Turning point for artificial intelligence?

This week, DeepMind introduced a new AI agent called Gato, capable of completing 604 different tasks “in a wide variety of environments”.

Citing DeepMind’s announcement, newspaper Daily Mail says Gato uses a simple neural network, which is a computer system of networked nodes that connect and act like neurons in the human brain.

As a result, Gato can chat, annotate images, arrange blocks with a robotic arm, and play 1980s video games…

Gato uses a simple neural network to complete 604 tasks. Photo: DEEPMIND

Even in a study published on the Arxiv electronic archive, DeepMind describes Gato as an “AI agent of general artificial intelligence”. Artificial General Intelligence (AGI) is a machine or program capable of understanding or learning any intellectual task that can be performed by a human.

After the announcement of DeepMine, page The Next Web new technology specialist commented: “Gato’s ability to multi-task is more like a video game console with 600 different games built-in than a game you can play in 600 ways. not AGI”. In addition, other comments did not appreciate Gato’s quality of completing each mission.

Before the above comments The Next WebMr. Nando de Freitas – DeepMind researcher and Oxford University professor – affirmed that DeepMind, through Gato, has solved the biggest challenges in the race to reach AGI.

According to Mr. De Freitas, the task now for scientists is to scale up AI programs, including upgrading computer power and adding data, to create a true AGI.

For a long time, AGI has been seen as a threat that could wipe out humanity in the future, intentionally or unintentionally. Dr Stuart Armstrong of the University of Oxford believes that the day will come when machines will operate at superhuman speeds compared to the human brain and when they will, they will take control of the economy, financial markets, transport, healthcare. …with no regard for us.

Dr. Armstrong gives an example: “Don’t let people suffer” – such a simple command could be interpreted by AGI as “kill all humans”, because human language is very powerful. easily misinterpreted.

Even genius professor Stephen Hawking before his death told the radio BBC: “The full development of artificial intelligence could mark the end of humanity”.

For its part, in a 2016 study, DeepMind admitted that a “strong mechanism” is needed to prevent machines from “completing a harmful sequence of actions”.

DeepMind was founded in London in 2010 and acquired by Google in 2014. The company is famous for creating an AI program that beat South Korean go player Lee Sedol in 2016.