Generative AI to Boost Productivity?

Justin Hutchings, GitHub

 

Greetings from Microsoft's Canberra office where I am attending a glitzy GitHub presentation. The bold claim made was that Copilot, Microsoft's implementation of ChatGPT, could end the computer productivity paradox. 

The paradox is that computers have not increased office productivity. We were invited to photograph the slide of claimed productivity increases. However, I can recall claims made decades ago for 4GL languages, not far from where we are, which would allow programs to write themselves. In practice they allowed amateurs to get into a mess. The current attempt is not like that, with tools for professionals. However, these tools only do the easy bits.

Some of the features demonstrated could be useful for teaching computing. In particular implementing rules about what code libraries can be used and what coding practices must be followed. The billing features might also be used for student hurdles, the idea being that to pass the course the student must put in a set amount of work (the quality of the work would determine their grade).

ps: At question time someone asked about legacy code. I guess we need Copilot COBOL. ;-)

pps: It is a long time since I have been in a corporate environment, and find sales pitches excruciating. One yesterday from a plagiarism detector company was particularly bad. But this Microsoft event today is okay.