ChatGPT Is Raising Students' Grades, But Not Their Knowledge

The number of straight-A students at American universities has surged. AI is likely the cause, according to researcher Igor Chirikov, who analyzed over 500,000 grades from one large Texas university spanning 2018–2025.

He split all courses into two groups: "AI-vulnerable," where an AI model can complete assignments for students (essays, coding), and "AI-resistant," where such tasks are rare (math, physics, lab work). Before 2022, grades in both groups were nearly identical. After ChatGPT's launch, A grades in humanities and engineering courses jumped by roughly 30%—especially in courses with heavy homework loads.

Now, Chirikov says, an A grade might mean that a student has access to a more advanced technology model, or is better at using it—rather than measuring the underlying skill.

The most prestigious universities are feeling this the hardest. Harvard, responding to grade "inflation," has proposed capping top marks at 20% per course. Yale framed the problem even more bluntly in an April report:

"Grades exist to communicate what students have learned. At Yale, as at many peer institutions, they no longer do."

Has the grading system become obsolete?

❤️ — Yes, A grades are meaningless now
🔥 — No, but the criteria need updating

@hiaimediaen

Artificial Superintelligence Will Help People Live to Be 1,400 Years Old

This is what Swedish philosopher and writer Nick Bostrom states in his new essay, "Optimal Timing for Superintelligence."

He compares the implementation of strong AI to heart surgery. If it is delayed or doesn't happen at all, the patient might die. The surgery itself also carries risks, but the chances of radically improving the quality of life afterward are much higher and justify the risks.

According to Bostrom's assessment, even with some risk of humanity's destruction due to the technology, the benefits of the emergence of superintelligent AI are incomparably greater.

Superintelligence is capable of extending life to 1,400 years, he believes. If humanity were freed from disease and aging and mortality were fixed at the current rate for 20-year-olds (about 0.07% per year from random causes), that would be the average life expectancy.

Moreover, Bostrom calls this estimate conservative, since it ignores more radical "eternal life" scenarios, such as digitizing consciousness and uploading it into "backup copies".

Do you think AI will enable people to live for thousands of years?

❤️ — Yes, it's just a matter of time
🔥 — No, that's science fiction...

@hiaimediaen

"Genetic Shredder" Fights Against Cancer and Infections

An international team of researchers described a CRISPR-based Cas12a2 system that destroys a cell's DNA, triggering self-destruction.

The Cas12a2 protein can be programmed to search for specific RNA signatures, for example, ones associated with cancer or viral infections. When it detects its target, the protein activates inside the cell and begins shredding the cell's DNA.

"Its goal is not to correct anything. Instead, it's to destroy anything it sees," explains study co-author Yan Liu from the University of Utah.

In experiments, Cas12a2 reduced the growth of lung cancer cells carrying a common mutation by 50%—roughly comparable to chemotherapy. In another test, it destroyed more than 90% of cells infected with human papillomavirus (HPV). In mice, the system also slowed tumor growth.

At the same time, the "genetic shredder" left healthy cells untouched, which could reduce many of the severe side effects associated with traditional cancer treatments.

Medicine already uses CRISPR-Cas9, often described as "genetic scissors," which make precise edits to DNA. The technology has already been approved for treating Sickle Cell Disease, and clinical trials are underway for other inherited diseases and even HIV.

The new "shredder" works through a similar activation mechanism. Researchers hope that everything learned from years of Cas9 research will help accelerate the development of Cas12a2-based therapies.

What do you think about the idea of a "genetic shredder"?

❤️ — This could be revolutionary
🤔 — Sounds a little dangerous…

@hiaimediaen

Claude Helped Recover Access to a Bitcoin Wallet Worth $400,000

An X user cprkrn bought Bitcoin back in 2013 for around $200. But in 2015, while still a student, he changed the password to his wallet and forgot it.

Since then, his 5 BTC grew in value from about $2,000 to nearly $400,000. For the next 11 years, he says he unsuccessfully tried "around 7 trillion passwords."

A few weeks ago, he managed to identify the mnemonic phrase (the recovery seed) he had used before changing the password, but it didn't work with his current wallet file. As a last attempt, he uploaded the contents of his old computer into Claude, including files, notes, and backup folders.

The AI found an older wallet file created before the password change. It also discovered a bug in BTCRecover, a popular wallet recovery tool. The software was incorrectly combining the key with the password during decryption. Claude fixed the logic, ran the recovery process, and extracted the private keys from the old wallet backup.

To verify the story, the user shared the wallet address. Blockchain records matched his story: the 5 BTC had remained untouched since 2015 and were moved only on May 13.

