You don't have a flying car, jetpack or ray gun, but you still think that they will definitely appear in the future. Why? Because we are surrounded by artificial intelligence. It's funny when people ask when we will have smart computers, because they already have them in their hands. The phone calls are made by artificial intelligence. Whenever you write something into the search bar, you are using data that is collected by smart algorithms. Our world is full of these limited artificial intelligence programs, which we classify as "weak", "narrowly focused" or "applied".
These programs are far from the sensitive, loving, or power-obsessed artificial intelligences we saw in science fiction, but this is temporary. All of these highly targeted AIs are like the amino acids in the primordial soup of the Earth. Ingredients for true human-grade AI are created every day, and we won't have long to wait for results.
AI jungle surrounds us
In the late 1980s, scientists decided to rethink how they search for AI. Rodney Brooks of the Massachusetts Institute of Technology (also one of the founders of iRobot) found a new approach. Instead of developing AI from the top down, he decided to build it from the bottom up. Instead of artificial thinking, he decided to take up artificial behavior.
The result was robots that structured their actions based on instincts and models. The iRobot Roomba doesn't vacuum the floor, speculating on what the carpet should look like in the end, it just goes through different cleaning procedures until the carpet is clean.
This is AI-powered behavior, and it's powerful.
Along with the increase in computing power, artificial intelligence began to actively develop in the 90s. Using modular and hierarchical methods like Brooks' approach, scientists began to create a variety of AIs to perform all kinds of tasks. They were not philosophical programs, they were practical. Analyzing data, tracking stock and orders, processing images - all these tasks fell on the shoulders of AI.
Promotional video:
The list of tasks is growing. We are slowly building a library of highly focused AIs that impress us more and more. Speech recognition and processing allows computers to better convert sound to text and vice versa. Google uses AI to analyze millions of YouTube videos.
Computer vision is also improving, thanks to which programs begin to better recognize objects, classify them and understand how they move. Focused AI not only gets better at understanding its environment, it begins to understand the difference between what a person says and what he wants.
By assigning different meanings to different situations, focused AIs can make choices that maximize their reward. Thanks to this, they learn to overcome obstacles, more and more accurately recognize objects and text, build more and more complex logical chains.
Artificial intelligence is also getting better at analyzing and synthesizing big data. People who create music or paintings do the same thing.
These are all building blocks for the next explosion of AI tools.
Need an AI guard? Computer vision with the interpretation of human actions? How about a program that answers your little son's endless questions? Or all at once?
Of course, everything is not so simple: take it, plug it in and work. But every single-minded AI worker gets better and gradually starts doing more complex things.
There are three key factors that will enable us to create powerful AI.
We need more computational power to spoof and simulate the brain.
We need to better understand how the brain works and how it processes information.
We need to find ways to train AI to perform increasingly complex tasks.
All these factors are already being fulfilled to one degree or another.
Futurists predict exponential growth in computing power. Various projects are exploring our brain and looking for a way to mimic its functions. A growing number of highly focused AIs, methods for developing them, understanding their work, as well as deep learning and neural networks, are gradually meeting the third requirement.
There are many approaches in AI research and not all of them are compatible, but the more programs that perform simple actions we create, the faster the system of hierarchical thinking at the human level will be built.
When will the first lifeform of a sensitive computer appear? It is hard to say.
But the beginnings of its appearance are already scattered across the advanced technologies that we use day after day. Just take your smartphone and ask the artificial intelligence to tell you about the weather, about the ruble exchange rate, about hot tours. Doing so will strike the strings, stretch the veins of many modern AIs, and they will play in unison.
And so much the better for them.
ILYA KHEL