we know very little about these things yet, we can probably safely draw the conclusion that a brain without a stomach or any other microbiome would act differently to what we're used to (we're in fact more microbiome than human!). Of course we don't know how yet. And we also don't know if this extends to other body parts. How much will we ever really know about how our muscles, blood, lungs etc affect our personalities, our capability to form thoughts and what those thoughts are?

Even if I could picture us being able to take a snapshot of an active brain the way it is at any given moment, it seems like one big leap to go from that to being to progress it forward in a natural way. So I could basically have a mental image of my consciousness as it was at some date in time, but getting that to be active on its own, to be able to react to the surrounding to form new feelings, memories and thoughts?

I think what I am trying to say is that if we ever want to create an AI that could call itself conscious we would first of course have to know every intricate detail about what makes a human conscious (and call me a pessimist but I never think we will). And then we need to successfully translate that into a program that can not only run all that data in the way it needs to but also be able to react to the outside world the way it needs to. It can not be separated from input, and that input has to be a lot more than just audibly and visually.

Asimov's Robot-series is also very good.

In a way this then becomes similar to the discussion about whether we would ever be able to reliable predict the future. I know at least one really interesting sci-fi series has been written about this subject, I am of course talking about Isaac Asimovs "Foundation"-series. I believe it is theoretically possibly to have so much data about everything going on that you could predict how something, by law of nature, is going to go down (although the Foundation series is about the laws of society, but the core concept is the same. You need a lot of data). But practically it will never be possible for us to calculate all the variables. The same will be true when it comes to creating an AI.

We will probably be able to create programs that are clever enough to outsmart humans in different areas, even intricate conversations like the Turing test (didn't they already do that by the way?). But create a self-conscious robot/computer like in the books and movies? No, I don't think so.

5 comments:

  1. Since the 90s I've been following the work of roboticists that are focused on making low-level AIs and robots and then evolving what they learn to make more complex entities.

    I think they're on to something. Our own brains aren't one big chunk of intelligence. There are layers, and those layers correspond to earlier periods in our species' evolution. We're mentally and physically built in layers.

    So there's a lot to be said about the theory that mind and body and soul are interlinked.

    ReplyDelete
    Replies
    1. The word "layers" triggers my inner Shrek, but I definitely get your point. I find that believable too and think it just goes toward proving how much it is still for us to understand before we are anywhere close to building a consciousness. I mean, do we even agree on what a consciousness would look like and be?

      Delete
  2. This speculation reminds me of articles about the Cognitive Revolution, where sometime between 70000 - 30000 BCE humans had a spark (or a switch, your choice of metaphor) in their brains where they were able to make the leap from what their fellow hominids --and other animals-- could talk about and jump into the "why?" question with both feet. Prior to the Revolution, humans could make sentences such as "Beware a tiger!" if they saw a tiger nearby, but saying things like "The tiger is down by the river and is headed this way!" was beyond them. Or humans were able to imagine things that never existed, such as a human with a tiger head.

    Whether AI will be able to make that similar cognitive leap remains to be seen, because only humans have made that jump so far.

    ReplyDelete
    Replies
    1. Indeed. Even though there are other very clever animals around us (dolphins, whales) and some that are genetically very similar to us, still there are many things that differ between us and many things that make us have the upper hand when it comes to controlling our environment. Do we know which factors these are and will we be able to replicate them? Like you say, time will tell.

      Delete
  3. Exactly - one of those roboticists I read about was focused on building robots that would have the same mental ability as ants - and he wasn't there yet. But he had the opinion that we couldn't even KNOW if we could make a true AI until we understood what made US functional intelligences. Smart guy.

    ReplyDelete