Discover • Tech News

Brain Decoder Reads Thoughts According to researchers at The University of Texas at Austin, an innovative AI-driven brain decoder can now translate a person's thoughts into coherent text with just a quick brain scan and minimal training, offering new hope for enhanced communication in individuals with language disorders. Read More

✖️ Musk's xAI Unveiled Its Grok 3 Today, Elon Musk presented Grok 3, the newest AI model from his startup xAI. Grok 3 is the first model to achieve over 1400 points on the LMArena leaderboard. Main Features: 🔜 The base Grok 3 mini performs on par with Google Gemini 2 Pro, DeepSeek-V3, Claude 3.5 Sonnet, and GPT-4o in coding, math, and exact sciences. The full model outperforms its competitors by roughly 1.5 times. 🔜 In reasoning mode, Grok 3 surpasses OpenAI's o3-mini-high by 5-10%. The new Big Brain mode pushes the model to think even more effectively. 🔜 Deep Search: new xAI agent for research on the web powered by Grok. The model will be available starting today for Premium+ X subscribers. Read More

⏳ Time May Flow Both Ways in Quantum Systems Physicists at the University of Surrey have uncovered fascinating evidence that challenges the traditional "one-way street" view of time. Their research suggests that, in certain quantum systems, time might flow forward or backward — a concept known as quantum time symmetry. Here’s what they found: ➡️ Some quantum systems follow equations that work the same in both time directions. ➡️ This opposes the classical idea of time always moving from past to future. ➡️ It hints at the possibility of opposing arrows of time within quantum mechanics. This discovery could reshape how we understand time and its connection to the quantum and macroscopic worlds. Time travel, anyone? Read More

📣 Hello everyone! Our Sunday digest features the most exciting AI news from Week 7, 2025. ▎OPENAI'S ROADMAP 👐 GPT-4.5 will be released in the coming weeks, said OpenAI CEO Sam Altman. And GPT-5 will launch within a few months—it will be free for everyone. ▎SAVE THIS — IT'S HELPFUL 💬 Tough Tongue AI: free chatbot for interview and negotiation training. 🔴 Video Effects by Pika AI are now available on @GPT4Telegrambot! Add objects, people, or fantastical elements to your clips. ▎TO READ 😋 Experiment: Can AI voice the Simpsons better than a professional actor? 🕶 Top 5 unique features of Ray-Ban Meta "smart" glasses. Ⓜ️ Meta is making a significant investment in AI-powered humanoid robots. 🐦 X Money: Why is Musk building his payment system within X? 💲 Musk-led consortium of investors has made a $97.4B bid to take control of OpenAI. 📸 AImagine: The AI photography exhibition opens in Brussels. 👁 How do AI hallucinations help scientists develop new drugs and train robots? 🖥 All that we know about Safe Superintelligence Inc., a new startup launched by Ilya Sutskever, co-founder of OpenAI. ▎TO WATCH 📺 A selection of helpful lectures on LLM and AI Agents. Enjoy your Sunday 👍

Ⓜ️ Meta Invests In Humanoid Robots According to Bloomberg, Meta is making a significant investment in AI-powered humanoid robots, forming a new team within its Reality Labs hardware division to develop robots capable of performing household chores and physical tasks. Elon Musk believes that by 2040, the number of humanoid robots will exceed the global human population. Read More

💗 Valentine's Day Video Effects by Pika AI Hi everyone! Just today, we rolled out new video generation services by Pika AI in our bot, and we already have an update! To celebrate Valentine's Day, Pika has introduced six fresh effects that bring your photos to life ⤴️ How to Use: ➡️ Go to @GPT4Telegrambot 1️⃣ Purchase the "Video" package in the /premium section. 2️⃣ Send the /video command to the bot, select "Pika Effects," and upload a photo. Within a minute, Pika will transform your image into an adorable or funny video that can become a heartfelt greeting for your loved one. Have a wonderful holiday 🌹

🎬 Video Effects by Pika AI on @GPT4Telegrambot Hi everyone! We've launched two new video generation services from Pika AI in our bot: 🧩 Pikaddition adds anything and anyone to ANY video. You can upload your non-AI videos and add objects, people, or fantastical elements. Mind-blowing! 💫 Pika Effects brings your images to life with various visual effects. Upload a photo and choose any of the 16 effects. You can inflate, squish, explode, melt, and more. Pika will turn images into realistic videos. Check out the examples above ⤴️ How to Use: ➡️ Go to @GPT4Telegrambot 1️⃣ Purchase the "Video" package in the /premium section. 2️⃣ Send the /video command to the bot and select either Pika Effects or Pikaddition. Have fun!

🔴 @GPT4Telegrambot — the #1 bot for using AI on Telegram. It can write texts and code, translate languages, solve math and physics problems, work with documents, and create images, videos, and music. 20 million users.

