Human wins coding contest … but AI finishes a close second

“Humanity has prevailed (for now!),” so said Polish programmer Przemysław Dębiak after emerging victorious in a coding contest whose competitors included an advanced AI model created by AI giant OpenAI.

Dębiak, who happens to be a former employee of OpenAI, was showing off his programming skills at the AtCoder World Tour Finals held in Tokyo last week. 

The 10-hour session left the Pole, who competed under the name Psyho, feeling “completely exhausted” and “barely alive,” according to a post on X.

As for the AI model, well, presumably it felt just fine and could’ve carried on without complaint. 

It’s believed to be the first coding contest to have allowed the participation of an AI model, with the technology gaining prominence for its rapidly improving programming smarts.

OpenAI’s AI model participated as part of a special “Humans vs AI” exhibition match conducted within the Heuristic division of the contest. Called OpenAIAHC, the AI model took on 12 top-ranking human programmers — with Dębiak among them — on a 10-hour optimization challenge.

The Polish programmer managed to finish in first place by a margin of just over 9%, with the AI placing second — ahead of all of the other human contestants.

Interestingly, Dębiak said in comments after the contest that it was the AI model’s presence that motivated him to keep going, as he could see during the match-up that he was just ahead of the model, and he very much wanted to keep it that way. Without the AI’s participation, his score would have been “much, much lower,” Dębiak told Business Insider. 

The AI’s impressive performance is a clear demonstration that advanced AI is making real progress when it comes to edging out the world’s best human programmers in open-ended, creative problem-solving tasks. Indeed, it’s hard to think that it won’t come out on top at next year’s contest … if it’s invited back, that is.