AI's potential for mass stupefaction

Recent project from a researcher the MIT Media Lab claims LLMs make you dumber.

Project link: https://www.brainonllm.com

> We assigned participants to three groups: LLM group, Search Engine group, Brain-only group, where each participant used a designated tool (or no tool in the latter) to write an essay
> performed using ChatGPT
> 54 participants
> We used electroencephalography (EEG) to record participants' brain activity in order to assess their cognitive engagement and cognitive load
> Brain connectivity systematically scaled down with the amount of external support: the Brain‑only group exhibited the strongest, widest‑ranging networks, Search Engine group showed intermediate engagement, and LLM assistance elicited the weakest overall coupling
> the LLM group's participants performed worse than their counterparts in the Brain-only group at all levels: neural, linguistic, scoring

Is this a qualitatively new problem, or merely an acceleration of existing cognitive problems previously precipitated by our use of the Internet as a mental crutch?

My gut take is that it is simultaneously not a real problem (we compensate for worsening mental powers with increasingly adept use of technology) and worse than you could possibly imagine (such technological dependence dramatically accelerates our alienation from other humans).

> We hooked up ELECTRODES and did a heckin SCIENCE… it turns out people THINK LESS when they make ChatGPT write an essay for them, than when writing an essay themselves.

> scale labeled “p values” from “more significant” to “less significant”, no numbers

You don’t hate clickbait academia enough.

I think if you read books about people reading the Bible and trying to decipher whatever makes sense from 3,000-year-old books, they really are just an older version of modern LLMs. People not understanding text data but still interpreting it.

The big problem is that most people use writing for thinking, and if people start polluting words with voice speaking, then our writing is going to be measurably less dense.

For example, I wrote most of this using Wispr Flow. I recognize that it's helpful for communicating but it also pollutes it because people usually just speak whatever comes to mind, whereas writing requires thinking.

I think writing and speaking increasingly are going to become less relevant. Whether somebody agrees with you or not depends on the words and phrases that people use—somebody agrees, or they disagree, or they don't do anything. There's no need to speak that much or to read that much.

For example, it looks like if somebody uses hyphens, that might be an indication that somebody is using the Wispr Flow app. This whole thing is just a DDoS attack on our basic sensibilities in writing. Yet, I don't see AirChat replacing communication. Looks like it's time to go back to the real world because the internet became wholly fake.