The idea of AI, or “artificial intelligence”, is one that’s pretty central to transhumanism. You may even hear of something referred to as a machine “superintelligence”, usually in relation to discussions about a technological singularity.
The thing is, “artificial intelligence” comes in different flavors. When transhumanists talk about AI in future scenarios they are often referring to something known as “strong” AI. When we are talking about current day artificial intelligence such as Siri, Cortana, or even the cutting edge IBM Watson system, it’s referred to as “weak” AI. These two terms can be more than a bit confusing, so I thought it would be a good idea to write a short piece clearly separating the two ideas in language anyone can understand.
So, What’s “Weak” Artificial Intelligence Exactly?
All the AI we have created to date can be classified as weak AI. Why? Simply because they have such narrow applications. Our current artificial intelligence systems are not generally intelligent. They can make decisions and solve problems in a very limited area, but give them anything outside of that scope and you may as well ask a sack of bricks for its opinion.
An easy example to understand is the chess-playing computer. It made international headlines when IBM’s Deep Blue defeated a human chess grandmaster, Gary Kasparov. Did this mean that the artificial intelligence was smarter than Gary? Clearly not. It was (a bit) better at playing chess, but it couldn’t debate you on politics or appreciate art. Heck, it couldn’t even play any other board games. Back then, regular people didn’t really have access to AIs of any stripe. These days you have an AI assistant in your pocket, and specialized weak AI such as self-driving cars are going to be a fact of life very soon.
These weak AIs are also becoming incredibly good at their specific tasks. Although the term “weak AI” may make the impression that these are useless applications, weak AI may in fact be the most useful type of AI in general. After all, we don’t want cars that ponder the meaning of life or virtual assistants that learn to dislike us. Smart and purposely-built AIs are a fantastic solution to a lot of problems, and they are not going to go away.
And “Strong” Artificial Intelligence?
Strong AI is both easier and harder to explain. Basically, if you think of movie AIs or AIs from other media, that’s usually strong AI. Think Skynet or HAL9000. OK, I know both of those are actually evil characters, so maybe they aren’t the best examples. Unfortunately, kind and benevolent artificial intelligence don’t make for a good movie, so most of our examples are of bad guys.
Nonetheless, strong AI are intelligences that are at least at the level of what we’d generally consider human-level intelligence. It also includes the idea that the AI is apparently self-aware. I say “apparently” since self-awareness and consciousness as an experience are both things you can’t really prove objectively. Strictly-speaking, the only person I know for sure to be conscious and self-aware is myself. I have no way to know if people who aren’t me aren’t actually just so-called philosophical zombies.
Since we ourselves don’t really understand exactly why human beings are conscious, or the fine details of consciousness, it’s difficult to predict if or when we’ll achieve strong AI. Many transhumanists believe that once strong AI becomes a reality, we’ll very rapidly experience an intelligence explosion that will precipitate a technological singularity. Since the AI can design its own successor and do so at a speed that is beyond human comprehension, we’ll almost immediately give birth to a superintelligence that is only limited by the laws of physics, although not necessarily just the laws we know about.
At this point, strong AI is a similar kind of idea to fusion energy. We know it’s possible for fusion energy to exist because we see it in nature. We also know that human-level intelligence exists because we are living examples of it. However, knowing that the destination is waiting for you is different to knowing how to get there; the path to strong AI may be one we never discover.