❎ New Grok 4.1. Honest and empathetic

xAI, Elon Musk's startup, has just released Grok 4.1. It's based on the same base as Grok 4, but with extra post-training emphasis on empathy, personality, and helpfulness.

Key Features:

✅ Far fewer hallucinations: Significant improvement in benchmarks.

✅ Arena leader: In the LMArena ranking, based on blind tests, Grok 4 Thinking surpassed Gemini 2.5 Pro and took first place. The non-reasoning model is in second place.

✅ The model is more creative, compassionate, and empathetic. But from xAI's own examples, it tends to produce much longer answers that can feel overly emotional at times.

➡️ You can try Grok 4.1 for free on the web, in apps, and via X.

@hiaimediaen

🎶 97% of People Can't Distinguish AI-Made Music from Human-Made Music

In a blind test of 9,000 people recently conducted by Ipsos Digital for streaming service Deezer, 97% couldn't distinguish AI-generated music from human‑made tracks. Each participant heard three songs, two produced entirely by algorithms.

71% of the respondents were surprised by these results and 52% felt uncomfortable by not being able to tell the difference.

While half of those surveyed agree that AI will play an important role in music creation in the coming decade, 80% say that streaming services should label AI tracks, and 52% believe that such music has no place in the charts alongside human-made songs.

@hiaimediaen

🤖 Two Founders Pretended To Be AI And Built A $1B Company

Sam Udotong, co-founder of the call-transcription startup Fireflies AI, admitted that at the start, he and his partner, Krish Ramineni, pretended to be AI. They joined meetings as a "bot," took notes by hand, and sent summaries 10 minutes after the call.

This went on for almost a year. By 2017, they had earned enough to rent an apartment, realized the idea worked, and decided actually to automate everything.

Still, many in the comments insist that Fireflies effectively misled clients and violated their privacy.

Later, Ramineni clarified that their earliest users were friends and acquaintances who knew about it. Today, he says, calls are genuinely processed by AI—and handled confidentially.

"Wanted to clear up the confusion for everyone asking. No human can take notes for millions of people every day, 24/7," he jokes.

💰 Today, the startup is valued at $1B and claims 20M users.

@hiaimedien

💉 An "Injectable Microchip" to Treat Brain Disorders Created at MIT

Deep brain stimulation is used to treat Parkinson's and Alzheimer's disease, epilepsy, depression, and multiple sclerosis. However, it requires risky brain surgery to implant electrodes.

Engineers at MIT have developed microscopic chips called Circulatronics. These can stimulate the brain from the inside without surgery. The cell-sized devices are delivered through an intravenous injection.

The team attached the chips to monocytes—immune cells that naturally cross the brain's protective blood-brain barrier. These living cells "cloak" the electronics from the immune system and carry the chips straight to areas of brain inflammation.

The chips don't need batteries. A pulse of near-infrared light shines through the skull, hits the chip, and powers it to activate nearby neurons.

🐀 The technology was tested in mice. Thousands of microelectrodes formed a network that precisely matched the shape of the inflamed brain region.

The devices didn't interfere with the surrounding tissue. Once activated, they stimulated neurons within a 30 μm radius—far more precise than traditional electrodes.

⏱ Human trials are expected to begin within the next three years.

@hiaimediaen

🧬 Preventive: a Startup Creating Gene-Edited Babies

San Francisco startup Preventive aims to protect children from genetic diseases before birth by editing embryos' genomes. The company has already raised $30 million. Investors include OpenAI CEO Sam Altman and Coinbase CEO Brian Armstrong.

Geneticist Lucas Harrington, founder and CEO of Preventive, says his team is committed to "complete transparency in research" and does not plan to conduct clinical trials on humans until the technology is proven to be completely safe.

He believes that preventive gene editing in human embryos "could be of the most important health technologies of the century."

💩 However, correcting the genes of embryos for reproductive purposes is prohibited in many countries, including the US, Russia, and China.

Chinese scientist He Jiankui, whose experiments resulted in the birth of the first "gene-editd twins" with immunity to HIV, spent three years in prison for this.

According to sources, Preventive is considering foreign jurisdictions—primarily the UAE—to continue research in case of legal challenges in America.

Is genetic engineering of embryos ethical?

❤️ — Ethical and even necessary
🤯 — It should be banned

@hiaimediaen

💬 GPT-5.1: A New Model from OpenAI

OpenAI has released GPT-5.1, an updated model that feels more human and conversational while also improving efficiency and overall performance. It addresses earlier feedback—many users had found GPT-5 too robotic compared with GPT-4o.

Main Features:

➡️ The Instant model is now more talkative yet still responds quickly and follows instructions better. It also knows when to take extra time to think through complex tasks.

➡️ The Thinking version has become more effective for everyday use. It now reasons more carefully on complex problems without getting stuck on simple ones. It uses fewer jargon words and undefined terms, resulting in clear, natural answers.

💡 Paid subscribers are already getting access. GPT-5.1 will gradually roll out to everyone and replace the base GPT-5. For the next three months, paying users will still be able to switch back to the old model.

A GPT-5.1 Pro release is also coming soon.

@hiaimediaen

At the recent Central Asia—US (C5+1) summit, Kazakhstan clearly staked its claim to regional AI leadership, signing about a dozen agreements with the world's leading AI companies.

💻 Groq, Oracle, HP, Cisco, and, of course, Nvidia want to develop data centers in the country. The latter has agreed to supply $2 billion worth of chips for the construction of an exaflop AI center.

But at the same time, Kazakhstan will have to resolve its energy crisis. The data center's capacity is 100 MW, yet the republic is forecasting a 900 MW generation deficit this winter. Without energy, there will be no data centers. 

Still, the country already has the only AI supercomputer in the region with a capacity of 2 exaflops. Telegram plans to use it for its own AI projects.

👨 More acute than the electricity shortage is the lack of AI specialists. According to KPMG, 24% of businesses are having difficulty finding AI-qualified personnel. Overall, only 8% of companies in the country have a clear AI strategy.

👨‍💻 Despite all the problems, top AI companies are actively trying to enter the emerging market. OpenAI will give the country's teachers access to ChatGPT Edu, and Perplexity will provide a free annual Pro subscription to all Kazakh citizens.

Moreover, not only American players are keen to expand into the region. The Kazakh AI model AlemLLM was developed in partnership with 01.AI, the unicorn of former Google China head Kai-Fu Lee.

@hiaimediaen