- DeepMind expert suggests the hardest tasks to create a human-like AI are solved
- The London firm wants to build an 'AGI' that has the same intelligence as humans
- This week DeepMind unveiled a program capable of achieving over 600 tasks
Nando de Freitas, a research scientist at DeepMind and machine learning professor at Oxford University, has said 'the game is over' in regards to solving the hardest challenges in the race to achieve artificial general intelligence (AGI).
AGI refers to a machine or program that has the ability to understand or learn any intellectual task that a human being can, and do so without training.
According to De Freitas, the quest for scientists is now scaling up AI programs, such as with more data and computing power, to create an AGI.
Earlier this week, DeepMind unveiled a new AI 'agent' called Gato that can complete 604 different tasks 'across a wide range of environments'. Gato uses a single neural network - a computing system with interconnected nodes that works like nerve cells in the human brain. It can chat, caption images, stack blocks with a real robot arm and even play the 1980s home video game console Atari, DeepMind claims.
The Next Web that said humans alive today won't ever achieve AGI.
De Freitas tweeted:
'It's all about scale now! The Game is Over! It's about making these models bigger, safer, compute efficient, faster...'However, he admitted that humanity is still far from creating an AI that can pass the Turing test - a test of a machine's ability to exhibit intelligent behaviour equivalent to or indistinguishable from that of a human.
After DeepMind's announcement of Gato, The Next Web article said it demonstrates AGI no more than virtual assistants such as Amazon's Alexa and Apple's Siri, which are already on the market and in people's homes.
'Gato's ability to perform multiple tasks is more like a video game console that can store 600 different games, than it's like a game you can play 600 different ways,' said The Next Web contributor Tristan Greene.Gato has been built to achieve a variety of hundreds of tasks, but this ability may compromise the quality of each task, according to other commentators.
'It's not a general AI, it's a bunch of pre-trained, narrow models bundled neatly.'
ZDNet columnist Tiernan Ray wrote that the agent 'is actually not so great on several tasks'.
'On the one hand, the program is able to do better than a dedicated machine learning program at controlling a robotic Sawyer arm that stacks blocks,' Ray said.For example, when a chatbot, Gato initially mistakenly said that Marseille is the capital of France.
'On the other hand, it produces captions for images that in many cases are quite poor.
'Its ability at standard chat dialogue with a human interlocutor is similarly mediocre, sometimes eliciting contradictory and nonsensical utterances.'
Also, a caption created by Gato to accompany a photo read 'man holding up a banana to take a picture of it', even though the man wasn't holding bread.
DeepMind details Gato in a new research paper, entitled 'A Generalist Agent,' that's been posted on the Arxiv preprint server.
The company's authors have said such an agent will show 'significant performance improvement' when it's scaled-up.
AGI has been already identified as a future threat that could wipe out humanity either deliberately or by accident.
He believes machines will work at speeds inconceivable to the human brain and will skip communicating with humans to take control of the economy and financial markets, transport, healthcare and more.
Dr Armstrong said a simple instruction to an AGI to 'prevent human suffering' could be interpreted by a super computer as 'kill all humans', due to human language being 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.'
DeepMind, which was founded in London in 2010 before being acquired by Google in 2014, is known for creating an AI program that beat a human professional Go player Lee Sedol, the world champion, in a five-game match in 2016.
In 2020, the firm announced it had solved a 50-year-old problem in biology, known as the 'protein folding problem' - knowing how a protein's amino acid sequence dictates its 3D structure.
DeepMind claimed to have solved the problem with 92 per cent accuracy by training a neural network with 170,000 known protein sequences and their different structures.
What isGoogle's DeepMind AI project?
DeepMind was founded in London in 2010 and was acquired by Google in 2014.
It now has additional research centres in Edmonton and Montreal, Canada, and a DeepMind Applied team in Mountain View, California.
DeepMind is on a mission to push the boundaries of AI, developing programs that can learn to solve any complex problem without needing to be taught how.
If successful, the firm believes this will be one of the most important and widely beneficial scientific advances ever made.
The company has hit the headlines for a number of its creations, including software it created a that taught itself how to play and win at 49 completely different Atari titles, with just raw pixels as input.
In a world first, its AlphaGo program took on the world's best player at G, one of the most complex and intuitive games ever devised, with more positions than there are atoms in the universe - and won.