Computer scientists are a step closer to perfect softwares now that they created and algorithm which allows artificial intelligence to learn like humans.
- Machines can recognize simple patterns
- The goal is to make computers learn faster with only a few examples
- Scientists hope to bring artificial intelligence to the same level of intelligence a human child has
You might think computers are a lot smarter than us and you might be right in some ways. However, the human mind has a unique way of functioning and we are able to easily recognize the things that surround us, we become aware of them and we understand them by making connections. After seeing and understanding we are able to recreate what we see either through describing it in words, or drawing it.
What scientists have done through extensive research was to develop an algorithm that mimics some of the human learning abilities. Thanks to this algorithm computers can now easily recognize simple visual concepts and then draw them. This helps to reduce the time it usually took artificial intelligence to understand any new concepts.
To get a better understanding of how the human brain works think about the fact that a person only need two examples of a thing, whether that thing is an animal or a letter from another alphabet or a new dance move, to understand the object. Of course the understanding is quite basic only after two examples, but they can later recognize objects that are similar or come from the same ‘family’. So what we recognize and understand are actually patterns.
A computer does not perform as well as a human does when it comes to pattern recognition because the task is too complex for it. A computer can need thousands of examples to begin to be as accurate as a human.
Obviously the purpose of the research was to speed things up a bit and try to bring artificial intelligence closer to human intelligence in pattern recognition. Recognizing a pattern from only a few examples will accelerate the learning process. The computers went from recognizing a pattern after thousands of examples to recognizing it after a couple and then started generating other similar examples. That’s basic learning.
For this particular research scientists have applied a BPL framework which produces different outputs for each execution. So the computer shouldn’t reproduce the same examples it saw, but generate other examples, just like two people having a different writing.
The next goal is to try and bring the machines to level up with the intelligence of a human child. In this way they should be able to learn and also create new things. But artificial intelligence is still far from human’s, regardless of the human’s age.
Image source: www.bing.com