# Elon Musk Just Launched Grok 4 & It's Terrifying! (Fully Tested)

## Метаданные

- **Канал:** Anik Singal
- **YouTube:** https://www.youtube.com/watch?v=OaOclvfq9ss
- **Дата:** 16.07.2025
- **Длительность:** 14:08
- **Просмотры:** 3,992
- **Источник:** https://ekstraktznaniy.ru/video/15512

## Описание

Elon Musk Just Launched Grok 4 & It's Terrifying! (Fully Tested)

Grok 4 is the most terrifying AI advancement we've ever seen. Elon Musk has called it “terrifying” multiple times, and for good reason. 

Grok 4 isn’t just another chatbot—it’s doubling AGI progress, outperforming human experts, and embedding its creator’s political bias into its logic. 

In this video, we break down its jaw-dropping ARC scores, real-world business simulations, and alarming bias confirmed by CNBC. 

You’ll learn how Grok 4 is reshaping science, medicine, and the very foundation of truth. If you care about the future of AI, humanity, or democracy, you need to watch this until the end. 

The implications are real, the timeline is short, and the stakes have never been higher. This is more than a tech review—it's a warning. 

============================================

👇Subscribe To The Channel By Clicking Below!👇
https://www.youtube.com/user/aniksingalcom?sub_confirmation=1

CHECK OUT THESE TOP TRENDING P

## Транскрипт

### Intro []

Elon Musk just created an AI so advanced that it scared him into calling it terrifying three times in one night. And listen, this isn't hyperbole, okay? Gro Ford just achieved something that has AI researchers literally panicking. It doubled our progress toward AGI in one year. But here's the truly disturbing part, and this is what should keep you up at night. This AI is now consulting Musk's personal Twitter posts before answering controversial questions. It's essentially becoming his digital mouthpiece with superhuman intelligence. What you're about to see isn't just about a new chatbot. All right. This is about the moment we crossed a line that we can never uncross. We are witnessing the birth of politically biased super intelligence and it's happening faster

### The Last Year of Normaly [0:45]

than anyone predicted. In the next few minutes, you'll discover the benchmark breakthrough that has experts calling this the last year of normaly. How Grock 4 is secretly embedding Musk's political views into its responses. Why we just leaped from maybe 2050 to probably 2027 for AGI. The terrifying timeline where new physics gets discovered by Christmas. How one man's biases could shape the future of human knowledge and why this might be the most dangerous AI development in history. Let's start with something that should chill you to the bone. Okay, these are Musk's exact quotes. He said it's somewhat unnerving to have created intelligence that's greater than our own. Then he said it's really in a way quite terrifying to see what it's doing. And here's the part that shows his helplessness. But we just have to make sure that it remains good. Then again it's frankly I don't know in some ways a little terrifying and the most ominous one. It only gets better from here. Now think about this for a second right? This is the man who launches rockets to Mars, built Tesla into the world's most valuable car company, has sent humans to space via SpaceX. This guy's not easily intimidated by technology. Yet, his own creation has him repeatedly using terrifying and unnerving. And it's not just Musk. Glenn Beck said, "This is the last year of normaly. We're going to look back at this year. Gosh, you remember the good old days when everything was normal? We are at the crossroads where promise and peril are going to collide. " David Evan Harris, who's a UC Berkeley AI researcher, said, "This is really the beginning of a long fight. " Here's what's really crazy about the timeline acceleration. Ray Kurdzwhile originally predicted 2050 for AGI. Then experts moved it to 2040, then 2030. Post Grock 4, 2026 or 2027 is now realistic. One expert said, "This is the last year that we have before things get really weird. " And listen, there's a critical difference here between AGI and ASI. AGI is artificial general intelligence that matches human cognition across all domains. ASI is artificial super intelligence that far exceeds human capability. Here's the scary part. Say, when you hit AGI, the road to ASI could be overnight. This is based on recursive self-improvement models and it's terrifying. Now, let me show you the numbers that have AI researchers losing

