AlphaCode, An Artificial Intelligence Machine That Can Write Programs Better Than Humans

JAKARTA - The artificial intelligence (AI) company DeepMind, has recently developed a new coding engine capable of outperforming many human developers, called AlphaCode.

AlphaCode is able to write computer programs at a "competitive level", ranking among the top 54% of humans in competitive settings.

The performance of the AI model is assessed using ten challenges hosted on Codeforces, a platform that attracts developers looking to flex their coding muscles by competing with others to solve specific problems.

While these challenges are not typical of the average day-to-day workload, the ability to solve the problems they encounter in a creative way is an excellent indicator of programming ability. AlphaCode is the first AI system capable of competing with humans in this context.

DeepMind Alpha Alpha Code

Acquired by Google in 2014, DeepMind has long been at the forefront of artificial intelligence development. Most notably, the company developed AlphaGo, which became the first computer program to beat a human player in the ancient Chinese game Go.

DeepMind's R&D activity extends to many areas where AI can be feasible, but the company appears to have found a new, fruitful opportunity with AlphaCode.

The model was pre-trained on a large set of open-source code taken from GitHub and refined using a smaller data set collected from programming competitions, a move described as "critical to performance".

While many humans outperformed AlphaCode in testing, its ability to create new solutions to complex problems highlights the potential for AI to be integrated into the development landscape in the years to come.

“Solving competitive programming problems is a very difficult thing to do, it requires good coding skills and creative problem-solving in humans,” explains Petr Mitrichev, a competitive programmer from Google, as quoted by Techradar.

"I'm very impressed that AlphaCode can make progress in this area, and excited to see how the model uses its statement understanding to generate code and guide its random exploration to create solutions," he added.

What's more, DeepMind says its research into the area is still in its early stages and therefore the level of performance on offer with AlphaCode is just the beginning.

“Our explorations into code generation leave ample room for improvement and hint at more interesting ideas that can help programmers increase their productivity and open up the field for people who aren't currently writing code,” wrote DeepMind. "We will continue this exploration, and hope that further research will yield tools to improve programming and bring us closer to problem-solving AI."