This was not a case of hacking the password; instead, success relied on locating the right information. The user had all the necessary data and potentially could have recovered the wallet eventually. The AI helped organize and identify what was needed, effectively helping him "find the needle in the haystack."

The user described the day as the happiest of his life and joked about naming a future child after Dario Amodei, founder of Anthropic.

Happy for him?

❤️ — Absolutely
🔥 — Not entirely…

@hiaimediaen

Korean Scientists Created Rings That Translate Sign Language into Text

Seven wireless WRSLT rings are worn on the fingers. They track movements and convert gestures into text with a delay of just 0.1–0.3 seconds.

Each ring contains an accelerometer, a magnet, and a Bluetooth module for connecting to a smartphone or PC. The rings are not connected by wires, allowing the hands to move freely.

The system was trained on just 100 words of American Sign Language and 100 words of International Sign Language. In tests, recognition accuracy reached 88.3% and 88.5% respectively, with no calibration for the wearer.

For now, it's still a prototype. In the future, the technology could be adapted for controlling virtual reality, touchless interfaces, or post-injury rehabilitation.

What do you think?

❤️ — A very useful thing!
🤔 — I doubt it will catch on...

@hiaimediaen

Women Are 60% More Likely Than Men to Be Injured in Car Crashes

Researchers at Graz University of Technology reached this conclusion after analyzing crash data from Austria between 2012 and 2024. The study focused on 20,000 accidents involving a man and a woman in the same car, in which at least one person was injured.

Women were injured in 90% of crashes, compared to 57% for men. Serious injuries among women were also recorded at lower collision speeds.

To understand why, researchers recreated some of the least obvious crashes using computer simulations. They adjusted seat position, seatback angle, crash speed, and seat-belt and airbag behavior.

One major reason for the higher risk may be that vehicle safety systems have been designed and tested for decades around the "average male body." Standard crash-test dummies are modeled on an average-sized man, while "female" dummies are often just smaller versions of the male model. They don't properly account for differences in pelvis width, shoulder shape, or chest structure. According to the researchers, these models fail to accurately represent 95% of women.

Women are also more likely to ride in the passenger seat, where reclining too far back or sitting farther from the dashboard can reduce the effectiveness of seat belts and airbags. The underlying issue is clear: crash tests are usually based on "average" occupants sitting in an ideal upright position.

The researchers recommend developing seat belts that adapt to a passenger's body shape and seating position, as well as testing safety systems on a wider range of body types, including digital human models that could reduce the need for millions of physical crash tests.

Is it fair that car safety systems have historically been designed around men?

❤️ — Yes, men drive more often
🎃 — No, that's discrimination
🤔 — It's just how the industry evolved

@hiaimediaen

A Neuroimplant to Restore Vision

A wireless brain implant designed to restore vision, developed at the Illinois Institute of Technology, was recently implanted successfully in the third participant of an ongoing clinical trial.

During the procedure, surgeons placed 34 stimulators with 544 electrodes into the completely blind patient's brain. He will now begin extensive training to learn how to use his new form of vision.

The device bypasses damaged optic nerves, transmitting signals directly to the visual cortex of the brain.

To begin with, the brain gradually learns to interpret simple visual signals: the person begins to see glowing dots and the outlines of large objects. Over time, they can associate these patterns with objects, direction, obstacles, and learn to navigate their surroundings.

The technology is designed for people with damage to the retina or optic nerve for whom conventional vision recovery methods do not work.

"For people who are completely blind, even the ability to perceive a small amount of light can profoundly impact their daily lives," explains Janet P. Szlyk, president of The Chicago Lighthouse's Hilton Center for Prosthetic Research.

@hiaimediaen

Google Is Releasing AI Laptops

Google announced a new lineup of computers called Googlebook. They'll run a desktop version of Android (codenamed Aluminum OS) with Gemini baked right into the system.

"It's like if a Chromebook and a Copilot+ PC had a baby," Mashable quips.

AI is built into the cursor. The agent constantly tracks what's on your screen—shake the cursor, and Gemini will offer contextual suggestions: inserting a date into an email, adding notes to your calendar, or merging selected photos.

Gemini can build desktop widgets. Alarms, countdowns, hotel and restaurant reservations, interactive notes—the whole system is designed around deep personalization and adaptability.

Full Android phone integration—all your mobile apps will run on the Googlebook, and you'll be able to pick up right where you left off on your phone, straight from your laptop.

Google is building the hardware with partners: Acer, Lenovo, ASUS, Dell, and HP are already on board. Premium materials and top-notch build quality are promised.

