GitHub just rolled out an interesting new feature that helps programmers (read, data scientists) write code more efficiently.
In this week’s Tech Tuesday, I’ll explain what this new feature does, how it’s being used, and how this type of technology might lead data scientists of the future.
GitHub is a platform that holds the code written by many programmers. It stores revisions of the code we all write each day and allows data scientists to roll back code to a previous version, if that sort of thing is necessary. It’s invaluable to teams of developers and used everywhere I’ve ever worked (If it’s not GitHub, then it’s some other version control platform, but Microsoft’s GitHub is by far the most popular).
GitHub recently rolled out an A.I.-driven tool to help programmers be more efficient while completing the code they spend all day writing. This new tool is called Copilot. Here’s what it does, in the words of GitHub’s CEO:
[Copilot is] a new A.I. pair programmer that helps you write better code. It helps you quickly discover alternative ways to solve problems, write tests, and explore new APIs without having to tediously tailor a search for answers on the internet. As you type, it adapts to the way you write code – to help you complete your work faster.
This is a big deal, as programmers routinely spend significant portions of their days searching for solutions to problems or ways to solve problems more efficiently.
One popular website, Stack Overflow, hosts the answers to what must be millions of questions written by programmers. Data scientists routinely spend hours poring over this site, looking for the optimal way to solve a problem.
Now, GitHub claims that Copilot can do that for us. Instantly.
It works by identifying coding patterns across billions of lines stored in GitHub’s public code repositories. After all, they have the biggest collection of code in the world.
How Much Will Copilot Help?
I predict that Copilot will work best for problems that have known solutions, such as how to efficiently sort a list of numbers, or how to compute a rolling average over an array of numbers. It’ll be a huge time saver when we need boilerplate or repetitive code patterns.
Where it will break down, of course, is in the novel solution to specific problems. For example, what if the data you’re computing the rolling average over has many spurious values? Is the average still the right thing to do?
In other words, we’ll still need senior level oversight to ensure that the solution matches the problem.
As Copilot matures, I can imagine it saving coders appreciable time that would otherwise be lost to finding trivial solutions or copying and pasting the same snippets over and over.
But a reality in which we can talk into a device about the problems we want solved and where an A.I. creates it for us from scratch – in a manner in which it works and generalizes well to all sorts of inputs – is still far, far away, and will likely require human guidance (at a minimum) for a long time to come.
Where some may see the end of humans in the loop for completing data science tasks, I see a long road ahead. Data are simply too varied and the breadth of problems too great for a single tool to force data scientists into obsolescence.
Of course, this is not what Copilot claims. Their claims are that they can increase efficiency and help get products delivered sooner. From what I have seen and heard about this tool, I believe it will save companies a lot of time otherwise spent doing the mundane.
All the best,
Anthony Bourdain Documentary Sparks Backlash
A new documentary about Anthony Bourdain has sparked a debate over its use of artificial intelligence to stitch together voiced quotes by the late celebrity chef – effectively bringing his voice back to life. Is it ethical to use A.I. to create deep fakes to speak on behalf of the dead? I’d love to hear your thoughts. No doubt we’re in for more controversy like this in the future.
As the Use of A.I. Spreads, Congress Looks to Rein It In
The White House, lawmakers from both parties, and federal agencies are all working on bills or projects to constrain potential downsides of A.I.. At the same time, parts of the US government are working to place limits on algorithms to prevent discrimination, injustice, or waste. Where will all this lead, as A.I. is now regularly used for face recognition and forensic algorithms for face, fingerprint, and DNA analysis? Read more here.
WHO Issues First Global Report on A.I. in Health
Artificial Intelligence holds great promise for improving the delivery of healthcare and medicine worldwide, but only if ethics and human rights are put at the heart of its design, deployment, and use. This is tricky territory as we rely more and more on A.I. in health. Here’s the World Health Organization’s newly published guide on this, listing six guiding principles for its design and use.