Currently sick, laid up in bed. but i keep seeing this "AI makes you dumb" line in response to the MIT Media Lab paper that came out recently and it's unbelievably incorrect:

- The paper is about educational outcomes for 3 constrained groups (search-only, brain-only, LLM-only). if you read the paper, a finding is if students only used ChatGPT to write an essay, they became more lazy, relying solely on LLMs. not sure that's a shocking finding, but now it's in the literature.

- A more important finding as suggested by the authors shows that well trained, non-lazy brains (brain-only participants when eventually allowed to use LLMs) showed a significant surge in activity when allowed to collaborate with LLM tools.

- A more pressing finding is that unfettered LLM usage may lead to a "a likely decrease in learning skills". The flipside is that we will get much "better" at output (albeit in a state of hegemony). Do we care about that? I do - but the labour market likely won't.

- The "using AI makes you dumb" line was widely reported by press and people who didn't read the paper ('cos why read a scientific paper when you can just make up an opinion based on your biases?). 

- As the main author at Media Lab knows full well that people don't read papers she literally made a website with an FAQ to allow laypeople to avoid misunderstandings. The first FAQ is that this paper does not suggest that using LLMs make us "dumber"

- The authors knew this would attract such headlines. They even rushed it out as a pre-print to avoid being scooped and have this big news and cultural story. 

Academia is very much part of the attention economy especially with AI being a ripe topic for doom and gloom as well as awe inspiring hope. 
So expect more of this and please read things!

Website: https://www.brainonllm.com/
Paper pre-print (non-peer reviewed): https://arxiv.org/pdf/2506.08872

i read the entire paper that everyone’s citing as “ChatGPT makes you dumb” and made a 90 second video on it if you have the attention to watch it

I saw a YT shorts of Christophe explaining the same (?) study here: https://youtube.com/shorts/-zTQ6tVHSwA

But I got to give you the win here,  you actually added the very important bit about the "switcheroo"! Instead of "ChatGPT is bad" you should say "ChatGPT can be bad if used improperly" (like you said 😁)

building farcaster | surfing @ venice, ca | eating @ gjusta / gjelina | “lucky me, lucky mud” 🤠