The research presented by “Apple” on the techniques of deep thinking in artificial intelligence models caused a wide global uproar, as the research indicates that the great inference forms of artificial intelligence are unable to think well despite what companies are trying to promote, according to the “Mashable” website.
“Apple” presented the research several days before the annual developer conference, in which it indicated that the models such as “Oben AI or 1”, “or 3” and “Deep CICA1” completely fail to solve complex problems with the increase in the degree of failure as the complexity of the problem increases, and this research comes from the team that published a previous research paper that showed the shortcomings of great linguistic intelligence models Last year.
The “Apple” research shows that the performance of deep thinking models is a little better than the usual linguistic models of artificial intelligence in the face of minor problems, but with the increase in the complicate rate and the problem reaching an average stage of complexity.
That is, the model fails to solve challenges and reach the desired results, and the research describes this failure that deep thinking models enjoy good performance when facing software or arithmetic problems, but they fail to face logical problems.
In order to implement these tests, Apple has fully relied on traditional logical problems, such as the Hanoi Tower or the river crossing problem that usually includes a group of pets, predators, some plants and the degradation of blocks in a specific composition.
While skeptics of the abilities of artificial intelligence found their goal of this research, artificial intelligence experts find that research does not necessarily mean that deep thinking models are bad and do not do their work well, but they indicate that these models at the present time are weaker in logical thinking of humans.
“Humans actually suffer from a group of well -known restrictions that are identical to what Apple’s team discovered,” Gary Marcus said on his blog. Artificial intelligence with humans while trying to solve these logical puzzles.