TL;DR: Wait until machine learning solutions solve real business problems.

Alphabet's DeepMind, Microsoft's Github, and OpenAI are leading the Artificial Intelligence coding segment.

With different approaches, they try to solve coding challenges.

They use [transformer-based](https://en.wikipedia.org/wiki/Transformer_(machine_learning_model) language models to generate code.

AlphaCode achieved an estimated rank within the top 54% of participants in programming competitions by solving new problems that require a combination of critical thinking, logic, algorithms, coding, and natural language understanding.

Deep Mind's product was trained and competed against humans at Codeforces contests.

With overfitting claims, it outperformed several people at those contests.

Results are available online.

A careful look at the datasets shows basic algorithmic solutions with output code full of code smells many human counterparts make.

Variables with bad names, Lots of ifs, too long methods, global functions are some examples of these smells many code reviewers would not approve.

The three current solutions (AlphaCode, GitHub Copilot, GPT3) focus on competitive programming, an area where performance is the ultimate goal and readability is not important.

The final frontier

Model domain entities and business behavior instead of boring and repetitive algorithms full of code smells.

Try to understand the entities and the behaviour on your simulation software.

References

Why GitHub Copilot is not a Threat to your Job

I've Recently Learned About GPT3 - This is my Journey

What should you do?

Stop competing against smarter machines than you.

Learn and model rich domains.

You will keep your job for a few more years...