The authors conclude that the work of neural networks is currently primitive compared to human language. Artificial intelligence systems are able to extract very limited information from text.
Fear of a robot invasion
The authors humorously write that if you are afraid that robots will rise up and attack people, you just need to lock yourself in your house, paint the handle so that it is invisible against the background of the door, attach a large poster of a school bus or a fancy-painted toaster to the front door, put a banana peel and nails on the floor, and for additional safety, you can put a table in the robot's path. All of these are insurmountable obstacles for robots. Moreover, in the process of overcoming obstacles, a modern robot will probably run out of battery.
Apocalyptic scenarios about a robot invasion are much closer to fiction than to reality.
being must be able to evaluate:
where is it located;
what is happening around her;
what needs to be done right now;
how to achieve this;
what needs to be done over the longer term to achieve existing goals.
All these issues must be considered continuously in the form australia number data of cycles that are continuously repeated.
In the field of artificial intelligence, progress is observed only in some parts of the cycle, while others remain without solutions.
Smart machines should have common sense and be much more reliable than existing ones
The authors believe that if our goal is to create human-like artificial intelligence, we need to move from systems that use statistical correlations as their primary learning tool to systems that understand the world through a basic understanding similar to that of humans. Machines need to have what we call common sense—the basic knowledge that we expect all humans to have. It's knowledge about how people and objects typically behave in different situations, how they use objects, what can be done with different things, and what does and doesn't happen in different situations.
To create artificial intelligence similar to human intelligence, it is necessary to embed in machines the principles that allow humans to learn and understand the world: the ability to abstraction, compositionality, cause-and-effect representations, the understanding that objects exist over time. It is necessary to embed in artificial systems the foundations in the form of an understanding of time, space, and causality. Therefore, when creating artificial intelligence, computer science must be enriched with knowledge from other disciplines, in particular cognitive science.
According to the authors, solving the problem of common sense and deep understanding is extremely difficult, but only this will help us create truly intelligent artificial intelligence.