Google says it’s about to achieve “human-level” artificial intelligence – what the project is about

Advances in artificial intelligence (AI) make progress beyond all logical reason. Algorithms are acquiring new skills every day and the ability to learn is becoming faster and more accurate.

Google has been developing the AI ​​pencil for 20 years and is now ready to leverage its expertise in an architecture called Catan artificial brain where it tries to do it replicate the functioning of human thought. However, the announcement was heavily criticized by other researchers.

What is General Artificial Intelligence (AGI)

The news that has spread is that the British company owned by Google, deep mindit might be about to catch up what is known as general artificial intelligence (AGI). That is, it has the ability to understand or learn any human intellectual task without specific training.

This new AI was able to complete 604 different tasks “in a wide range of environments”, using a single neural network: a computer system with interconnected nodes that functions like nerve cells in the human brain.

This new combination of algorithms has the ability to do things like chat, captions, stack blocks with a real robotic arm, and even play the 1980s Atari home video game console, says DeepMind.

To put it into context, Nando de Freitas, a DeepMind researcher and professor of machine learning at Oxford University, admitted that humanity is still a long way from creating an AI that can pass the Turing test an examination of the ability of a machine to exhibit intelligent behavior equivalent to or indistinguishable from that of a human being.

Cat: the doubts of the experts

While the Google Matrix considers Gato an evolutionary leap in this research, other scientists specialize in this field they don’t see it as something so revolutionary. They also didn’t hesitate to question DeepMind for its powers.

“It’s good work, but it doesn’t seem like a major step on the road to anything,” Mike Cook of the Knives & Paintbrushes research group told TechCrunch. “It might seem like he’s able to make a cup of tea or easily learn 10 or 50 other tasks and he’s not,” he said.

Another similar assessment, cited by El Confidentcial, was made by Matthew Guzdial, professor of computer engineering at the University of Alberta (Canada) “It’s a bit exaggerating”.

In defense of this innovation, Freitas tried a response through his Twitter account and defended that it all comes down to scalability.

“It’s about making these models bigger and safer, but also more computationally efficient, faster sampling, smarter memory, more modes, innovative data. Solving these challenges of scale is what AGI will generate, ”Freitas commented.

Gary Marcus joined this debate, one of the references in this matterthrough Twitter he refused that Gato has anything to do with general artificial intelligence, other than what he calls ‘alternative intelligence’, which “uses enormous amounts of data, often derived from human behavior, as a substitute for intelligence “.

SL