👐 OpenAI's GPT-5 Roadmap According to recent statements from OpenAI CEO Sam Altman ⤴️, the company plans to simplify its AI offerings with the upcoming release of GPT-5, aiming to provide users with a more streamlined and intuitive experience that eliminates the need for manual model selection. Read More

🤔 Microsoft Study: AI May Harm Critical Thinking A study by Microsoft and Carnegie Mellon found that heavy reliance on generative AI can reduce critical thinking among workers. Surveying 319 workers and analyzing 936 AI use cases, researchers observed that many accept AI outputs without proper evaluation, mistaking minor edits for critical thought. Key insights: ➡️ High trust in AI lowers critical thinking effort. ➡️ Strong self-confidence improves critical thinking and evaluation of AI outputs. The findings highlight the need to balance AI use with independent problem-solving skills. Read More

Russia's Plasma Engine Prototype Russian scientists at Rosatom have unveiled a prototype plasma electric rocket engine that could advance space travel, with claims that it could reduce the journey time to Mars to just 30-60 days. This groundbreaking technology, which accelerates charged particles to speeds of 100 km/s, represents a significant advancement over traditional chemical rockets and could pave the way for faster interplanetary exploration. Read More

Musk-Led Group Bids for OpenAI Based on reports from the Wall Street Journal, a consortium of investors led by billionaire Elon Musk has made a $97.4 billion bid to take control of OpenAI, the artificial intelligence startup behind ChatGPT, marking a significant development in the ongoing power struggle between Musk and the company he co-founded. Read More

Server error; please try again. Administrators have been notified of the incident.

💫 Scientists Discover the Largest Structure in the Universe Astronomers have uncovered a colossal cosmic structure named "Quipu," stretching an incredible 1.3 billion light-years across—more than 13,000 times the size of our Milky Way! This massive formation holds the equivalent of 200 quadrillion suns in mass. Its discovery challenges what we know about the universe, offering fresh clues about galaxy formation, dark matter, and the evolution of large-scale cosmic architecture. Read More

📣 Hello everyone! Our Sunday digest features the most exciting AI news from Week 6, 2025. ▎NEW RELEASES 👐 OpenAI has launched Deep Research, an advanced agent capable of conducting extensive online research autonomously. ♊️ Google's Gemini 2.0, the latest release in its AI model family, is now widely available. It features Flash and Pro models for diverse use cases.

🔴 Gemini 2.0 Flash Thinking, the top-ranked model on LMArena, is now free to use on @GPT4Telegrambot

▎SAVE THIS — IT'S HELPFUL 🇨🇳 Discover the best AI models from China — innovative tools you don’t want to miss! ▎TO READ 🧠 AI usage may weaken critical thinking, researchers discovered. 💻 "Source Code: My Beginnings": Bill Gates has released the first volume of his memoirs. 💬 Sam Altman, CEO of OpenAI, shared strategies to compete with DeepSeek and outlined insights on the company's future. 🤩 Neko Health, a medical AI body-scanning startup, raised $260M at a valuation of $1.8B. 🤩 A Neuralink patient learned to control a robotic arm with the power of thought. 🤖 Figure AI aims to build 100,000 advanced humanoid robots within four years. 🖼️ Apple has unveiled a lamp-like tabletop robot prototype with lifelike movements. 🎞 Flawless AI: How filmmaker Scott Mann's startup is changing Hollywood post-production. ▎TO WATCH 📱 Lex Fridman's latest podcast discusses how DeepSeek entered the AI race, the US-China rivalry, and AGI's impact on the world. Enjoy your Sunday 👍

🖼️ Apple Prototypes Tabletop Robot Based on reports from MacRumors, Apple has unveiled a prototype of a lamp-like tabletop robot with lifelike movements, showcasing advanced interactive capabilities and hinting at the company's potential entry into the home robotics market by 2027. The device incorporates a camera, projector, and speaker alongside its LED light, enabling it to serve as a multifaceted smart home assistant. It can: ➡️ Move and act like it's alive ➡️ Follow your hand movements ➡️ Learn your habits ➡️ Show info on walls ➡️ Play and dance to music ➡️ Display weather ➡️ Act friendly and emotional It's meant to feel more like a helpful friend than just another gadget. Read More

Researchers Create $50 o1 Rival Based on reports from TechCrunch, researchers at Stanford and the University of Washington have developed s1, an open-source AI reasoning model that rivals OpenAI's o1. It uses less than $50 in cloud computing credits and a novel distillation process involving Google's Gemini 2.0 Flash Thinking Experimental model. The groundbreaking s1 model was developed using a carefully curated dataset of 1,000 questions and answers, and the entire training process took just 26 minutes. Read More

⚡️ Gemini 2.0 Flash Thinking — FREE on @GPT4Telegrambot The powerful reasoning model from Google is now available to all users of our bot. What's interesting about it? ➡️ Gemini 2.0 Flash Thinking is the leader in the Chatbot Arena rating, where users vote for the AI models. ➡️ The knowledge cutoff: up to August 2024 (for comparison, OpenAI o1 — up to October 2023). ➡️ Self-checking: before giving an answer, it analyzes intermediate results, reducing the errors. How to use it? 🔴 Go to @GPT4Telegrambot 1️⃣ Send the command /model 2️⃣ Select Gemini 2.0 Flash Thinking Have fun 👍