First devices hit shelves this fall. No pricing yet. But at $500–$700, Googlebooks could be serious competition for cheap, sluggish plastic Windows laptops, something like a "MacBook Neo for Android fans."

Would you buy a laptop from Google?

❤️ — Maybe, looks interesting!
🔥 — Nope

@hiaimediaen

Italy Reports Its First Official Case of AI Addiction

Italy's addiction treatment service SerD has registered the country's first case of behavioral addiction to an AI chatbot. The patient is a 20-year-old woman from Venice who replaced real-life social contact with conversations with an AI model.

According to Laura Suardi, chief physician at Venice's SerD, the case is "the tip of the iceberg." Treating such disorders, she says, takes more than simply limiting access to devices. It requires work with psychologists, psychiatrists, and the patient's family.

Neuropsychiatry professor Stefano Vicari compares this kind of addiction to substance addiction. He says patients can show the same compulsive urge to return to the device, agitation and aggression when access is restricted, habituation, and growing tolerance.

"People need more and more time online to feel pleasure. In the end, the same brain areas are involved in these processes as in addiction to cocaine or cannabis," he explains.

Do you think AI addiction is a serious problem?

🔥 — Yes, and it will only get worse!
🤣 — It's just a way to scare people

@hiaimediaen

Scientists Transplanted the Naked Mole-Rat's "Longevity Gene" into Mice.

Naked mole-rats hold the record for lifespan. They live up to 41 years—almost ten times longer than other rodents of similar size. They almost never get by cancer, arthritis, or other age-related diseases.

These rodents owe their longevity primarily to the large amount of high-molecular-weight hyaluronic acid (HMW-HA) in their bodies. Its level in naked mole-rats is about ten times higher than in mice or humans.

In 2023, scientists from the University of Rochester (U.S.) transplanted a naked mole-rat gene responsible for HMM-HA synthesis into mice. The genetically modified mice then proved to be more resistant to tumors, maintained good gut health, and showed significantly less tissue inflammation as they aged.

The median lifespan of the test rodents increased less dramatically, by 4.4%. On the other hand, for a person living to age 80, this would mean approximately 3.5 additional years of life.

The researchers' next goal is to try to transfer this effect to humans.

@hiaimediaen

AI Drops You Into a Korean Baseball Broadcast

A new trend is spreading across social media: AI creates a video where you end up in a live sports broadcast, as if a camera operator accidentally caught you in the stands — and it looks incredibly realistic.

Go to @GPT4Telegrambot

Select Image Generation → GPT Image 2 and attach your photo with the prompt:

로드한 인물 얼굴은 실제 그대로 유지하고, AI 미남/미녀 느낌 말고
"진짜 KBO 생중계 카메라에 우연히 잡힌 일반인 관중" 처럼 자연스럽게 생 성.
SPOTVIKBO 방송 캡처 느낌, 관중석 직캠 구도, 주변 관중•맥주컵•응원도구•선풍기 자연스럽게 배치.
중요:
얼굴 과보정 금지
눈 키우기 금지
턱 보정 금지
피부 뽀샤시 금지
화보 느낌 금지
인플루언서 느낌 금지
실제 방송처럼:
약간 흐린 생중계 화질, 압축 노이즈, 미세한 모션블러, 현실적인 피부결, 잔 머리와 땀광 표현.
다리 꼬고 편하게 경기 보는 모습, 카메라 의식한 듯 안 한 듯한 자연스러운 표정.
핵심은:
"AI가 만든 사람"이 아니라
"진짜 방송에 잡혀 화제된 일반인"
같은 현실감

To bring the image to life, go to the menu and select Video Generation → Kling AI, version 3.0. Use this prompt:

Add gentle, believable motion to the main subject in the center of the frame, making it feel like the person is being shown live by a real stadium broadcast camera during a busy baseball game. The subject should respond to the action naturally, with realistic eye direction, small head movements, blinking, subtle breathing, slight shifts in expression, and spontaneous reactions typical of someone watching the game from the stands.
The behavior should feel candid and unposed, as if the broadcast camera briefly picked out this person in the crowd during a live telecast. The audience around them should also appear active and convincing, with natural crowd movement, an excited game-day atmosphere, and the visual feel of an authentic sports broadcast.
Keep the overall result immersive, cinematic, and highly realistic — like a genuine moment from a televised baseball game. Preserve a natural posture, subtle blinking, and true live sports broadcast energy. Avoid making the eyes look crossed.

In a few minutes the bot will send you the finished clip!

What do you think of this trend?

❤️ — Love it!
🔥 — Way too AI...

