Google Deepmind says he has made a breakthrough of “historic” artificial intelligence like computer in deep blue beating Garry Kasparov in chess in 1997 and One AI Beat a human Go champion in 2016.
A version of the model of IA Gemini 2.5 of the company has solved a complex problem of the real world which disrupted human computer programmers to become the first model of AI to win a gold medal during an international programming competition held earlier this month in Azerbaijan.
In a performance that technological society called a “deep jump in the solving abstract problem”, it took less than half an hour to determine how to weigh an infinite number of possibilities in order to send a liquid via a network of conduits to a set of interconnected tanks. The goal was to distribute it as quickly as possible.
None of the human teams, including the best interpreters of universities in Russia, China and Japan, has done things.
The AI failed two of the 12 tasks it has been set, but its overall performances classified it in second place out of 139 of IT programmers of the strongest college in the world. Google said it was a “historic moment, towards AG (general artificial intelligence)”, which is widely considered as a human level intelligence to a wide range of tasks.
“For me, this is a moment that is equivalent to a deep blue for failures and alphago to go,” said Quoc le, vice-president of Google Deepmind.
“Even greater, it is more of the real world, not just a forced environment (as Chess And come on) … because of this, I think this advance has the potential to transform many scientific and engineers. He quoted the conception of drugs and chips.
The model is an AI for general use, but has been specially formed to solve very hard coding, mathematics and reasoning problems. He carried out “as a coder of the best 20 in the world,” said Google.
“The resolution of complex tasks to these competitions requires deep abstract reasoning, creativity, the ability to synthesize new solutions to the problems never seen before and a real spark of ingenuity,” said society.
Speaking before the details were made public, Stuart Russell, professor of computer science at the University of California in Berkeley, said that “the vintage statements seem exaggerated”.
He said that the AI systems have been going on well on programming tasks for some time and that the breakthrough of the failures of deep blue had “essentially no impact on the real world of applied AI”.
However, he said that “to obtain an ICPC (International Collegiate Programming Contest) question, the code must actually work correctly (at least on a finite number of test cases), so this performance can show progress towards the manufacture of coding systems based on AI sufficiently precise to produce high quality code”.
He added: “The pressure on AI companies to continue to claim breakthroughs is enormous.”
Michael Wooldridge, Professor Ashall of the foundations of artificial intelligence at the University of Oxford, said that it looked like an impressive achievement and that “being able to solve problems at this level was exciting”.
But he asked how necessary the computing power was. Google refused to say, in addition to confirming that it was more than available for an average subscriber to its Google AI Ultra service of $ 250 per month using the light version of Gemini 2.5 Deep Think in the Gemini application.
Dr. Bill Poucher, executive director of the ICPC, said: “Gemini to join this arena and to obtain results in terms of gold, marks a key moment in the definition of AI tools and the academic standards necessary for the next generation.”
After promoting the newsletter
Four breakthroughs of Machine Intelligence
1957 The Percetron
Frank Rosenblatt, an academic at Cornell University, has determined that it should be possible to create an “perception and recognition automaton”. He named the pierceptron and said An electronic system would be able to learn to recognize models of optical, electric or tonal information “in a way that can be closely similar to the perceptual process of a biological brain”.
The following year, he built the device, which was the size of a small room. He was considered one of the first breakthroughs of artificial intelligence based on neural networks.
1997 Big Blue
In May 1997, Big Blue of IBM became the first computer system to defeat a reigning world chess champion in a match under the controls of the standard tournament. He beat Garry Kasparov in what has become an inflection point in calculation power, but the competition was close.
Kasparov won the first game, Deep Blue the second followed by three draws. Deep Blue won match 6 to guarantee victory. He showed how the computation power of the brute force could create a system to overcome a human, although a narrow task. “The computer is much stronger than anyone who did not expect,” said Kasparov, conceding defeat.
2016 AlphaGo
Go is one of the most complex games ever designed, and one of the world’s players in the world was Lee Sedol, a South Korean professional. In 2016, Deepmind, the British company of AI created by Demis Hassabis, took it back with its Alphago computer.
He won 4-1 and some of his movements seemed to display a truly original thought. Movement 37 in particular fell into traditions. Hassabis said: “It could be the first glimpse of a brilliant and daring future where humanity exploits AI as a new powerful tool, helping us discover new knowledge that can solve some of our most pressing scientific problems.”
2020 Alphafold
Another breakthrough of Hassabis and Deepmind was an AI program which can predict how proteins fall back into 3D forms, a very complex fundamental process to understand the biological machines of life. The Royal Society, the 360 -year -old London scientific institution, called It is “an amazing advance”.
When researchers know how a protein folds back, they can start discovering mysteries such as the way insulin controls blood sugar levels or how antibodies fight viruses. After other iterations, the system helped Hassabis and his colleague John Jumper Share a Nobel Prize For chemistry in 2024.