Google's new AI is capable of writing its own code

According to a share on Twitter, AlphaCode is capable of solving open problems that require critical thinking, logic, and language understanding.

Through many tests and evaluations, AlphaCode can only perform simple tasks, not enough to meet the complex tasks of today's programmers. However, this AI tool is said to be the premise for future technological breakthroughs.

Previously, Mikhail Mirzayanov, founder of Codeforces, who had raised many doubts about AlphaCode's capabilities, had to acknowledge this new AI technology: 'AlphaCode's capabilities exceeded my expectations. Surely this technology will have a promising future.

Besides, AlphaCode's intelligence also raises some deep concerns. Some argue that when self-coding AI tools like AlphaCode are too smart, they can create malware themselves, even triggering unintended cyber attacks.

Before AlphaCode, Microsoft launched self-coding AI software called GitHub Copilot in June 2021, with the ability to analyze existing code, thereby recommending code and debugging. Not long after that, OpenAI also launched Codex AI with the ability to program itself in 12 languages.

Lesley Montoya

Update 05 February 2022