So Google has gone into the automation process for code development, and not surprising given how popular GitHubs Copilot is. The new system was open to beta testers and sign ups on the 17 of May, so this is a pretty new product. What’s cool is that if you’re a Google Workspace user, this might be a tool that comes to your subscription.
There are four main ideas behind this system:
Duet AI helps Google Cloud be more accessible and personal for all types of users, regardless of their skill level, by providing them with support whenever they need it. Developers can get code recommendations from Duet AI, data engineers can get insights from prompts, and business users can create apps via chat.
Generate a holistic picture in any format you want, wherever you are in Google Cloud, with generative AI at its center, providing a holistic picture across functions, services, and tech stacks.
Using Codey, a foundation model that powers Duet AI, we pre-trained it with Google Cloud-specific content, such as documentation and sample code, and fine-tuned it based on the patterns and behaviors of Google Cloud users to deliver smarter, contextual recommendations.
Google’s AI Principles include responsibly developing technology. Model learning and development will not be performed on users’ code or recommendations. Data and code privacy are both protected, as is the integrity of the knowledge space from which AI models are derived.
So, for me this is interesting as it is going right into one of the more interesting markets. And as an instructor, using something like Duet or Copilot is going to make meaningless many student coding exercises in college. How do we test basic coding proficiency and understanding of what the code does if we can just have an AI do it.
At least with Github we could go and read the notes, or have it flag in a plagiarism checker.
Then the flip side of this there are only so many basic ways of coding a solution, code reuse is an accepted standard and expected along the way.
So, is AI generated code just code reuse?
It’s going to be interesting