reid-lab.org
Canadian 🇨🇦 Assistant Professor at Tilburg University 🇳🇱, interested in the neuroscience of decision making, aging, and Alzheimer's disease. Also a dad 🤓
More at https://www.reid-lab.org
115 posts
777 followers
783 following
Prolific Poster
Conversation Starter
comment in response to
post
Doable 😅
comment in response to
post
As a larger point, though:
I actually don't mind if students use ChatGPT, it's not going to go away, and even good students will use it. Universities will need to adapt to this.
It could be useful to have the LLM provide a response to a question and then ask students to critique its response.
comment in response to
post
To clarify, by "good", I mean "not lazy". That includes students who may struggle with the content but are engaged and put in the effort to learn.
It's also a disservice to them if classes get derailed by students who don't.
But point taken, some effort should be made to teach these things.
comment in response to
post
I'm pretty sure the good students understand this.
It's important to set the expectations from the start, yes. But at university level, students should already understand that failing to think for themselves will ultimately mean they will fail assessments and are wasting everybody's time. 🤷
comment in response to
post
I agree... how, tho?
The alternative is wasting time scolding students instead of teaching content. LLMs aren't going away. Best to teach students how to use them critically and constructively.
Along with the value of thinking though problems rather than blindly reiterating ChatGPT. 🤷
comment in response to
post
Well, also to their students 🤷
comment in response to
post
These ideological conflicts dominate on the right because of conservative movements increasingly inhabiting fantasy worlds that do not correspond to reality. I don't think "look, we've improved peer review" is going to reach these people.
comment in response to
post
Is the problem now tractable? Possibly (although still maybe not), using existing casual discovery algorithms.
The brain is highly recurrent, however, throwing a new spanner in the works.
My current thinking: with resting state it's probably futile. With perturbations it is probably doable. 2/2
comment in response to
post
If we stick with the idea that all 86b neurons need to be factored into any casual inference, I'm with you.
But I think there are many forms of casual inference that do not require this. As a thought experiment, pretend that it's okay to use aggregate signals from ~100 brain regions. 1/2
comment in response to
post
Didn't mention whether tea would be served tho...
comment in response to
post
About time we addressed that!