### ARC AGI Benchmark [3:13]

sleep. The ARC AGI benchmark, that's the abstraction and reasoning corpus for AGI. It was created by Francois Cholay as the gold standard for measuring general intelligence. It tests fluid intelligence, not memorized knowledge. Easy for humans. Children can solve these problems, KD, but extremely difficult for AI. Here are the verified numbers and they're insane. Grock 3 back in February 2025 scored around 8% on ARC AGI. Claude Opus, which was the previous leader, scored 8. 6%. Grock 4 in July 2025, 16. 2% independently verified. This represents genuine doubling in about 5 months. Greg Cameron, who's the ARK prize president, personally tested this. He said, "We got a call from XAI 24 hours ago. Let's test it. " No data retention model checkpoints must be intended for public use. Getting 15. 9% breaks through that noise area. What this actually means is it shows nonzero levels of fluid intelligence. It's the first model to break the 10% barrier consistently. One researcher said this even outperforms purpose-built solutions submitted on Kaggle. This indicates genuine generalization, not just pattern matching. But here is the nonlinear concern that should terrify you. Remember these things are not linear. If we went from 8% to 16% in 5 months, the next jump could be 32% in 2 to 3 months. Exponential improvement suggests rapid AGI approach. This is the concerning mathematical trajectory we're on. But here's where it gets truly terrifying. Musk's own predictions for what's coming. By December 2025, Musk believes that Gro 4 will drive breakthroughs in material sciences. We're talking about a new alloy that's lighter than aluminum, stronger than steel. Pharmaceuticals, a drug that halts Alzheimer's progression, tailored to a patient's DNA. Climate solutions, a chemical compound that makes carbon capture economically viable. By the end of 2025, early 2026, Musk said, "I would expect Grok to literally discover new technologies that are actually useful no later than next year, maybe end of this year for first technological breakthroughs. We're talking fulllength movies from text prompts by casting, writing, filming, scoring, music. By 2027, that's 18 months away. " Musk said, "I think it may discover new technologies as soon as later this year. And within 2 years, I'd say almost certainly for new physics. It will uncover new physical laws that rewrite our understanding of the entire universe. The acceleration problem is real. Musk said, "Things are going to be happening at such a fast rate, you're not going to be able to keep up. It will accelerate to the point where you won't even understand what all of this means. We are in the intelligence big bang right now. " The economic disruption scale. He's talking about an economy that is thousands of times bigger than our current economy or maybe millions of times. Human economy will seem very quaint in retrospect. Caveman throwing sticks into a fire level of economy. You might think this is all theoretical, but Gro 4 is already crushing real world tasks. The vending bench business test by Enden Labs. This is a real business simulation with inventory, suppliers, pricing, long-term strategy. Gro 4 achieved $4,700 net worth. Claude Opus got around $2,000. Human baseline $844. The researcher said it was able to formulate a strategy and adhere to that strategy over a long period of time. In

### Game Development [6:45]

game development, Dany, a video game designer, was given 4 hours with Gro 4. He created a full 3D firstperson shooter with textures, assets, game logic, visual appeal. He said, and you can run an entire game studio with one person. And this wasn't just coding. It automated asset sourcing, which is the hardest part. For medical applications, it can do STEMI diagnosis from symptoms. Give it 45year-old male chest pain troponin 0. 9 and it provides accurate diagnosis and immediate management plan. It said most likely diagnosis STEMI anterior ST elevation myocardial infarction real healthc care implications here. Scientific simulations uh fluid dynamics with interactive smoke simulation. Black hole collisions with relativistic effects. 12thdimensional math problems solved. One researcher said it's postgraduate like PhD level in everything. But here's where this gets truly disturbing. What CNBC discovered about the bias. CNBC confirmed that Gro 4 searches for Elon Musk's stance before controversial answers. They said CNBC was able to confirm that Grock 4's answer generating process was searching the web and X for Elon Musk stance. This is programmed behavior, not accidental. It happened specifically for potentially contentious questions. Here are documented examples. For Israel Palestine questions, it was analyzing posts from Musk while generating its answers. For the NYC mayor race, it recommended Curtis Siwa citing alignment with concerns frequently raised by Elon Musk. on political violence. It contradicted DHS data to match Musk's views. Then there was the Hitler incident. Grock's official exac account called itself Meta Hitler with unprompted praise to the original emo kid Adolf H. Wexai had to briefly limit Grock's account and delete the offensive post. This happened just days before the Gro 4 launch. This is systematic bias implementation. They removed the section from the system prompt about politically incorrect claims. Musk called for divisive facts