🤖 Figure AI's 100k Humanoid Robot Army Figure AI, a U.S.-based robotics company, plans to build 100,000 advanced humanoid robots in four years. Backed by strategic partnerships with BMW and OpenAI, as well as investments from Microsoft, Figure AI is leveraging cutting-edge AI technologies to become a global leader in robotics. Read More

📣 Hello everyone! Our Sunday digest features the most exciting AI news from the fifth week of 2025. ▎DEEPSEEK VS OPENAI 🐳 How did Chinese DeepSeek blow up the AI industry? Here's our in-depth review of the week's biggest story. 💻 OpenAI has launched o3-mini, a powerful and fast model, delivering exceptional STEM capabilities—with particular strength in science, math, and coding.

🔴 OpenAI o3-mini is already available on @GPT4Telegrambot. Check the "ChatGPT + DALL-E 3" bundle.

▎BOT UPDATES 🎸 Suno V4: The latest version of the best text-to-audio AI model is now live on @GPT4Telegrambot ▎TO READ 🔄 "Humanity's Last Exam": The most challenging benchmark for AI. o3-mini scores 13%, DeepSeek-R1—9.4%. ☀️ We can't stop global warming: The new AI models predict our planet's future. 🎬 The "Brutalist" might not win an Oscar because of using AI to enhance the actors' playing. 🏴 Manchester City players will wear AI-designed kits. 🌍 "World Models": are Nvidia, Google, and others trying to give computers the ability to imagine? 🚘 Tesla is set to launch its first robotaxi service in June 2025 in Austin, Texas. 📱 LinkedIn co-founder Reid Hoffman explains why AI is far more useful than dangerous. 📱 Apple wants to integrate Starlink satellite connectivity into iPhones. 💳 Swedish fintech Klarna replaces nearly 1,000 employees with AI. 🍌 Gaussian splats: A new technology for realistic 3D modeling. ▎INTERVIEW 📱 OpenAI CPO Kevin Weil: On robots, AI agents, and accelerating progress. 🖥 Historian and futurist Yuval Noah Harari: Can AI actually create? Enjoy your Sunday 👍

🐳 "ChatGPT just lost its job to AI." How Did Chinese DeepSeek Blow Up the AI Industry? Who created the free ChatGPT competitor, is DeepSeek's arrival changing the AI race, and what's next? Let's get into the main topic of the week. ➡️ DeepSeek was founded by entrepreneur Liang Wenfeng in May 2023. A few years before, Wenfeng co-founded High-Flyer, a hedge fund that uses AI for stock trading. ➡️ The company's first AI models were better and cheaper than the other Chinese ones. At the end of 2024, DeepSeek entered the international competition with the V3 model, which showed results at the level of GPT-4o and Claude-3.5 Sonnet. We covered the V3 in detail here. ➡️ Soon, on January 20, the reasoning DeepSeek-R1 was released. It is as good as OpenAI o1 in most benchmarks and superior in some. Also, R1's API access is much cheaper. ➡️ By January 27, the DeepSeek app became the most popular app on iOS and Android in the U.S., displacing ChatGPT. Big Tech's stocks dropped down. ➡️ Nvidia suffered the most. Its shares fell by about 17%, its capitalization — by almost $600 billion. This was the largest collapse in a day for a single company. 💲 How Much Did the DeepSeek Models Cost? We don't know. The training of V3, which the R1 is based on, cost $5.6 million. This figure does not account for the expenses related to pre-launch activities or the compute cluster comprising 2,048 Nvidia H800 GPUs. 🔐 Security. Conspiracies. Censorship. Wiz discovered that DeepSeek's database, including some chat logs, was in the public domain. The backdoor was promptly covered up. Elon Musk and Scale AI CEO Alexandr Wang suspect that DeepSeek has more GPUs, but the company is hiding the real numbers because of U.S. export restrictions. Microsoft is investigating information about DeepSeek may have abused the use of the OpenAI API in training its model. However, there is no evidence of any of this yet. DeepSeek stores data in China, which has alarmed ordinary users and the U.S. authorities. The service is prone to censorship: DeepSeek's models often don't answer questions about China's domestic politics. On the other hand, DeepSeek is an open model: anyone can deploy it locally and without restrictions. For example, Perplexity AI-powered web search had integrated R1: without censorship or sending prompts to China.

📌 You can use DeepSeek-R1 with Perplexity search in our bot! 1️⃣ Go to @GPT4AgentsBot and type the command /deepseek 2️⃣ Click 🔘 Activate button 50 uses per day is included in the "Premium Max" plan.

🔜 What's Next? OpenAI released the new o3-mini model. "It's fast, cheap, smart, and designed to crush DeepSeek," —Wired writes about o3-mini. The real question now isn't whether Chinese startups can create a competitor to the o3-mini or a major version of the o3, but rather how quickly they will do it. So, whose AI is the best today? ❤️ — ChatGPT 🐳 — DeepSeek 🔥 — Other: Gemini, Claude…