Google AI team DeepMind says new coding bot rivals humans

Alphabet-owned AI company DeepMind claims to have created an AI capable of writing programming code, finding new solutions to interesting problems, and doing so at the level of mid-ranked human participants in competitions. coding.

Dubbed “AlphaCode” and detailed in a pre-print article [PDF]the tool is said to advance over previous automated coding efforts by displaying the ability to tackle “problems that require a combination of critical thinking, logic, algorithms, coding, and natural language understanding”.

Previous efforts to create code that codes have not been able to achieve this level of sophistication, but have succeeded decently when asked to handle simple mathematical or programming tasks.

The DeepMind article explains that the AlphaCode developers addressed three important things:

  • Finding a clean dataset to work with, and since coding contests aren’t hard to come by, the data was easily secured.
  • Design a suitable algorithm, modeled on transformer-based architectures used for natural language processing or image recognition.
  • Create lots of sample solutions, then filter them to find work that would be useful for the problem at hand.

This problem was to participate in coding contests organized by Codeforces.

One of the challenges faced by AlphaCode is described as follows:

AlphaCode’s solution follows.

Screenshot of AlphaCode output

AlphaCode code and explanation of how it works. Click to enlarge

AlphaCode has done this task quite well. The pre-printed log says he placed in the top 54.3% of contestants he…er…contested.

Yes, that means he beat a lot of humans. Humans who might just code for a living.

The article considers the wider impact on the world of code than code, suggesting that one potential risk is “the advanced risks of AI” as automated coding “could lead to systems that can write themselves and s ‘improve recursively, rapidly leading to more and more advanced systems’.

Before we have to worry about that, the article discusses whether automated coding is bad news for human programmers.

The authors’ conclusion is that developers should be OK because “writing code is only part of the job, and previous instances of partial programming automation (e.g. compilers and IDEs) have only moved programmers to higher levels of abstraction and open the field to more people”.

More information about AlphaCode can be found in this GitHub jar and repository.

Read them carefully, dear reader. They could be competition for your next job. ®

Comments are closed.