DeepMind’s AI Coder Won’t Replace Humans Yet

… but that doesn’t mean it doesn’t want to

Key Takeaways

  • DeepMind’s AI coding engine is as good as an average human programmer.
  • The AlphaCode engine comes up with creative solutions to coding problems.
  • AI might be best when it augments human labor instead of replacing it.
The words "Artificial Intelligence" written using a typewriter

Markus Winkler / Unsplash

Research company DeepMind says that AI coding engines can write programs as well as a human. Are robots finally coming for software developers' jobs?

When DeepMind put its AlphaCode engine to work on coding challenges designed to test humans, it finished in the top 54 percent, making it as good as an average human. That might sound like it's ready to be deployed for live use. You could fire the worst half of your human coders, then replace them with AI coding bots, right? Not yet.

"With AI companies, writers are needed more than ever. The real benefit of AI writers is that they provide research and tools that speed [up] the process of what needs to go into the content. I imagine that AI coding engines will do the same for programmers. It will make them more efficient, making it easier to get started with developing a structure for their applications, and speed [up] the process of coding," John Cass, co-founder of AI company AIContentGen, told Lifewire via email. 

Support, Not Supplant

The promise of AI is that it can replace humans in menial tasks or supplant humans in expensive jobs. But in practice, we're not there yet. If you've ever used AI apps to edit your photos, for example, you'll know there's still plenty of cleanup to do after the tool is finished. At the very least, the human is reduced to clicking a button to cycle through AI-created options, then picking the best. 

In the case of DeepMinds' AlphaCode engine, its AI is trained to tackle coding challenges. Examples provided on the AlphaCode project page are finding optimum ways to arrange roads and buildings or coming up with strategies to win board games. These might not be useful in the workplace, but DeepMind's AI showed one important trait: Creativity. 

"I can safely say the results of AlphaCode exceeded my expectations," Mike Mirzayanov, founder of Codeforces, a site that organizes coding competitions, said in the Deep Mind blog. "I was skeptical because even in simple competitive problems, it is often required not only to implement the algorithm but also–and this is the most difficult part–to invent it."

Graph showing how DeepMind's AI compares to human programmers


The most likely scenario, to begin with, at least, is for human coders to use AI tools to help them work. And other companies, Microsoft for instance, are working on AI tools to help programmers work faster by doing a lot of the busywork for them.

In a way, we're all used to using AI tools every day, and we know the pitfalls and frustrations they bring. Autocorrect, for example, is supposed to make typing faster on little on-screen keyboards, but in practice, you end up changing your typing style to better trigger the autocorrect suggestions. 

So, will human coders really be replaced by AI? Unlikely.

"Coders will still be in the driver's seat, as writers are with AI content writers," says Cass. "In a way, the new AI writing tools mean even more job security for writers because they will have the expertise on how to use and get the best out of the more sophisticated tools for the foreseeable future."

Art Official Intelligence 

There are a few ways to view AI in creative pursuits. One is that it removes the grunt-work and lets the human focus more on the creative aspects. The human becomes more of a film director instead of the screenwriter of the actor. We can take a step back and view the whole project from a higher level, unconcerned with the nitty-gritty details needed to achieve our visions. 

"It will make them more efficient, making it easier to get started with developing a structure for their applications..."

On the other end, AI creativity is still algorithmic creativity. It will invent solutions, write novels, or filter our photographs, but perhaps not in a way that resonates with other humans in the way art can. 

Between these extremes are artists like Brian Eno, who lets home-grown AI-created music run in the background while he’s in the studio. When something catches his ear, he saves it for later use.

AI creations can inspire humans in directions we might not normally go. Or AI can dictate how we work, so we end up as menial babysitters for the machines. Like any tool, then, it’s how we use it that counts.

Was this page helpful?