Why ChatGPT Could Be Making Us Less Intelligent: 6 Key Reasons

Not only is ChatGPT reportedly getting worse over time, it could be making us less smart too. A Stanford University study evaluated the performance of ChatGPT in March 2023 and June 2023 and found, “the behavior of the ‘same’ LLM service can change substantially in a relatively short amount of time.” Their recommendation was continuous monitoring of LLM (large language model) quality.

But what about the humans using ChatGPT? If LLMs are getting worse, what is happening to our focus, attention, memory recall and common sense? Is the real question how the humans using the tools are performing over the same time period, and who is continuously monitoring our quality? In some ways, ChatGPT could be making us smarter. In other ways, we might have degraded too. Here are 6 reasons why that could be the case.

How ChatGPT could be making us less smart

1. Dependence on instant answers

As soon as you open up ChatGPT, Claude, or your favourite LLM, you can type in any prompt and get a response. With instant answers now available, why would you expend any energy thinking? Sure, we’ve always had Google. But sometimes finding the right answer using Google takes more effort. Google gives a list of web pages to read and scrutinize. But with ChatGPT you get one answer, and that’s it. It’s never been quicker to summon information, which might be costing our ability to do it.

Over-reliance is a danger, as is the decline in critical thinking and problem-solving skills. We can simply skip the process of research and analysis on our own. So what’s the answer? Take longer to think. Use ChatGPT to confirm what you already concluded, not work it out for you. Use the enhanced capability to outproduce, not remove the need to think.

2. Reduction in memory retention

What’s that actress called? The one who was in that film, married to that guy, friends with that other actress? Recalling information used to involve furrowed brows and cogs whirring in our brains. Multiple people would chime in to give clues and figure out the answer together. Now, there is just no need to remember anything. We’ll get an answer faster than googling it. Not only do LLMs give responses, we can upload our own PDFs. Tools exist to ask questions of our own content. We don’t even need to remember what we wrote.

Requiring less information on tap might lead to storing less information in the brain. In a bid to become more efficient, memory cells diminish, memory retention goes out the window, and cognitive abilities slowly decline. Test yourself without it and see how you function.

3. Erosion of writing and communication skills

If an entrepreneur can learn how to prompt ChatGPT effectively and create their website homepage copy, create free resources or have their emails written, where’s the incentive to improve their written communication? Growing accustomed to automated assistance can make us less adept at expressing thoughts and ideas effectively solo.

The valuable skill is now not writing well, the valuable skill is prompting well. Prompt engineers are earning $300k salaries. The better you can prompt, the better an LLM can create the right output. Whether or not that’s making us less smart is anyone’s guess. It’s simply a different use of our brainpower. The art of written communication is now mastered in a novel way.

4. Limited understanding of complex topics

You can use ChatGPT to supercharge your learning, but how deep is the knowledge you’ll collect? While it’s brilliant at simplifying complex information, generating fun mnemonics and creating a structured study schedule, its light approach to education could mean your grasp of complex topics becomes far below your true capability. Rather than attending lectures and poring over textbooks, we’re asking for the top level of information; just enough to get by.

Writing its findings into pitches, presentations and courses might impress an audience at first, but when they ask further questions you’ll be found right out. Relying solely on ChatGPT for learning can cost in-depth knowledge and train you out of grasping intricate subjects, after which only basic brain function ensues.

5. Propagation of misinformation

One danger of constantly querying ChatGPT is that responses are not always accurate. Language models generate answers based on patterns in the data they were trained on. Relying on them could see you believing things that simply aren’t true. Instead of digging deeper to find the real story, we’re taking what we read at face value and assuming it’s right. Bias and misinformation propagates and rumours spiral out of control.

Not questioning further or failing to fact check is not a good idea. If someone in your team brought you information, you’d expect them to reference their sources. If a friend told you an exaggerated story, you’d ask to see proof or verification. Do the same with ChatGPT instead of blindly following its tales.

6. Reduced social interactions

A business owner replacing half their team with AI tools has eliminated half their social interaction each working week. Once those tools get more proficient, it could be reduced even further. Not only can you replace team members with technology, you can replace very human roles with ChatGPT, including training it to act as your personal AI business coach.

But humans need other humans. We know that isolation, over time, costs social skills and emotional intelligence, making us less good at understanding others and connecting with them. Could that make us less smart? Conversation and companionship via ChatGPT, meaning fewer face-to-face social interactions and reduced ability to function with other people. It’s a risk worth thinking about avoiding.

Could ChatGPT be making us less smart? Dependence on instant answers, reduction in memory retention, erosion of writing and communication skills, less able to grasp complex topics, believing and spreading misinformation and reduced proficiency in social interactions. Be aware of the pitfalls to take all the benefits without these very real costs.