Your Brain on ChatGPT
Our AI overlords are here to stay in the worst possible way. Cyberpunk genre was pretty spot on in hindsight. There has been worry about how these bots are being used for exams to cheat. Now, exams have been become more and more frequent over the years. It was pretty harsh even when I was a kid. So I get the need behind wanting to skip to the end.
Unfortunately, a preliminary MIT study has shed light on a potential problem with relying on chatbots to do all heavy lifting. It’s not like migrating from pen to a keyboard (or typewriter). Here, it’s the bot equivalent of asking your friend to do your homework.
I should go onto mention that this MIT study warns that it is still way too early to tell. They used a limited sample size and published ahead of being peer reviewed due to rapid state of things (and possibly current political climate).
The Study
In a nutshell, the study introduces the concept of cognitive debt. It points out that people who use GPT have a harder time reciting the answers compared to search engine and “brain-only” groups. When the GPT group was told to think on their own and brain-only group was given said bot, the latter excelled at their results while the former struggled. Despite the small sample size, the results were pretty damning.
As someone who has used GPT bots to brainstorm world building ideas (being one person and all), I can say that this is somewhat true. I was already building off of an established canon that simply needed polishing. I was looking for anything to bounce my ideas off of. My case fits the latter results. So there is some merit to these preliminary findings.
Consequences
I have seen two videos on subject so far. One was a week or so ago now from someone who is justifiably anti-AI, given the state of things. Another is an optimistic YouTuber who simply uncovers and explains AI research in a friendly manner. While the latter is a bit more bias, they both came to the same conclusion despite all the caveats: MIT has a point.
You don’t need to look far at what this bot dependency will cause than Twitter. Elon, the guy who pledged to get rid of bots, now has it as it’s top selling point with the Grok platform. People are using it as a fact checking tool instead of going to pople who do this for a living. The thing was already notoriously questionable as it is before going full MechaHitler.
At first, I thought that was someone else making a Wolfenstein 3D reference about it. I had no idea the bot legit said that! Of course, it tried to brush it off as a “joking” but we’ve all heard that before.
Despite feeling like the worst case scenario, the majority of the population outright reject AI being shoved down our throats. Self included. They were already a little iffy about it during the short lived IoT era. Even those willing to give the benefit of the doubt still see it working on limited use cases. The people with cognitive debt are likely to be those on Twitter, which already has an ever shrinking user base as it is.
Simple Solutions
Now, the study focused on SAT tests and exams. It’s just the subject of much debate. They were never about grading anyone or calling these people dumb. Just trying to make sense out of our fears.
But, like I said, exams are harsh. It’s no wonder students are resorting to extreme measure. The system needs to adapt to its people. Not the other way around.
The short term solution is to simply let people write their essays. In the long run, I feel the study’s methods provides answers. It was about noticing errors. On average, one group had less than the other. So, why not just help people fix said errors instead of failing hard working students?
