GitHub Copilot, a personal AI pair programmer is generally available to all developers this Jun. Is this another Clippy, “the World’s Most Hated Virtual Assistant”, or something better? What do you think?
Beginning of the end?
On Jun. 21, 2022, Copilot graduated from beta into a paid service for $10 per month. It’s a good step forward. According to GitHub, more than 1.2M developers used the preview in the past year. In the code it assisted, Copilot contributed nearly 40%. Who will say no to “an extra helping hand” 🦾.
There is no doubt Copilot & a bunch of code autocomplete systems have become much better recently, thanks to advances in Machine Learning. “Confidently”, Amazon also announced CodeWhisperer this Jun. Nowadays, ML-powered coding auto-complete systems can, not only solve “multiple-choice” problems but also do “short-answers”. Don’t worry yet, this is not even the beginning of the end for human coders. Because the machine coder won’t fully replace them anytime soon yet. But it is, perhaps the beginning of the end of passive autocomplete tools, or mechanical coding interview even . I’m feeling lucky because I can’t wait to let JARVIS take over tedious coding tasks from me. So that, I may spend more time on design & development.
To charge or not to charge
GitHub may have many good reasons to charge Copilot. After all, no one wants to work for “free”. And, it is surely clever to ask for Return on Investment. But if I may, my counterarguments are:
- There are many free alternatives & substitutes. Why push users to switch?
- There will be much fewer users paying for the service. As there is still a lot of room to explore new value, why kill the goose for the eggs so early?
- There are still conservancies such as a consensus of fair use, Intellectual Property, etc. to generate code from other code even indirectly. Why create even more controversy by monetizing directly?
A longer game
To be fair, charging Copilot may be an attempt for GitHub to learn something. For example, I’m very interested to know how many, who & why people will pay for what kind of tasks. Nevertheless, there are better ways to eventually win in the longer run.
For example, some kinds of freemium may be a better business model for such services. Why not offer premium capabilities for enterprise & even solution providers. There is not only much less conservancy when the code snippets are from the customers’ own codebase and sample codes, but also more desirable for enterprises to keep codebase coherence across generations of developers. And, solution providers may make it easier for their customers to use & themselves to sustain.
Anyway, coding may be the part easier to be automated. It’s also a part of the development workflow today. From experiences with previous code generation technologies & even frameworks or support libraries, how much effort engineers end up on understanding & debugging them are very important. Those efforts can drag the adoption when they are too heavy. If anyone can reduce them to a certain threshold, it’s generally good when machines can do the heavy lifting to augment humans, instead of something to be worry about.
Full Disclosure
The opinions stated here are my own, not those of my company. They are mostly extrapolations from public information. I don’t have insider knowledge of those companies, nor a whatever expert.
