DeepMind develops AI programmers, which are said to reach intermediate levels in humans

Following the development of the Go robot AlphaGo and the artificial intelligence predictive protein folding system AlphaFold, DeepMind, an AI company owned by Google, has announced the creation of a blockbuster programming robot system, AlphaCode, and the level of programming has reached the average level of human programmers.

DeepMind says AlphaCode's ability to write computer programs has reached a very competitive level. In a race against human programmers, AlphaCode can rank at a moderate level, ranking among the top 54 percent, marking an important step toward autonomous coding for humans.

Oriole Vinyals, principal research scientist at DeepMind, said the study is still in its early stages, but preliminary results show the possibility of creating an autonomous system that solves the challenges of coding programs. "In this area, only humans can do it at the moment." Vinyals said, "In the long run, we are excited about AlphaCode's potential to help programmers and non-programmers write code, increase productivity, or create new ways to make software." ”

AlphaCode still can't beat the best human programmers at the moment. Critics also warn that machine-generated code must be examined with particular care for errors or unforeseen effects, including vulnerability to cyberattacks.

But DeepMind ultimately hopes that this AI tool can solve open-ended problems that require critical thinking, logic, and language understanding. Vinyals said that in the short term, AlphaCode can become an assistant to programmers, while in the long run the system will be able to write its own software.

Mike Mirzayanov, founder of Codeforces, a site that challenges AlphaCode, said in a joint statement with DeepMind: "I can say with certainty that AlphaCode's results exceeded my expectations and I think AlphaCode has reached the level of a promising new competitor. But implementing an algorithm is not the most difficult, inventing an algorithm is. ”

AlphaCode isn't the first AI tool to generate computer code, and many tech companies are developing similar applications. Last June, Microsoft and AI Labs OpenAI adapted language generators into automated programs capable of completing code strings. To end users, these systems are like Gmail's smart writing capabilities, doing whatever people are writing.

"While human programmers may still have an advantage today, it's only a matter of time before AI will surpass humans in many areas." Xiao Jianxiong, founder and chairman of auto-driving company AutoX, told the first financial reporter.

AI is considered the most effective tool to help humans understand the world around them, and it is also a must for tech giants. DeepMind was acquired by Google in 2014 for $600 million, Microsoft funded OpenAI in $1 billion in 2019, and Meta also has a strong AI Research division that leads the world's most advanced AI research and development efforts.

Demis Hassabis, co-founder and CEO of DeepMind, has said: "DeepMind's ultimate vision has always been to build universal AI to maximize the pace of scientific discovery and research. ”

But it is still difficult to make a profit from AI. DeepMind made its first profit in 2020 after losing £500 million in 2019, with a profit of around £44 million, but the source of the profit was not disclosed. The company's AlphaFold algorithm for predicting protein folding last year made its way into drug discovery.

A founder of a domestic artificial intelligence software company told the first financial reporter that DeepMind's revenue source is limited and the investment is huge. DeepMind does not sell products directly to consumers and has not announced any deals with private companies other than parent company Alphabet. But it does sell software and services to Alphabet's Google, YouTube and Moons landing X.

But most experts believe that the tasks that the human brain can achieve are theoretically possible by a physical system. "It's just a matter of time and money." Computer expert Herbert Jaegar told First Financial Reporter, "See who is willing to pay huge sums of money to develop these technologies." ”

DeepMind currently has about 1,000 employees. High salaries are an important means of attracting talent from DeepMind, which employs scientists from the world's leading ai research fields, most of whom have diplomas from Oxford, Cambridge, Stanford, and MIT, and whose annual salaries typically exceed a million dollars.