However, with ChatGPT, Copilot, and other LLMs, it becomes much easier to cheat on these interviews. No longer does a potential cheater have to find a ghost-coder to provide answers live - they just need to type in answers into a LLM and wait for it to generate some plausible code. While LLMs will probably not let someone pass who can't code at all (unless they just use straight LeetCode, in which case those interviewers get what they deserve), it will make it much easier to come up with a first draft and iterate, and therefore make the interview much easier, even for experienced interviewees.
I wonder if, as a result, companies will start requiring in-person interviews for the final interview round again. I've already heard suggestions that college classes that want to test the essay-writing abilities of their students move back to bluebooks in a proctored exam hall, rather than allowing students to take tests remotely. To be clear, I haven't heard of tech companies moving away from pure-remote interviews, but perhaps with the hiring rate slowing down, the vulnerability of remote interviews to cheating via LLMs will cause them to be deprecated.
If employers really wanted to hire people who can actually write software interviews would be a conversation stepping through a process to solve a complex problem. Examples of complex problem spaces: Performance, Accessibility, Security, Transmission Control.
From the perspective of the candidate the quality of questions and candidate filtering indicate whether the employer is just trying to put bodies in seats, jockey configurations, or advanced copy/paste. Writing original software isn’t that hard, but it does take practice most employers won’t provide.
Bring back pure whiteboard and I will be almost as good at this stuff as I was when I was 22. Little slower on the recall, but still good.
If colleges are unable to detect AI-written essays then they have bigger problems. LLMs have no concept of anything other than a probability of what words and phrases go together (which is why it can be so confidently incorrect).
To be honest, it was inevitable.