Chat GPT Is Causing Cognitive Decline | ✉️ #72

Hey! 👋

A few days ago, the preprint of a study conducted at MIT on how frequent use of ChatGPT affects the cognitive functions of our brains made some buzz on the internet. The results, of course, are not comforting. Although it's only a preprint, and the work has yet to be peer-reviewed and replicated by other research groups, I think it's worth paying attention to anyway. I'm not qualified to properly interpret the results of the study, but what it describes is enough for me to make some adjustments to the way we use generative tools in our daily lives and work.

For example, the researchers found that 83% of the subjects were unable to cite their own essay that they had written with the help of AI. The conclusion is self-evident, you should at least read what you generate :) duh...

Further, the researchers found that when creating the same piece of text, the people who used the AI to do this got 47% fewer new neural connections in the process, even if they read what they were writing or understood what text they were creating. This is alarming to say the least. A direct path to dumbing down. The less neural connects you have = the dumber you get. This also might affect your long term neural activity resulting in early Alzheimer's. Researchers confirm that there is a direct correlation between the amount of effort invested in acquiring new information and the level of memorization. The more effort - the better the new information/knowledge is memorized.

Then the researchers delve into how this all leads to the accumulation of so-called “cognitive debt” and deterioration of the ability to think critically. But overall, the conclusions are clear. Using AI to facilitate routine activities - okay. Substituting it for acquiring our own knowledge - not okay.

But I wonder how this will change us in 20-30 years? Could it be that by memorizing less and being more stupid overall, we will free up brain capacity for more things that we can do, and in sum become more efficient and advanced versions of ourselves? Well, sort of like how now a high school kid with a smartphone in his hands will be clearly smarter than Aristotle. You know what I mean?

What We've Shared

• Built-in AWS tools for cost analysis and optimization: 94% of companies overspend on cloud—don’t be one of them. In this article, Kirill Shirinkin shows how to combine AWS Cost Explorer, Budgets, and Trusted Advisor to track, control, and cut your cloud spending
• runc, crun & Container Standards Wrap Up: In this Dockerless Course lesson, Kirill shows how to run containers with runc and crun, the runtimes behind tools like Docker. Discover what really happens under the hood — and why it matters.

What We've Discovered

• The Real Failure Rate of EBS: Fascinating learnings of EBS failures and how often do they happen: "That’s the true rate of failure of EBS: it’s constant, variable, and all by design".
• Scaling Prometheus: From Single Node to Enterprise-Grade Observability. An overview and comparison of open source tools to scale Prometheus horizontally.
• Teaching a new way to prevent outages at Google: Goes without saying - you don't need to copy Google. Still, it's good to know about some big tech level ways of attempting to prevent outages, in this case using System Theoretic Process Analysis (STPA).
• In S3 simplicity is table stakes: Posted on 19th birthday of S3, this article shares some behind the scenes on how S3 team constantly tries to make things simpler - and how complex doing so sometimes is.
• System Diagrams are Performance Caches for Cognitive Load: If you needed another reason to create diagrams of your systems, this article gives an excellent one.

The 73rd mkdev dispatch will arrive on Friday, July 11th. Subscribe here to get it. See you next time!