A new AI coding challenge just published its first results – and they aren’t pretty

A brand new AI coding problem has revealed its first winner — and set a brand new bar for AI-powered software program engineers. 

On Wednesday at 5pm PST, the nonprofit Laude Institute introduced the primary winner of the Ok Prize, a multi-round AI coding problem launched by Databricks and Perplexity co-founder Andy Konwinski. The winner was a Brazilian immediate engineer named Eduardo Rocha de Andrade, who will obtain $50,000 for the prize. However extra stunning than the win was his closing rating: he gained with appropriate solutions to only 7.5% of the questions on the check.

“We’re glad we constructed a benchmark that’s truly laborious,” stated Konwinski. “Benchmarks needs to be laborious in the event that they’re going to matter,” he continued, including: “Scores could be completely different if the large labs had entered with their largest fashions. However that’s type of the purpose. Ok Prize runs offline with restricted compute, so it favors smaller and open fashions. I really like that. It ranges the enjoying discipline.”

Konwinski has pledged $1 million to the primary open-source mannequin that may rating increased than 90% on the check.

Just like the well-known SWE-Bench system, the Ok Prize exams fashions in opposition to flagged points from GitHub as a check of how effectively fashions can take care of real-world programming issues. However whereas SWE-Bench is predicated on a set set of issues that fashions can practice in opposition to, the Ok Prize is designed as a “contamination-free model of SWE-Bench,” utilizing a timed entry system to protect in opposition to any benchmark-specific coaching. For spherical one, fashions have been due by March twelfth. The Ok Prize organizers then constructed the check utilizing solely GitHub points flagged after that date.

The 7.5% high rating stands in marked distinction to SWE-Bench itself, which at the moment reveals a 75% high rating on its simpler ‘Verified’ check and 34% on its tougher ‘Full’ check. Konwinski nonetheless isn’t certain whether or not the disparity is because of contamination on SWE-Bench or simply the problem of amassing new points from GitHub, however he expects the Ok Prize undertaking to reply the query quickly.

“As we get extra runs of the factor, we’ll have a greater sense,” he informed TechCrunch, “as a result of we anticipate individuals to adapt to the dynamics of competing on this each few months.”

It would look like an odd place to fall quick, given the big selection of AI coding instruments already publicly accessible – however with benchmarks changing into too straightforward, many critics see initiatives just like the Ok Prize as a vital step towards fixing AI’s growing evaluation problem.

“I’m fairly bullish about constructing new exams for current benchmarks,” says Princeton researcher Sayash Kapoor, who put ahead an analogous concept in a recent paper. “With out such experiments, we will’t truly inform if the problem is contamination, and even simply focusing on the SWE-Bench leaderboard with a human within the loop.”

For Konwinski, it’s not only a higher benchmark, however an open problem to the remainder of the trade. “In the event you hearken to the hype, it’s like we needs to be seeing AI docs and AI attorneys and AI software program engineers, and that’s simply not true,” he says. “If we will’t even get greater than 10% on a contamination free SWE-Bench, that’s the fact examine for me.”