Expert View: The Possibility of General Artificial Intelligence

One of the challenges of following news about developments in the field of artificial intelligence is that the term “artificial intelligence” is often used indiscriminately to mean two unrelated things.

The first use of the term AI was more accurately called narrow AI. It's a powerful technique, but it's also very straightforward: You take a bunch of data about the past, use a computer to analyze it and find patterns, and then use that analysis to predict the future. This type of AI touches our lives multiple times a day as it filters spam from our emails and directs us through traffic. But because it is trained on past data, it only works where the future is similar to the past. That's why it can recognize cats and play chess, because they don't change from day to day on a basic level.

Another use of the term AI is to describe what we call general AI, or commonly known as AGI. Except in science fiction, it doesn't exist, and no one knows how to make it. General artificial intelligence is a computer program that is as intelligent and versatile as humans. It can teach itself completely new things it has never been trained on before.

The difference between narrow AI and general AI

In movies, AGI is Data in "Star Trek", C-3PO in "Star Wars" and "Blade Runner" clone. While it might seem intuitively that narrow AI is the same thing as general AI, just a less mature and complex implementation, this is not the case. General AI is different. For example, identifying spam does not computationally equate to true creativity, something that general intelligence can do.

I once hosted a podcast about artificial intelligence called “Voices in Artificial Intelligence.” It's interesting because most great science practitioners are approachable and willing to be on podcasts. So I ended up with over 100 great AI thinkers having an in-depth discussion on this topic. There are two questions I ask most guests. The first question was, “Is general artificial intelligence possible?” Almost everyone—with only four exceptions—said it was possible. Then I ask them when we can build it. The answers vary, some are available within five years, and some are as long as 500 years.

Why is this?

Why do almost all of our guests say that general artificial intelligence is possible, but provide such broad estimates of when we will achieve it? The answer to this question is Going back to something I said before: we don't know how to build general intelligence, so your guess is as good as anyone else's.

“But wait!” you might say. “If we don’t know how to make it, why do experts overwhelmingly agree it’s possible?” I ask them this question too, and I usually get are different versions of the same answer. They believe that we will build truly intelligent machines, which is based on a core belief: humans are intelligent machines. Because we are machines, reasoning is like this, and have universal intelligence, manufacturing has universal intelligence. Intelligent machines must be possible.

人与机

What is certain is that if people are machines, then those experts are right: general intelligence is not only possible, but impossible Avoided. However, if it turns out that people are more than just machines, there are things about people that may not be replicable in silicon.

What’s interesting is the difference between these hundreds of AI experts and everyone else Disconnect. When I give a talk to a general audience on this topic and ask them who thinks they are a machine, about 15% of people raise their hands afterward, compared to 96% of AI experts.

On my podcast , when I refute this assumption about the nature of human intelligence, my guests usually accuse me—very politely, of course—of indulging in some kind of magical thinking that is anti-science at its core. “If it weren’t for biological machines, we would What else could it be?"

It's a fair and important question. We only know of one thing in the universe with average intelligence, and that's us. How could we happen to have such great creativity? Do we really Don’t know.

Wisdom is a Superpower

Try recalling the color of your first bike or the name of your first grade teacher. Maybe you haven’t thought about either of those things in years , but your brain will probably be able to get them back with little effort, which is even more impressive when you consider that the "data" isn't stored in your brain like it is on a hard drive. .In fact, we have no idea how it is stored. We may find that each of the one hundred billion neurons in your brain is as complex as our most advanced supercomputers.

But That's where the secret of our intelligence lies. It gets trickier from there. It turns out we have something called a mind, and it's not the same thing as the brain. The mind is all the three pounds of goo in your head can do , like having a sense of humor or falling in love, it seems like it's not what it's supposed to do. Your heart doesn't, and neither does your liver. But somehow you do it.

We are not even sure that the mind is just a product of the brain. Many people lose up to 95 percent of their brains at birth but still possess normal intelligence and often do not learn of their condition until a diagnostic test later in life. Furthermore, it seems that much of our intelligence is not stored in our brains but is distributed throughout our bodies.

General Artificial Intelligence: The Complexity of Consciousness

While we don’t understand the brain or the mind, it actually gets more difficult from there: General intelligence likely requires consciousness. Consciousness is your experience of the world. A thermometer can tell you the temperature accurately, but it doesn't feel warm. The difference between knowing and experiencing is consciousness, and we have little reason to believe that a computer can experience the world like a chair.

So now we have a brain that we can't understand, a mind that we can't explain, and as for consciousness, we don't even have a good theory to explain how mere matter can possibly have an experience. However, despite this, AI proponents of AGI believe that we can replicate all human abilities in computers. This seems like magical thinking to me.

I don’t say this to belittle anyone’s beliefs. They're probably right. I just think the idea of ​​general artificial intelligence is an unproven hypothesis rather than an obvious scientific truth. Building such a creature and then controlling its desires is an ancient dream of mankind. In modern times, it has been around for centuries, perhaps starting with Mary Shelley's Frankenstein and then manifesting itself in the 1,000 stories that followed. But it's actually much older than that. We have imagined this as far back as we have writing, such as the story of Talos, a robot created by Hephaestus, the Greek god of technology, to defend the island of Crete.

Somewhere deep within us longs to create such a creature and control its awesome power, but so far there's no sign that we actually can.

The above is the detailed content of Expert View: The Possibility of General Artificial Intelligence. For more information, please follow other related articles on the PHP Chinese website!