Book a call

News · February 2, 2022

DeepMind's AlphaCode reaches median level in programming contests

// ai.news :: 2022

DeepMind built a system that writes complete solutions to competitive programming problems and scores around the middle of human participants. It is a research result, but it shows how fast code generation is moving.

What happened

  • DeepMind said on February 2, 2022 that it evaluated AlphaCode by simulating participation in 10 recent Codeforces contests that were newer than its training data.
  • According to DeepMind, AlphaCode's estimated rank placed it within the top 54% of participants, roughly the level of a median competitor.
  • DeepMind released its competitive programming dataset of problems, solutions and tests on GitHub.
  • TechCrunch reported that the model was trained on selected GitHub code plus a collection of coding problems and solutions, compared it with OpenAI's Codex behind GitHub Copilot, and said it was not yet a threat to software engineers' jobs.

Why it matters

Contest problems are short, well specified and come with tests, which is exactly where code generation is strongest. Real product work is the opposite: vague requirements, legacy code and trade-offs. So AlphaCode does not change hiring plans today.

What it does change is the direction of travel. Teams that already write clear specs and good tests will get the most from AI coding tools as they mature. That is worth investing in now, whether the code is written in-house, by contractors, or with a machine's help.

Sources

  • DeepMind: Competitive programming with AlphaCode
  • TechCrunch: DeepMind's AlphaCode AI writes code at a competitive level

This post is CMM’s own summary and commentary on publicly reported news. Facts are attributed to the sources linked above; please read them for full details. CMM is not affiliated with these publishers or with the companies mentioned, and trademarks belong to their owners. Prepared with AI assistance. Corrections or requests: [email protected]. See our editorial policy.

Want help with this?

Tell us what you need, and we will suggest the right team model on a short call.

Talk to us