DeepMind: Deep learning algorithms can overcome human intelligence in many ways: from image sorting, to speech reading from lips, to accurate predictions for the future. But despite their hyper-human levels of competence, they are disadvantaged at the rate at which they learn.
Some of the best mechanical learning algorithms need hundreds of hours to study and learn classic video games, something a man can learn in an afternoon. The fact may be somewhat related to neurotransmitter dopamine, according to a publication of the subsidiary companyGoogle's DeepMind in the journal Nature Neuroscience.
Post-learning or the process of quick learning from examples and the acquisition of rules from these examples over time is believed to be one of the ways in which people acquire new knowledge more effectively than algorithms. However, the main mechanisms of post-learning are currently poorly understood.
In an effort to shed light on the process, DeepMind researchers in London modeled human physiology using a recurring neural network, a type of neural network that is capable of internalizing previous actions and observations and learning from these experiences. The system that mathematically optimizes the algorithm over time through tests and errors is reportedly using dopamine, a chemical in the brain that affects feelings, movements, sensations of pain and pleasure, and plays a key role in the process learning.
Οι ερευνητές λοιπόν δημιούργησαν ένα παρόμοιο σύστημα σε έξι πειράματα νευροεπιστημονικής μετα-μάθησης, συγκρίνοντας την απόδοσή του με εκείνες των ζώων που είχαν υποβληθεί στις ίδιες δοκιμές. Μια από τις δοκιμασίες, γνωστή και ως Πείραμα του Harlow, έδωσε στον αλγόριθμο να επιλέξει δύο τυχαία επιλεγμένες εικόνες, η μία από τις οποίες συνδεόταν με μια ανταμοιβή. Στο αρχικό πείραμα, μια ομάδα πιθήκων έμαθαν πολύ γρήγορα μια στρατηγική για τη συλλογή των ανταμοιβών. Επέλεγαν ένα αντικείμενο τυχαία την πρώτη φορά, αλλά αμέσως μετά τα objects who possessed the reward.
Ο algorithm he worked more or less the way the animals did, choosing images that were directly associated with rewards from novel images he had not "before seen". In addition, the researchers noted that learning took place through the neural network, supporting the theory that dopamine plays a key role in post-learning.
The study of dopamine shows that medical science has much to gain from neural network research, just like computer science.
"Harnessing data from AI that can be applied to explain findings in neuroscience and psychology highlights the value of each field to the other," the DeepMind team says. "Moving forward, we expect many benefits from the opposite direction as well, having instructions from this one organization brain circuits to design new models that learn from augmented AIs”.
- Data centers: Using the heat of the servers for heating!
- Microsoft, Apple, Google, Facebook, Amazon and captivity
- The most bizarre planning principles
