AIs learn by inductive inference (same as us) and deductive inference (same as us). They also have some rules coded explicitly by their creators (we do to if you consider our instincts and our genes). But most of what they know, they know from learning. I can’t think of anything that, in principle, they couldn’t learn.

Sure, social interactions and empathy might be difficult to figure out — but we humans did it once. Why couldn’t an AI figure it out a second time?

A quick note on feelings & emotions: they may not work exactly the way that other things in our brain work but there is nothing magical about them that we couldn’t build too. Binary logic (like a computer uses currently) might not be the right way to build them (they are more like analogue systems from older machines, IMO) but I don’t see any reason why we could not build those too.

I want to counter this by asking what exactly would AI figure out on its own and how? Our system of feelings and emotions and reactions to other people’s feelings and emotions is a very complex and delicate system, partly physical, partly mental. AI is fed with all kinds of observations that we have made ourselves. Indeed, the AI could conclude things based on that. And indeed, that could be dangerous if it is connected to something that could kill us. It also could say some funny and surprising things (like John Oliver cites in his show). And that is why we must keep watch over what it actually does. But the way it learns is nothing compared to the way our biological system and mental system developed from hormones for example, and learned things from for example complex interactions with others. It only samples data, it performs logic and it calculates. On the other hand, if one has the view that human feelings and emotions are also only data, logic and calculations then my argument does not help to alter that. Except maybe that we can look how empathy develops in primates and then when one realises how complex that is, it may be possible to see that it is difficult to ‘figure that out’ on your own with only electricity as a power source. https://link.springer.com/article/10.1007/BF02382862

What’s an example, Betty, of something that an AI will not be able to do because it doesn’t have emotions, feelings or a mind?

Thank you, Betty!

AI can get pretty smart without feelings and emotions. I saw a chart in a book recently (Other Minds) that plots human intelligence against other kinds of intelligence (chimpanzees, dolphins, octopuses, AI, aliens). The vertical axis plots raw intelligence (are they smarter than us or not?). The horizontal axis plots how similar they are to us. An octopus is super smart (though not as smart as us) but it has its brains in its arms. Who knows if it has emotions?

This is kind of my point about separating the ordinary stuff the brain does (thinking, calculating, seeing, moving) from consciousness. You can get a very smart machine long before you figure out how consciousness works.

Eventually, we will make an AI smart enough that it can think for itself and invent new stuff without involving us at all. There are university departments that specialise in planning for how to make AIs “safe” when they get smarter than us. It’s definitely coming.

But emotions are just chemicals and electricity coursing through the body, initiated by thoughts in the brain. There’s no reason they can’t be simulated too.

Have a read of “Superintelligence” by Nick Bostrom to learn about all the ways AI can get super smart (and kill us) without us being involved at all. I think you’d be surprised by how much AI can do already.

As we get into the topic, I’m really surprised at how diverse the opinions are about what mind is and what it does.

For me, it’s obvious that my dog can think and can reason about problems and even imagine chasing rabbits in the movie in her head. I think lizards, parrots and octopuses can do that too. They all have minds.

The idea that mind can come from somewhere other than my brain just blows my mind. Sorry, Descartes! Sorry, K!

A computer can calculate, combine in an amazingly fast pace ridiculously amounts of data, true. But the mind is more than that. Emotions and feelings are not a simple question of comparing data in my opinion. But you may be right, Kevin. Your blog is wonderful, you are doing us all good.

Your essay made me think… if my brain was bubblegum and still processed information, then I would still have a mind if defined as a “thought/stimuli processor”. If so, animals have minds too… But that sounds kind of weird to me. To me having a mind also has something to do with human intelligence and reasoning. That’s why we say “you’ve lost your mind!” when someone is being unreasonable. At least these are my first thoughts anyways. ??

Thank you for posting

We don’t have to teach the AI how our minds work. We just have to teach them to learn and they will figure the rest out on their own.