Strong AI will not happen this century

Imagine a world where things look the same twenty years later. Did things look very different twenty years earlier? You could fool me. Strong AI is a 22nd century problem at the earliest. Worrying about a world with Strong AI is a waste of time. You are thinking about science fiction.

 

I watched a video where by 2035, the speaker claimed we would have AI as smart as us. I remember Back to the Future where they predicted flying cars and hover boards by 2015. Don’t expect technology to improve because others say it will. The amount of effort to realize strong AI, it’s dismal. We are not even close. I don’t think we can have progress. What does your confirmation bias say?

The Vision problem is more complicated than previously realized. Sixty years ago, we knew we could make machines capable of moving in a natural environment. That is an easy thing to do because the world is made up of simple objects. But when we see things, how do we act in accordance? What does that mean?

There is almost an infinite amount of ways to look at a finite set of objects. Just knowing that fact is Devastating. Vision is a profoundly complicated problem. You can’t approach the vision problem without first figuring out the embodiment problem. If you don’t have a body, you can’t actually see. How else would you map something from the world unto actions? Looking at things and figuring out how to map your action onto it, is what seeing means. Seeing is mapping the world unto action.

To illustrate, a set of phenomenon can be seen an infinite number of ways. I see a book in front of me, I can open it up and read it. I see the function. I see the purpose. A beanie bag is a chair not because it has four legs and a seat, but because you can sit in it. Most of what we see in the world, we see it functionally. We see how we can use the object. That process limits our interpretation, but there are an infinite amount of interpretations. You can decompose objects into tiny details, you do this when you attempt to draw realistically.

AI people ran into this problem and they are working on it right now like mad. Looking at the world turned out to be exceptionally complex. The post modernists realized this. Say you take Shakespeare. There is an infinite number of interpretations you can derive from it. Hamlet is so complex and sophisticated. The individual words, a phrase, a sentence, a paragraph, and a chapter each are infinite in interpretations. There is an infinite amount of ways to interpret a text and therefore, the world.

Yes the world is super complicated, but you have to work out a game that you can play. It’s terribly un-useful to interpret from Hamlet that you should kill yourself and your family. We look at the world for useful tools to navigate, cooperate, compete, and continue our existence in a maintainable way.

 

There is an infinite set of interpretations; we have a tremendous number of constraints imposed to figure out how to live and thrive over multiple periods of time and multiple environments together. Figuring out these constraints are a mystery rather than a very hard problem. Constraints change and are unique. It may be hubristic to think we can understand the constraints for any given situation. That is why the vision problem is a very complex one.

 

I’d go far and say figuring out these constraints are a, ‘we don’t know problem and this kind of thing may be beyond our comprehension like how a refrigerator is beyond the comprehension of a dog.’ Wow, so dismal. Even with superb computational power, we don’t have the programs written out of what to even do for those constraints. People are going to disagree on the constraints. The Atheists think we are rational beings. Hmm, I don’t think so, and the scientific literature doesn’t agree either [1]. Just take a proper psychology course.

It’s a mystery why people do anything. Maybe you think you are a good person because you have running water, electricity, food, air condition, and you pay taxes. It’s mystery why people pay taxes. In Greece, they pretend to pay taxes. If everybody decided to stop paying taxes, the government wouldn’t have the ability to do anything. This isn’t easy.

Don’t be scared of technology, there’s not going to be a skynet. Here’s a contrary thought, technology isn’t going to get better because you expect it to. In the 1980s they thought technology was going to get so much better because just because. I have a feeling that hover boards and flying cars are simple compared to AI.

Technology won’t get better if you don’t build it.

 

[1] reading the book, Influence just makes that so clear.