Browse Comments — Raw (as collected)

Oh good, another study telling me AI is making us dumber. Let me summarize: researchers at Carnegie Mellon, Oxford, MIT, and UCLA gave people fraction problems, let them use AI for 10 minutes, took it away, and watched them struggle and from this they concluded AI impairs human cognition. Groundbreaking stuff, guys. Truly. Here's what they actually proved: if you give someone a crutch and then kick it out from under them, they fall. Someone call Stockholm. There's a Nobel Prize waiting. What they didn't study and clearly don't have the operational experience to understand is what happens to the brain of someone using AI to build something that has never existed before. The Fall of modern education is real!

It's quite simple, our brain has the innate tendency to lazyness, it's programmed in for purposes of efficiency and capacity restraints amongst others.

This is the wrong lesson to draw from AI. Of course, if you give people AI for fraction problems and then suddenly take it away, some will perform worse. That measures short-term tool dependency under artificial test conditions. But give people a real-world problem, such as air pollution in Punjab, Pakistan, and the result is very different. Without AI, most people would probably give up in the first five minutes. The problem is too complex, too interdisciplinary, and too overwhelming. With AI, a much larger number of people can begin. They can ask better questions, challenge their assumptions, explore causes, compare solutions, model tradeoffs, and keep working. The danger is not AI use. The danger is passive AI use. We should not teach people to outsource thinking. We should teach them to use AI to extend thinking. Amplify the thinking! The goal is not to make humans dependent. The goal is to help more humans participate in solving problems that were previously beyond their reach.

"people who have been doing a task for a little while are faster than people who just started" Groundbreaking

"Intelligence is not the ability to store information, but to know where to find it." - attrubuted to Einstein by a human who hallucinated sources. "The AI-assisted group also had a much higher rate of simply skipping questions once their access to AI was removed.⁠ " This may actually be a sign of higher intelligence that doesn't waste time on stupid questions that AI can answer.

It’s driving me insane. Can we do this? “An easy fix? One line!” It says! Two hours later it’s created 3 new bugs, crashed the system, destroyed a week’s work and it hasn’t added the one line fix! Claude is becoming utterly useless.

Feels like you might get the same results using an electronic calculator?

Desmond Sheppard Ok let us look at these studies for reference they point to an effect over time. The MIT Media Lab Longitudinal Study (2025),The ACM Critical Thinking Access Timing Experiment (2026)ScienceDirect Review on Cognitive Atrophy (2026) Longitudinal Study on Employee Self-Efficacy (2025) these are studies that indicate an effect that alarming. I maintain my position on bias or vested interested. My personal beliefs have nothing to do with the results of these studies. Where was the muutal respect applied when growing this tech into a job replacement and ruination of lives tool?

Tom Kirkman I would say Actual Idiocy. There's nothing Artificial about it!

Dr. Fred J. MIT study did not say AI physically "modifies" or damages the brain. Joint study by MIT, UCLA, Oxford, and Carnegie Mellon found that using AI can cause a temporary decline in problem solving skills and motivation, known as "cognitive offloading." Study found that people who used an AI assistant for 10 to 15 minutes performed significantly worse and gave up more easily when AI was taken away, compared to those who never used it. It was Gary Kucher above -- not the MIT study -- who said "just 10 minutes of AI use [can lead] to impaired brain performance.⁠"

Productivity gains from AI are hitting a token-cost wall as costs to AI model users are rising due to decreasing AI model efficiency. As shown in the chart here, the latest AI model releases are less efficient than preceding versions. The operating cost increases from prior models range from +35-75% (i.e. Gemini 3.5 Flash +75% vs. Gemini 3.1 Pro, GPT 5.5 +70% vs. GPT 5.4, Opus 4.7 +35%). In reality, Google, OpenAI, and Anthropic have all raised effective prices in the last six months, degrading the unit economics of frontier AI as they charge more because each model burns more compute per task. Note that companies are pulling back from AI agent use as budgets have been materially exceeded. For example, Uber exhausted its full year 2026 AI budget in just four months. Earlier this week, Microsoft canceled its internal Claude Code licenses after token-based billing made the cost untenable, this move by a company with effectively infinite cloud resources. AI promised productivity gains, but users are balking at the higher costs. #AI #TokenCost #Productivity

This sounds like a math teacher in the 70's when calculators were first introduced into classrooms.

Artificial intelligence is not good for Natural intelligence. Always chose natural over artificial!

The real risk is not that AI helps people think. It is that constant assistance quietly reshapes cognitive behavior itself. Once AI becomes ambient, effort allocation changes: less persistence, less independent reasoning, less tolerance for uncertainty. The dangerous part is not the tool. It is how quickly humans adapt around its availability.

Robert S. Yes like I said it’s horse manure news

John W. C. Your indignation is righteous, but placement matters just as much. Your position isn't the problem here.

We will adjust as we had with the calculator. And I would argue that it doesn't hurt your brain, it just changes the sets of skills needed and valued Unfortunatlly the system based on greed does not fit well with the newer technologies. Instead of creating an utopia is creating a distopia. Due to the system based on greed not being compatible. It will take a very big war before that happens.

Then you are using AI the wrong way.

Our brain works like a muscle: use it, or lose it.

The part this post leaves out changes the conclusion entirely. The 20% drop only hit participants who asked AI for direct answers. People who used AI for hints or clarifications showed no significant impairment. Same study, same data. That is not a technology problem. It is a design problem. I build Companion OS for people in crisis and recovery. The AI never solves for the person. It holds space while the person does the thinking. Dependency or capacity. The design decides. Preprint: https://arxiv.org/abs/2604.04721