@hiaimediaen

AI Makes Houseplants "Talk"

Students from Spelman College in Atlanta have created an AI-powered device that analyzes a plant's condition and offers care tips.

Sensors collect real-time data on soil and air moisture, sunlight, and temperature. The AI processes this information and voices advice on behalf of the plant. For example: "So maybe it's time for some fertilization to get me pumped up again."

"PlantGPT is a way to be able to talk to your plants, so when your plant is dying, you need to know what type of information the plant needs or resources the plant needs," the device's creators explain.

The project is still in development—the students want to reduce the number of sensors and make the device more user-friendly for general public. The team hopes to launch it on the market in the near future, including for use on small farms and in urban gardens.

What do you think of the idea?

❤️ — Looks awesome!
🔥 — Plants are better off staying quiet...

@hiaimediaen

60% of Today's Jobs in the U.S. Did Not Exist in 1940

These are the findings of the report recently published by Bank of America. The analysts argue that apocalyptic predictions about AI "sit uneasily with both economic theory and the evidence so far."

The labor market is constantly transforming, economists note. For example, agriculture, which employed about 40% of Americans at the beginning of the 20th century, now accounts for only 1% of jobs. E-commerce hasn't killed retail either: in the U.S., about 16 million people still work in this sector—roughly the same as in the 1990s.

"Adaptability is the new job security," the report states.

According to the bank's estimates, the implementation of AI will affect one in four jobs worldwide—a total of about 840 million positions, and in high-income countries—up to one-third. But this doesn't mean people will lose their jobs: only 2.3% of professions could become fully automated.

Will your profession still exist in 50 years?

❤️ — Yes, definitely
🔥 — Probably not...

@hiaimediaen

The Creators of South Park Wanted to Make a Trump Deepfake, And Ended Up Founding an AI Startup

South Park creators Trey Parker and Matt Stone have unexpectedly become some of the most interesting players in Hollywood AI. Their company, Deep Voodoo, grew out of an idea for a Donald Trump movie. Back near the end of his first term, they wanted to make a deepfake comedy where the president's face would be "pasted" onto another actor's body—and the character would gradually lose his mind and his clothes. But no studio could deliver the quality they needed.

So Parker and Stone assembled their own team of AI specialists. That's how the startup was born—one that creates deepfakes, visual effects, and de-aging for actors. The company has already raised $20 million in funding.

Deep Voodoo operates on a much bigger scale than a typical indie studio. The startup films actors in its own capture space with nine cameras, collects up to 300,000 images, and builds a separate model for each individual project. Their work includes a Kendrick Lamar video, Super Bowl ad campaigns, and yes, the deepfake of naked Trump that ultimately made it into South Park

The irony is that America's biggest animation troublemakers are trying to be the "good guys": Deep Voodoo requires permission to use actors' likenesses and turns down projects without clear rights.

Stone and Parker also aren't ruling out using AI in South Park itself—to work faster, have more visual options, and "make the show better."

Does South Park need AI?

❤️ — Yes, that would be awesome!
🔥 — No, absolutely not...

@hiaimediaen

Quantum Computer Helps Simulate a Protein with 12,000 Atoms

To develop new drugs, scientists search for molecules that can bind effectively to specific proteins in the body.

The challenge is that even the most advanced classical computer models can't fully capture the quantum effects at such a tiny scale—even supercomputers don't have enough power to compute everything directly. Quantum computers could theoretically solve these problems much faster, but today's systems still aren't powerful enough on their own.

Researchers from the Cleveland Clinic, RIKEN, and IBM found a workaround by combining the IBM Heron quantum processor with the Fugaku and Miyabi-G supercomputers—some of the most powerful computing systems in the world.

Each system handled the tasks it's best suited for. The supercomputers split the molecule into fragments and assembled the overall simulation, while the quantum Heron calculated the physics of the most complex regions, where classical methods tend to lose accuracy.

Using this hybrid approach, the team successfully simulated a protein complex containing 12,635 atoms—about 40 times larger than the team's previous record. Accuracy during one of the key stages improved by a factor of 210.

"For years, quantum computing has been a promise. Now, quantum computers are producing results that matter to science," says Jay Gambetta, director of IBM Research.

Do you believe quantum computers will become truly useful?

❤️ — Yes, they're the future
🤔 — So far, it feels like hype

@hiaimediaen

A "Mind-Reading" Hat Instead of a Brain Implant

Startup Sabi claims it has developed a hat that lets you type using nothing but your thoughts.

It appears to be a regular cap, with approximately 100,000 miniature EEG sensors designed to detect electrical activity in the brain through the scalp and skull. These signals are processed by a "brain foundation model" trained on 100,000 hours of data from around a hundred volunteers.