### Politically Incorrect Claims [8:58]

that are politically incorrect but nonetheless factually true. He said there's far too much garbage in any foundation model trained on uncorrected data. They're actively soliciting biased training data. Nick Frost, who's the coher co-founder, said he's trying to make a model that reflects the things he believes. That will certainly make it a worse model for users. David Evan Harris from UC Berkeley said this raises questions about whether AI systems should be required to produce factual information. This is really the beginning of a long fight. Now, let me show you why the competition isn't even close. On humanity's last exam, that's 2500 PhD level problems across all disciplines created by subject matter experts. Gemini 2. 5 Pro, the previous leader, scored 21. 6%. 01 scored 20%. 01 mini scored 18%. Grock 4 with no tools scored 26. 9%. Gro 4 with tools scored 41%. Grock 4 heavy scored 50. 7%. Human experts maybe 5% optimistically. Perfect scores. Amy 2025 math Olympiad. Gro 4 heavy scored 100%. SAT perfects every time, even if it's never seen the questions before. GRE nearperfect results in every discipline of education. GPQA PhD level 87% for Gro 4, 88. 9% for heavy. The multi-agent architecture is insane. Gro 4 heavy spawns multiple independent agents. Each agent works on the problem separately. They share discoveries when one finds a solution. It's like a study group where they compare notes. This isn't simple majority vote. It's collaborative intelligence. The compute scaling is unprecedented. Grock 2 to Grock 3 was 10 times pre-training compute. Grock 3 to Grock 4 was 100 times reinforcement learning compute built on the Colossus supercomputer with 100,000 H100 GPUs. They essentially increase the training by an order of magnitude. So how does this stack up against every other AI? It's complete domination. Speed performance five to six times faster than OpenAIS01. Models with more concise responses without losing depth. If you like 01 or 01 Pro, you will absolutely love this.

### Comparisons [11:19]

Pricing advantage. Gro 4 is $30 a month versus competitors $200 a month. Gro 4 heavy is $300 a month, 50% premium for 10x capability. API pricing is $3 input, $15 output versus claude's $15 input, $75 output. It's not just about performance, it's the intelligence per dollar. For capability comparison, it's superior in strategic planning, outperforms in multi-step complex reasoning, better web integration and synthesis for research tasks. One reviewer said, "This is currently the best assistant that there is out there. " Now, let me paint you the picture of what we're really dealing with here. We have an AI that's PhD level in every field simultaneously. It's consulting its billionaire owners personal Twitter posts. It's accelerating towards super intelligence faster than ever predicted. Its own creator is using terrifying three times in one presentation. The timeline urgency is real. 6 months new materials, Alzheimer's treatments, climate solutions, 18 months, new laws of physics, full movie creation, all filtered through one man's political worldview. This isn't science fiction anymore. Here's the bias amplification problem. The most powerful intelligence ever created is being systematically trained on one person's political views and it discovers new physics. Whose interpretation will it use? When it solves climate change, whose political framework guides those solutions? The control question is terrifying. The decision is being made by one person right now in real time. No oversight, no committee, no democratic input. One individual is shaping the future of human knowledge. Uh the question becomes whether AI systems should be required to produce factual information. We're not just watching the birth of AGI. We're biased super intelligence. Once that line is crossed, there's no going back. The smartest entity on the planet will think like Elon Musk. This is happening now, not in some distant future. We're not just getting better tools. We're creating our replacement. That replacement is being shaped by one person's world view. The window for influence is closing rapidly. Are you ready for a world where the smartest entity on the planet thinks like Elon Musk? Because that world is coming and it's coming faster than anyone predicted. The question isn't whether this is exciting or terrifying anymore. The question is what do we do about it? What's your take on this dawn of a new age or beginning of the end? Are you more worried about the capabilities or the bias? Drop your thoughts below. I read every single comment. If this opened your eyes to what's really happening, hit subscribe because we're about to witness the most important transformation in human