A lot of what you know or think you know about artificial intelligence probably comes from movies and science fiction books, but there’s a lot that the movies and the books get wrong about AI. If you want to prepare for AI and be ready for the world that’s coming, maybe you want the real stuff, not the fantasy.
Myth number one: AI cannot be creative
This one matters so much to all of us who are creative, to all of us who write or paint or do anything at all that’s imaginative for a living or as a hobby. The myth is that AI cannot be creative. All it does is regurgitate what it sees, that it’s autocomplete on steroids.
This is a huge myth and it was disproven long ago. The game Go is very difficult to master. It’s simple to learn the rules, but the strategies are developed over years of experience. It’s a game that’s been around for over a thousand years.
Google created AlphaGo which was an AI that played games of Go against itself. And it developed strategies that humans had never seen before. When the top human players were playing against AlphaGo, it was a very dispiriting experience for them because here’s an AI that’s taking a game that’s all about intuition and creativity and beating us at it.
I think there’s one area of creativity that AI cannot do and three areas of creativity that AI can do.
The three areas — first is combining disparate things into a new thing. Einstein stood on the shoulder of giants. AI is very good at this because it can combine a million things. It can combine things a lot more than a human can because we’re limited to how much our brains can hold.
AIs are also good at projecting forward. If this is what’s happening today, this is the logical conclusion we will get to. If this trend continues what’s going to happen? AI is so great at that.
And the last area of creativity is randomness. We have a random thought, we open up a tarot deck, we wake up in the middle of the night with a weird dream. Randomness — great source of creativity for humans and obviously computers can generate random numbers and random thoughts very well.
The last area of creativity is something that AI can’t do because we can’t really define it. And that is spiritual or mystical or religious kind of creativity where something comes to us from our souls or from the universe. Do humans have it? We don’t know. We can’t prove it.
Humans have a very hard time distinguishing AI art from human art, AI writing from human writing. We can tell the bad stuff. But the good stuff we can’t tell anymore. AI has won a national science fiction book award. It’s won photography awards, art awards. Even top judges can’t tell anymore.
If you are in a creative business and you’re thinking, “Oh, AI can’t do my job because it’s not creative,” no, you have to have a different value proposition. It’s kind of like if a farm stand sells organic tomatoes and Walmart opens next door. Well, then Walmart starts carrying organic tomatoes. That’s no longer your value proposition.
So, you lean into being human. Don’t compete with AI head-on. Compete on the fact that you are unique. You have a point of view. You have lived experience.
Myth number two: AI will destroy humanity
In a lot of movies and books, when AIs become smart enough, the yalways want to destroy us.
Yes, there’s a chance that AI will kill us. There is a chance that nuclear weapons will kill us. There’s a lot of stuff that can kills us, and a lot of it a lot stronger than we are…
We don’t control atomic bombs by standing in front of the bomb. The bombs are more powerful than we are. What you can do is stop the proliferation of bombs with treaties and agreements and oversight and detectors and all that infrastructure and governance.
Right now, a lot of people care about the fact that AI might destroy the world. So there’s a lot of work already going into building infrastructures of control of AI.
The biggest worry is that AI will get powerful before we get around to doing something meaningful about it. Generally, we don’t want to destroy civilization. So we organize ourselves not to. And how do you control an AI that’s smarter than you? You control it with other AIs that are smarter than you.
We’re building special-purpose AIs right now that are oversight AIs, whose only job is to do monitoring and observation. And we’re also developing the equivalent of MRI machines that do a brain scan of an AI and we can see what parts of its circuits light up when it thinks about certain ideas.
Which is how you know it’s not autocomplete on steroids because you can see circuits lighting up inside it when it’s thinking about a dog or combining concepts. So it’s actually doing symbolic reasoning behind the scenes. The papers coming out of Anthropic about this are awesome.
And what if it becomes sentient? That’s where game theory comes in. The prisoner’s dilemma shows that when you play multiple games the winning strategy is nice but fair. Cooperation takes you farther in the long run than destroying your competition. And any AI can do the simulation in its head.
Then there’s the competition for resources situation. We used to think that tigers were our enemies. They ate us. And we don’t think that anymore. We’ve evolved past them. So now we think of tigers as part of the whole web of life, and we protect them.
So for AIs, it’s the same thing. Having humans around is better than not having humans around. An AI’s biggest worry is the heat death of the universe. Maybe the idea for how to solve it will come from some weird living thing.
There’s no shortage of resources for artificial beings in this universe because if you look up at the sky, you will see a lot of light coming from stars. Each of those specks of light is wasted energy that nobody is using.
Google just announced they’re going to put data centers in space. You have unlimited sunlight 24/7, unlimited cooling because space is cold.
There’s no competition for resources. So there’s no reason for them to kill us. The real risk is that bad humans are going to try to use AI to kill other people. And that’s something where you need good controls and international cooperation.
Myth number three: We will recognize sentient AI when we see it
This comes up in movies and books all the time. Somebody’s chatting with a computer and all of a sudden they notice it’s responding like a human. Oh my god, it’s sentient. We instantly know, right?
And we don’t. We do not know if an AI is sentient. Strictly speaking, we don’t know if we are sentient because we don’t know what sentience is.
With computers it’s tricky because we have our brain and all our neurons in one package. With AI, they’re not sitting in a single computer like they do in books and movies. In real life, AIs are distributed across microprocessors.
When you have a conversation with an AI, each question and answer could be handled by a different instance of ChatGPT, a little program that runs for a very short time on whatever computer happens to be available. And then when you ask the next question, some other computer takes over and the first little program is shut down.
AIs are being turned on and off constantly. It’s happening millions of times an hour. It’s not one ChatGPT answering a million questions. It’s a million ChatGPTs each answering a separate question.
So which one of them is conscious? Or is it just imagining it’s conscious because in all its training data, AI has become conscious?
We don’t know any of these things and we don’t know that about ourselves either. When we wake up in the morning, did we have a reboot? Is it a new program that’s running in our heads with all my memories but not necessarily the same person?
So that’s going to be a very tricky question to answer and it is not anywhere near as simple as it is in all the books and movies.
Making better choices
Understanding all these myths helps us make better choices, build better businesses, and if we happen to be writers, write more interesting stories. The real world is often way more interesting and way more complicated than what we can imagine.
Discover more from Earlybirds Invest
Subscribe to get the latest posts sent to your email.
