Dan Ports
I have heard a couple stories lately about faculty trying to replace their PhD students with LLMs to do research, and I regret to inform these people:
1) you are bad people who have misunderstood the concept of graduate students entirely.
2) I guarantee you that your students have already replaced you with LLMs because they provide more effective advising.
May 28 at 9:18pmWeb
Show threadShriram Krishnamurthi1d ago
@dan Hyperbole much?
Show threadDan Ports1d ago
@shriramk a bit, perhaps, but I have actually heard the profoundly bad take of "this is great, we won't need to deal with students anymore"
Show threadPatrick Lam 1d ago
@dan @shriramk That's the whole thing though. I'm reading a PhD thesis (defense next week) and part of it is that I provide comments to the student and hopefully they learn something. I would not want to do that for LLM text. Basically I kind of never want to read LLM text.
Show threadShriram Krishnamurthi1d ago
@va2lam @dan Aren't you reading Dan backwards?
Show threadPatrick Lam 1d ago
@shriramk @dan I can see the appeal of never having to read another PhD thesis. But educating PhD students is a key part of the job and results in more researchers in the future. I think that is consistent with Dan?
Show threadShriram Krishnamurthi14h ago

@va2lam @dan
I am 100% behind the idea that if you're a PhD advisor, or sit on a committee, you should be willing to read the dissertation yourself — or else don't serve in those roles.

They already know where to get an LLM's judgment. They asked for *yours*.

Show threadrk: it’s hyphen-minus actually1d ago
@dan @shriramk
Show threadShriram Krishnamurthi1d ago
@dan Sure, yes, that is bad in some ways.
Show threadjon crowcroft20h ago
@shriramk @dan post training alignment will make sure the gradLLM still hallucinates, while the advisorLLM still digresses. Progress will be maintanined although tuition fees to cover electricty costs will eventually bankrupt the USA. again.
Show threadDr. Bruce Simpson1d ago
@shriramk @dan I was trying to summarise some stuff earlier and wound up with an 86 point list of reasons why I will not use "AI": https://people.freebsd.org/~bms/no-thankyou-genai.txt
Show threadShriram Krishnamurthi1d ago
@bms48 @dan
Just 9 more to go!
Show threadThommy1d ago
@dan The great irony is those who are the most excited about replacing other humans with software are themselves the most actually replaceable by software.