DeepMind, a UK company owned by Googlecould be on the verge of reaching artificial intelligence (AI) at the human level.
Nando de Freitas, researcher at DeepMind and professor of machine learning at Oxford Universitydeclared that “the game is over” when it comes to solving the toughest challenges in the Artificial General Intelligence (AIG) race.
AGI refers to a machine or program that has the ability to understand or learn any intellectual task that a human being can without training.
According to De Freitas, the quest for scientists now is to scale up AI programs, for example with more data and computing power, to create an AGI.
Earlier this week, DeepMind unveiled a new AI “agent” called Gato that can perform 604 different tasks “in a wide range of environments”.
Gato uses a single neural network – a computer system with interconnected nodes that functions like nerve cells in the human brain.
He can chat, caption pictures, stack blocks with a real robot arm, and even play the 1980s Atari video game console, according to DeepMind.
Scroll down for video
DeepMind, a British company owned by Google, could be on the verge of realizing artificial intelligence at the human level (file photo)
Gato uses a single neural network — computer systems with interconnected nodes that function like nerve cells in the human brain — to perform 604 tasks, according to DeepMind
De Freitas’ comments came in response to an opinion piece published on The next web it says that humans alive today will never reach AGI.
De Freitas tweeted: “It’s all about scale now!” The game is over! It’s about making these models bigger, safer, more computationally efficient, faster…’
However, he admitted that humanity is still a long way from creating an AI capable of passing the Turing test – a test of a machine’s ability to display intelligent behavior equivalent to or indistinguishable from that of a human.
After DeepMind’s announcement on Gato, The Next Web article said it demonstrated no more AGI than virtual assistants such as Amazon’s Alexa and Apple’s Siri, which are already on the market. market and in homes.
“Gato’s ability to multitask is more like a video game console that can store 600 different games, than a game you can play 600 different ways,” said The Next Web contributor Tristan Greene.
“It’s not a general AI, it’s a bunch of neatly grouped pre-trained narrow models.”
Gato was designed to perform a variety of hundreds of tasks, but this ability can compromise the quality of each task, according to other reviewers.
De Freitas tweeted: “It’s all about scale now!” The game is over! It’s about making these models bigger, safer, more computationally efficient, faster…’
In another opinion piece, ZDNet columnist Tiernan Ray wrote that the agent “is actually not that good at multi-tasking”.
“On the one hand, the program is able to do better than a dedicated machine learning program for controlling a Sawyer robotic arm that stacks blocks,” Ray said.
“On the other hand, it produces captions for images which in many cases are quite poor.
“His ability to dialogue via standard chat with a human interlocutor is equally poor, sometimes eliciting contradictory and absurd statements.”
For example, when a chatbot, Gato incorrectly said that Marseille is the capital of France.
Additionally, a caption created by Gato to accompany a photo read “man holding banana for photo shoot”, even though the man was not holding bread.
DeepMind details Gato in a new research paper, titled “A Generalist Agent,” which was posted on the Arxiv Preprint Server.
The company’s authors said such an agent would show a “significant improvement in performance” when scaled up.
AGI has already been identified as a future threat that could wipe out humanity deliberately or by accident.
Pictured is a dialogue with Gato when he is asked to be a chatbot. A reviewer called Gato’s ability to chat with a human ‘poor’
Earlier this week, UK firm DeepMind unveiled Gato, a program that can chat, caption images, stack blocks with a real robot arm and even play the 1980s Atari video game console. Here are some -some of the tasks Gato was tested on in a DeepMind promo
Dr Stuart Armstrong of the Future of Humanity Institute at Oxford University has previously said that AGI will eventually make humans redundant and eliminate us.
He believes machines will run at speeds inconceivable to the human brain and avoid communicating with humans to take control of the economy and financial markets, transportation, healthcare, and more.
Dr Armstrong said a simple instruction to an AGI to “prevent human suffering” could be interpreted by a supercomputer as “killing all humans”, because human language is easily misinterpreted.
Before his death, Professor Stephen Hawking told the BBC: “The development of full artificial intelligence could spell the end of the human race.”
During his lifetime, the famous British astrophysicist, Professor Stephen Hawking (pictured), said AI ‘could mean the end of the human race’.
In a 2016 paper, DeepMind researchers acknowledged the need for a “big red button” to prevent a machine from performing “a sequence of harmful actions”.
DeepMind, which was founded in London in 2010 before being acquired by Google in 2014, is known for creating an AI program that beat human professional Go player Lee Sedolthe world champion, in a five-game match in 2016.
In 2020, the company announced that it had solved a 50-year-old problem in biology known as the “protein folding problem” – how a protein’s amino acid sequence dictates its 3D structure. .
DeepMind claims to have solved the problem with 92% accuracy by training a neural network with 170,000 known protein sequences and their different structures.
The company is perhaps best known for its AlphaGo AI program which defeated human professional go player Lee Sedol, the world champion, in a five-game match. Pictured is World Go Champion Lee Sedol of South Korea seen ahead of the first round of the Google DeepMind Challenge Match against Google’s AlphaGo program in March 2016
#Googles #DeepMind #verge #reaching #humanlevel #artificial #intelligence