I suppose we can never know.

"I'll have time next month when things finally calm down"

agree on all that. Some people gonna take shortcuts.

I'm ok with LLMs being used as an additional evaluation tool if clearly labeled as such. Eg a grant funder could do that.

When individuals use it for their 'own' assessment we're in trouble. Then a handful of corporate LLMs decide everything

or by getting something else funded and doing it on the side

Is there nothing wacky and fun and out there left for academic ML researchers to go for?

someone once told me that Watson, unlike Crick, was such an effective fundraiser because he had a deep set belief that rich people *must* be smart, and so treated them with deference, which they loved. Crick otoh did not.

Do you think there's any truth to that?

Watson was a racist who, "near the end of his life, faced condemnation and professional censure for offensive remarks, including saying Black people are less intelligent than white people"

🙈

Go n-éirí an bóthar libh

congrats Shane et al! Sounds like a very cool project, looking forward to seeing it progress

that is a good example of an LLM being used to complement human work for areas that LLMs are as good as/better than humans.

The worry though is that people are starting to use GenAI as a shortcut to do the bits they should be doing themselves

there is a bit of a discussion on here about using LLMs for peer review. But I think the general issue is worth discussing... I give my opinion!

absolutely agree it's ok to use software to assist your work - I definitely do. But imo some tasks (currently) remain better handled in a genAI-less way

100%, I use LLMs nearly every day for those kinds of tasks. It's great for tedious things that you can easily and rigorously verify yourself after.

That and brainstorming some maths questions... 'what is the best method for solving this problem' etc