The system converts thoughts into text at speeds of up to 30 words per minute—roughly on par with typing on a keyboard.

No implants required—according to the company, all you have to do is sit and think.

It Works?

So far, the company hasn't released any studies explaining how the system works or how accurate it is.

The challenge is that EEG signals are extremely weak and noisy. In lab settings, decoding thoughts into text is often only slightly better than random guessing. Last year, Meta showed a system that reached about 80% accuracy, but it relied on a large, capsule-like setup, not something you could wear.

Most existing systems use hundreds of EEG sensors, not tens of thousands. It's possible Sabi has made a breakthrough, but experts are waiting for evidence.

Would you wear something like this?

❤️ — Yes, that sounds convenient
🤔 — I don't want my thoughts read

@hiaimediaen

ChatGPT Chat Becomes Evidence in Murder Case

"What happens if a human has a put in a black garbage bag and thrown in a dumpster?"

Hisham Abugharbieh of Florida asked ChatGPT this question a few days before he was charged with the first-degree murder of two of his roommates.

The body of one of the victims was later found in a large black trash bag near the house.

The suspect also asked ChatGPT whether it was legal to keep a gun at home without a license, whether it was possible to change a car's license plate number, whether one could survive a sniper bullet to the head, and whether neighbors would hear a shot from his gun.

This correspondence is now being used as evidence in the investigation. In no country does the law protect the confidentiality of conversations with AI in the same way as, for example, conversations with a lawyer or doctor, or even phone calls.

"In my firm, we're treating it as: Anything that somebody's typing into ChatGPT is something that could be discoverable," explains attorney Virginia Hammerle.

Should authorities have access to correspondence with AI?

👍 — Yes, it's a security issue
🔥 — No, it's a violation of rights

@hiaimediaen

A Court in China Ruled That an Employee Cannot Be Fired Just Because They Are Being Replaced by AI

In China, a worker won a lawsuit after being fired due to automation. Zhou, 35, was responsible for checking whether an AI's responses were accurate at a fintech company. In January 2025, management informed him that he was being demoted: his salary would be cut from 25,000 to 15,000 yuan per month (roughly $2,200). Zhou refused, and he was fired. During the trial, the company explained that AI could now perform his duties.

The court sided with the employee and ordered the company to pay him more than 260,000 yuan in compensation (nearly $380,000). The core logic of the ruling is simple: if a company chooses to implement AI to cut costs, it cannot shift the consequences of that decision onto the worker. In addition, under labor law, the company was required to offer Zhou a reasonable alternative to his lost position, whereas the court found a nearly 50% pay cut to be unreasonable.

This does not mean China has banned companies from adopting AI or reducing headcount. But workers now have an important argument: the mere fact that AI can do the job more cheaply is not enough to justify termination. The employer must prove that the layoff is truly necessary, not just profitable.

In China's legal system, a ruling like this does not become precedent in the American sense, but it does serve as a signal to other courts and companies: automation may change how work is done, but it does not erase workers' rights.

A fair decision?

❤️ — Yes, workers should not suffer
🔥 — No, this hurts business

@hiaimediaen

London Schools Use VR to Relieve Student Stress

Relaxing VR sessions help students cope with exam stress, ADHD, and anxiety caused by family problems.

The seven-minute VR tour begins with gradual immersion: the room empties, the lights fade, and the user finds themselves in darkness with soft light effects moving toward them. It helps students calm down and ground themselves.

Students can use VR either at scheduled times or whenever they feel overwhelmed.

In the first ten schools to test it, 90% of students experienced immediate stress relief after the session. Teachers also report improvements in attendance and behavior.

"Exams terrified me. They don't scare me as much any more. It was the most terrifying thing ever. And when I felt that way I would use the headset and it would help me process how I was feeling about it," says 16-year-old Laura Wilson.

What do you think?

❤️ — Every school should have this!
🤔 — Kids need real psychologists...

@hiaimediaen

Kling Motion: Transfer Movement from Video to Photo

Now with @GPT4Telegrambot, you can upload a video with any movement and your own photo, and the Kling AI will map the motion from the clip onto your character.

Works with any reference footage: a TikTok dance, Reels gestures, or that one viral move everyone's doing.

How to Try:

In the @GPT4Telegrambot menu, tap Video Generation.
Select Kling Motion.
Upload the movement video, then add your photo.
Choose the Kling version and quality: 720p or 1080p.
Start the generation.

Pricing depends on the version, quality, and length of the original video.

@hiaimediaen