Artificial intelligence that can emulate human reasoning and even emotions is within Google’s reach, according to a leading AI scientist employed by the company.
Talking to the Guardian, Professor Geoff Hinton, an expert in artificial neural networks, said Google is “on the brink of developing algorithms with the capacity for logic, natural conversation and even flirtation.”
Google is working to encode thoughts as vectors described by a sequence of numbers. These “thought vectors” could endow AI systems with a human-like “common sense” within a decade, according to Hinton, who has been working at Google for two years.
Language and reasoning
Currently in early development, thought vectors may help AI master natural language conversation and make leaps of logic, two abilities that we perceive as distinctly human and that AI scientists have long sought to unlock.
Hinton said he even foresees AI that is so human-like that we will become attached to it, the way Joaquin Phoenix’s character in Her falls in love for his AI assistant.
Google has very practical applications for AI: currently, the company is using “thought vectors” to improve Google Translate, aiming to add a sense of meaning to the current dictionary-based translation algorithms.
The potential of AI is huge for Google; truly relevant search results, human-like translation, a genuinely helpful Google Now virtual assistant are just a few of the applications that only advanced AI can enable. Google has been acquiring top talent to further this goal, including through the $500 million acquisition of British startup DeepMind, considered the top player in deep learning algorithms.