Google’s AI is already capable of winning games without knowing its rules
One of the most advanced companies in AI issues is DeepMind which belongs to Google. In the past they developed AI programs capable of beating all their human opponents in various games, such as Go, Chess, etc … and now they have taken a new very important step with DeepMind MuZero, which is a new
AI program capable of playing any game without anyone telling it what the rules of that game are, and it becomes a “master” in it through self-learning and through trial and error.
This program watches how the opponent plays and learns what is allowed and what is not. When it begins to play a new game, it loses all the games, but it learns from his mistakes and the behavior of his opponents and it improves his knowledge of that game until it becomes a master at it in a short time.
If we think a bit about this it is how the human brain works, it also learns from its experience. But this program learns on its own and without anyone instructing it, and it does so very quickly.
It is based on some algorithms called “look-ahead search” that analyzes various potential movements and their possible results. I have said it in a very simplified way, but the internal calculations of the machine are surely very complex.
The great step of this DeepMind AI program is that it is able to “learn new things without having prior knowledge of them”; and that could be applied to a multitude of different applications.
From solving complex calculations, to developing new drugs, to …… almost anything we can think of. Or rather what the people of DeepMind can think of, because they are the ones who have the Know-How of their program and could adapt it to solve different types of problems.
From what some information indicates, DeepMind has already made a version to try to find a new means of improving video compression.
If we take into account that Google also owns YouTube, and they are the largest video managers in the world (1,000 million hours of video are viewed every day around the world), they would benefit greatly if they improved their compression and if they could transmit them to higher speed.
As we see, the possibilities are almost nameless. AI can be applied “to everything”, and that is why there are many people sensitized to the ethical use of this technology. There should be guidelines for ethical behavior to ensure that it is not used to harm people.
But there are many interests of very powerful organisms that will try to prioritize their interests over those of the general good.
The U.S. Force are already using these technologies (among other things) to launch missiles from U-2 spy planes; and I’m sure other countries will be doing similar things.
A shame, but we humans are like that.