We have always known computers as devices we can use to store and process data. But as it turns out, scientists have already found a way to teach them and make them learn like humans.
Researchers from Manchester Institute of Technology and New York University have found a way to teach computers by forming a set of algorithms that works the same way as humans learn. This particular algorithm is composed of unrecognizable alphabet letters that the team made.
The team has formed a 'Bayesian programme learning framework.' This allows the concepts to be represented as simple computer programs. One example of this is letter representation, where the computer will draw a certain symbol that corresponds to a specific letter.
According to the assistant professor of computer studies at the University of Toronto, Mr. Ruslan Salakhutdinov, it has always been hard to form machines that require the same small amount of data that humans need to learn. That's why the research is exciting, as it allows the machine to process data the same way as humans.
Over the course of the entire study and the implementation of the framework in the computer system, the software has been able to identify a symbol and write it correctly after just seeing one example. The team of researchers has tested the software's accuracy by drawing on a database of some of the world's most difficult languages.
The researchers then did a second test. This time they have presented a database consisting of over 1,600 characters from 50 alphabets from around the world and to make things even more difficult for the computer's recognition, and they have included a fictional alien alphabet from a TV show.
Another part of this particular test is having human subjects also try their best to recreate and draw the letters. Once the computers and humans are done, a set of judges was called on to identify which letter is drawn by whom. Much to their surprise, the computer-drawn letters are identical to those made by humans.
Mr. Brendon Lake from NYU has said that the result just shows that sometimes reversing how someone thinks about a problem will allow a better form of the algorithm to be developed.
