The term “artificial intelligence” is oftenI mean neural networks built on the technology of deep machine learning. Moreover, the technology of training neural networks is well established and is bearing fruit. However, not all scientists share the view that artificial intelligence should develop along this path. Someone even believes that such systems are “not worth trusting” and their development will not lead to anything good.
Why machine learning is bad for human development
In large-scale work published on the pagesTechnologyreview, a professor at New York University, a specialist in the field of cognitive science (cognitive science), Gary Marcus, told about the potential for widespread use of neural networks based on deep machine learning.
First, the scientist believes that technology hasexplicit limitations. In particular, there has been talk for a long time about what is required to create the so-called “real AI”, which is suitable for solving a wide range of tasks, and not just one specific one, as is happening now. Existing AI systems have already reached the peak of their development and they practically have nowhere to grow. In addition, you can’t just take and, say, first teach one AI to drive a car, and the other to force it to be repaired and then combine the systems, creating a universal assistant. Artificial intellects simply will not be able to interact, since they "studied in different ways."
You can train AI to play Atari betterhuman, but to make a good robomobile is unlikely. Although this task is also rather highly specialized. Deep learning works well in the analysis of big data, but the algorithms do not see a causal relationship and poorly perceive any change in conditions. Move the elements in the computer game by two to three pixels, and the trained AI will become ineffective. Make the go pitch not square, but rectangular, and artificial intelligence will lose even to a novice player.
How to make AI smarter
To make algorithms moreeffective, they need to be "taught differently." It is necessary to make sure that they begin to see the relationship of objects and the consequences of interacting with them. In this case, we will serve as the best example.
Recruit intern students and they througha few days will begin to work on any problem - from law to medicine. Not because all of them are smart. And from the fact that people have a general idea of the world, and not a particular one.
Moreover, what Marcus offers is not new at all. The example described above is how scientists imagined “classic AI.” Only in order for such an AI to work effectively, we need to program all possible outcomes in advance. And this is almost unrealistic. But there is a way. By the way, which way of AI development is preferred in your opinion? Tell us about it in our chat in Telegram.
See also: How artificial intelligence works
The solution may be a kind of symbiosis“Classical AI”, which sees the relationship and receives decisions in an understandable way, and deep learning, able to find a solution through “trial and error”. This may be some basic system of rules and regulations relating to the world. Based on them, AI systems will already be able to develop themselves in a certain area. Real artificial intelligence must realize how everything works in order to understand cause-effect relationships and easily switch from one task to another. Modern systems created using deep learning technology are simply not capable of this.