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

Counterpoint: the results may* have had more to do with the study design than anything. https://theconversation.com/mit-researchers-say-using-chatgpt-can-rot-your-brain-the-truth-is-a-little-more-complicated-259450

*Not saying it doesn't break one's brain, just that the study design isn't clear and the impact could be less than expected.

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