Where are we at 1. Generalization skills So lets start from the apparently simplest one to conquer generalization or the ability of an algorithm to perform multiple tasks . Among the various experts who are working on this aspect of the evolution of artificial intelligence we find Demis Hassabis CEO of DeepMind the advanced research laboratory owned by Google. According to Hassabis the key to arriving at a generalpurpose artificial intelligence which doesnt necessarily mean smarter but capable of doing more things is learning. In extreme synthesis the transfer of learning would make it possible to reuse for a second objective a model already used to complete a previous task without erasing everything.
Ability to abstract However in order to abstract our
Previously learned. The underlying idea is that this prior knowledge gained thanks to the first task will allow the AI to perform tter to trained Ecuador Phone Numr List faster and with less data than a neural network trained from scratch only on the second task reads on Hacker Noon . If one wants to arrive at AGI by exploiting. This technique transfer learning must applied in areas that are very distant from each other . Only in this way would it demonstrated that the machine possesses or at least can simulate the ability to abstract what has en learned in one field and then reuse it in another.
For now however significant results have only
en achieved in the transfer of closely related knowledge for example there are deep learning models that are capable of playing various video B2C Phone List games. New call to action 2. knowledge from single events or single skills another crucial element is needed the understanding of the relationship of cause and effect tween two related events . For example if you want to understand how rain works you need to know that clouds cause it. At the moment however artificial intelligences are only able to establish through statistics. That there is a correlation tween the presence of clouds and the presence of rain without however knowing which is the cause and which is the effect.