43 People Use ChatGPT To Age Up Childhood Photos, Then Compare Them Side-By-Side With Reality

There’s a lot of hype, joy, anger, and distrust surrounding generative AI tools. It’s a very emotionally messy subject. You’ll find many people who have embraced the ongoing technological shift, integrated tools into their lives, and swear by them. That’s on top of the countless people using these models for fun, to power up their businesses, or even for something as sensitive as therapy and relationship advice.

On the other hand, there are plenty of (healthy) skeptics out there, too, who think the over-the-top hype doesn’t match the actual usefulness of these models. Not to mention all the artists, writers, and creatives who don’t exactly enjoy having their hard work scraped and plopped into datasets used to train the latest AI models. Without their permission, of course.

And let’s not forget how many folks in white-collar positions are scared that they’ll be replaced by cheaper (and arguably worse-performing, unreliable, unimaginative) models.

Using LLMs (large language models) and generative AI might be handy, but it comes at a steep cost. Relying on them too much can erode your critical thinking skills, among other things.

Reporting on new research conducted by MIT’s Media Lab, Time Magazine explains that ChatGPT users had lower brain engagement than regular writers. The former “consistently underperformed at neural, linguistic, and behavioral levels” when using gen AI for essays.

According to the study, the individuals using ChatGPT actually got lazier with each essay they had to produce. By the end of the study, they often just copied and pasted whatever was generated.

While the study is relatively small and not yet peer-reviewed, the implication is more or less clear. Relying on LLMs might be convenient, but it harms brain development.

“What really motivated me to put it out now before waiting for a full peer review is that I am afraid in 6-8 months, there will be some policymaker who decides, ‘let’s do GPT kindergarten.’ I think that would be absolutely bad and detrimental. Developing brains are at the highest risk,” says the study’s main author, Nataliya Kosmyna.

In the meantime, other recent studies have found that people feel lonelier and socialize less the more they spend time chatting with ChatGPT.

In particular, people who trusted the chatbot and tended to get emotionally attached in real relationships felt lonelier and more emotionally dependent on ChatGPT.

Furthermore, discussing general topics led to more emotional dependence on ChatGPT, while discussing personal topics led to feelings of loneliness in the short term.

Even though many people have found uses for generative AI, like automating mundane work, these tools aren’t as powerful and reliable as some companies might have you believe.

Gary Marcus, a professor emeritus at New York University, author, and the founder of two AI companies, is one of the most insightful voices on the internet when it comes to speaking about the limitations and drawbacks of gen AI.

In a recent republished adaptation of his Substack essay in The Guardian, he covered Apple’s recent research paper about how LLMs and LRMs (large reasoning models) are limited in their reasoning, understanding, and general intelligence.

According to Marcus, Apple’s paper showed that models like ChatGPT, Claude, and Deepseek might appear to be smart. However, they collapse when complexity rises.

“In short, these models are very good at a kind of pattern recognition, but often fail when they encounter novelty that forces them beyond the limits of their training,” Marcus writes. He notes that the irony is that these models are explicitly designed for reasoning tasks.

Marcus points out that the recent paper amplifies an argument that he’s been making ever since 1998. “Neural networks of various kinds can generalise within a distribution of data they are exposed to, but their generalisations tend to break down beyond that distribution.”

When models work on tasks that are relatively close to the data they’ve been trained on, they perform fairly well.

However, “as they stray further away from that data, they often break down, as they did in the Apple paper’s more stringent tests. Such limits arguably remain the single most important serious weakness in LLMs.”

In other words, while useful, LLMs are still far too limited and flawed to lead to AGI (artificial general intelligence, which matches or surpasses human intelligence) that could “fundamentally transform society for the good.”

Feel free to share your thoughts in the comments at the very bottom of this post, Pandas. Which of these prediction vs. reality pics surprised you the most? How often and for what purposes do you use ChatGPT and other chatbots?

Do you think gen AI models would be able to accurately predict what you look like now, based on a single photo of you when you were little? How worried are you about your job being replaced by AI?