I’ve been using home coding assignments to assess candidates for SW and ML positions. I use custom assignments that mainly test candidate ability to come up with solutions and and write good software to implement them. As LLMs are everywhere it becomes very easy to come up with ideas, write code, improve code, etc. Personally I see it as a powerful tool to supercharge develop abilities. But as for skills assessment it is unclear to me how to cope with it. Ban? Allow?<p>WDYT?
I would allow it.<p>In a year or two everyone will use language models in their life. Also, GPUs become faster, there's a huge investment here, and improvements will keep coming for years. Also, we see that the larger the model, the smarter it is and we didn't get to the plateau. So, expect these model to take over more and more work.<p>At some point I believe knowledge of wrangling LLMs will be more important at some point than code.
> Personally I see it as a powerful tool to supercharge develop abilities<p>There's your answer: allow.<p>It's not going away, everyone will be using it soon, so get those early adopters IMO.<p>We're not even in the LLM era yet, kinda like we weren't in the smartphone era in 2007 when iPhone launched. Models (of all kinds) will be ubiquitous, probably more quickly than smartphones, because the value add is so huge.