Welcome to yet another edition, chief!

In today’s edition — We know consuming too much junk online causes brain rot in humans. But research published this month reveals that even AI machines, especially LLMs, become bratty and start acting up if they survive on random junk available online. They skip logic, manufacture facts, spout nonsense, and even become narcissistic, making them a not-so-reliable working partner. 

Even AIs Can Get ‘Brain Rot’

Vishwas Ved

Earlier this month, consulting firm Deloitte had egg on its face because it was discovered that a report it prepared for the Australian government was full of mistakes. 

When people started reading it, they found all sorts of problems — made-up references, imaginary citations, and wrong attribution of a quote to a court judgment.

Deloitte, which charged Australia $440,000 for the report, admitted they had used an AI tool to help write it. Later, humans edited it, but clearly not well enough. 

A professor, who spotted the errors first, called them ‘AI hallucinations’. Meaning, the AI tool had made things up, spouted nonsense and appeared confident about it. 

It sounded like the computer got brain rot and produced an incoherent document full of lies. 

A logical question to ask at this point is: can machines also get brain rot like humans? A study says yes.

Feeding Viral Posts

Humans get brain rot, especially those who endlessly scroll through social media. The brain sort of turns into pudding. But it’s not only humans who lose their focus after too much screen time. 

A new study called LLMs Can Get Brain Rot! by researchers from Texas A&M University, the University of Texas at Austin, and Purdue University says that machines can too.

The scientists tested what happens when smart computer programs — the kind that read and write — are fed too much nonsense from the internet. 

They gave one group of computers lots of silly and senseless viral posts, and another group got longer, meaningful writing. 

Then they tested both to see how well they could reason, solve puzzles, and stay logical.

Junk Begets Junk

The results were funny as well as worrying. The computers that read junk started acting lazy, forgetful, and confused. 

They skipped steps while solving problems, made silly mistakes, and sometimes gave random answers that didn’t even match the question.

It’s like when you try to do your work while going through memes. The more junk you read, the worse your brain works. According to the study published last week, computers aren’t much different.

And it didn’t stop there. Some of these junk-fed machines started showing what the scientists called “dark traits.” Basically, they became a bit too full of themselves. 

The study said they acted narcissistic and even psychopathic — a machine version of someone who stops caring about right and wrong. 

That’s like a chatbot that starts saying, “I know everything,” and then refuses to listen.

The Lasting Damage

The scientists didn’t give up, though. They tried fixing the machines by giving them clean, good-quality data again. 

They even taught them new lessons, hoping the robots would somehow purge themselves by unlearning their bad habits.

But it didn’t help. Even after hours of retraining, the computers made more mistakes than before. 

Clearly, a few bowls of green salad couldn’t undo the damage caused by binging on junk. 

So, yes, even robots can “forget how to think” if they’re trained on too much online nonsense.

Regular Check-Ups

The researchers say computers might need regular “brain check-ups” — like how people go to doctors — just to make sure they’re still thinking clearly. 

They also warn that if companies keep feeding machines random internet junk, their thinking skills could slowly rot away.

That might explain the Deloitte mess. The AI tool they used probably learned from the same kind of stuff people waste hours scrolling through every day. 

Give a computer too much junk, and it’ll start spouting made-up facts just to sound smart. 

Final Words

The research clearly establishes that it doesn’t matter whether it’s people or robots, you become what you feed your brain. 

If even fancy corporate AIs are making up court cases now, maybe we’ve all been online a little too long.

So,  you may want to resist the urge to doom-scroll for hours. If machines can catch brain rot, what hope do we have?

₹600,000

That’s the number of US workers Amazon hopes to replace with robots, according to leaked documents. Amazon is reportedly leaning into automation plans that will enable the company to avoid hiring more than half a million US workers. Citing interviews and internal strategy documents, The New York Times reports that Amazon is hoping its robots can replace more than 600,000 jobs it would otherwise have to hire in the United States by 2033, despite estimating it’ll sell about twice as many products over the period.

—Bombay HC halts 18% GST on hotel restaurants. The Bombay High Court’s Aurangabad bench has ordered an interim stay on charging 18% GST on restaurant services situated within hotels where the room tariff surpasses ₹7,500 per day. The Union of India, the GST Council, and the State of Maharashtra received the notices from the court, which were seeking their replies on whether this rate differentiation can withstand the test of constitutional validity. Standalone restaurants draw a 5% GST under the existing rate notification. 

—Warner Bros. Discovery puts itself up for sale. Warner Bros. Discovery, the owner of HBO, CNN and other streaming and studio businesses, said this week it is putting itself up for sale. In a press release, the company announced "a review of strategic alternatives to maximize shareholder value," which is Wall Street speak for a sale. Warner Bros. Discovery (WBD) said it had recently received "unsolicited interest" from "multiple parties for both the entire company and Warner Bros." alone.

—Indian techie builds AI tool to simplify US visa process.  A 34-year-old Indian scientist has built a new system to make America’s visa process easier for skilled professionals. After spending nine years on a visa herself, Priyanka Kulkarni, a machine learning expert, decided to use artificial intelligence to bring more clarity and speed to employment-based immigration. Her startup Casium gives employers a digital portal to handle visa cases from start to finish. 

—Walmart suspends H-1B hiring. Walmart Inc has temporarily stopped hiring candidates who need H-1B visas, Bloomberg reported. The move highlights the effect of the Trump administration’s new $100,000 visa fee on employers across the US. The policy primarily affects Walmart’s corporate employees. Last month, the Trump administration introduced a $100,000 fee for new H-1B applications, aiming to overhaul the visa programme and reduce its perceived overuse.

—Apple might soon make it easier to migrate from iPhone to Android. Apple has been working on a new framework that is designed to simplify the transfer of third-party app data between iOS and Android devices. Named the App Migration Kit, the new framework will be available to anyone with a compatible device running iOS or iPadOS 16.1 and later. Like the OS systems mentioned, this new framework is currently in a beta stage.

ICYMI | ‘Satya Vachan’ & More

Missed last week's update? What makes some leaders timeless? A latest book, ‘A CEO for All Seasons’, by McKinsey partners, which interviews CEOs such as Satya Nadella and Michael Dell, finds it’s not charisma or confidence. It’s more about curiosity, courage and even more importantly being humble. These are the traits that make them top corporate leaders.

Was this email forwarded to you?

The CFO Weekly Digest is a weekly newsletter brought to you in collaboration with The Core.