It's probably impossible to detect ALL languages without training for them specifically, but there's good generalization happening. Our model is a unified model rather than a separate model per language. We started out with language-specific models but found that the unified approach yielded slightly better results in addition to being more efficient to train.

Candidly, it's a bit of a black box still. We hope to do some ablation studies soon, but we tried to have a variety of formatting and commenting styles represented in both training and evaluation.

I'm sure you can but there isn't really an adversarial motive for doing that, I would think :)

Sure there is.

More details about how we eval'ed here:

https://www.span.app/introducing-span-detect-1

Recall 91.5, F1 93.3

I think you need to define which one is the positive, and which one is the negative?

Is AI generated code the positive?

I'd argue that knowing AI generated code shipped into production is the first step to understanding the impact of AI coding assistants on velocity and quality. When paired with additional context, it can help leaders understand how to improve proficiency around these tools.

That's not relevant to the comment you replied to.

That’s because many of the patients refuse to wear masks as well.

It's hard to wear a mask when you have a tube down your throat and/or have trouble breathing.

Cheap acquisitions

“Seattle Is Dying” is a great investigative report about the problems and potential solutions to homelessness. Also applicable to SF.

https://youtu.be/bpAi70WWBlw

I've seen this video and it's a Sinclair hit piece. Not worth the time of anyone who wants to engage in an informed discussion.

It's 2019 and this propaganda piece (yes, it's propaganda) is arguing mass incarceration is the solution.