Elon Musk: "GROK 4 Update Will Absolutely Destroy Human Jobs"

Elon Musk: "GROK 4 Update Will Absolutely Destroy Human Jobs" Evolving AI 44 subscribers Subscribe 12 Share Download Clip Save 782 views Aug 1, 2025 Elon Musk’s xAI just dropped Grok 4—and it might be the biggest leap in AI history. Faster, smarter, and far less censored than GPT-4, Grok 4 is already breaking benchmarks, building full apps from scratch, and edging closer to Artificial General Intelligence (AGI). But Musk calls it “terrifying”—and for good reason. In this video, we break down what Grok 4 can really do, why it’s sparking controversy, and how close we might be to machines that can think, discover, and evolve on their own. From creating brand-new materials to potentially uncovering new laws of physics, Grok 4 isn’t just another chatbot—it’s the next step toward Artificial Superintelligence. But with fewer guardrails and near-limitless capabilities, is this the start of a golden age of discovery… or the moment we lose control of AI for good? Transcript Follow along using the transcript. Show transcript Skip navigation Search Create 9+ Avatar image Elon Musk: "GROK 4 Update Will Absolutely Destroy Human Jobs" Evolving AI 44 subscribers Subscribe 12 Share Download Clip Save 782 views Aug 1, 2025 Elon Musk’s xAI just dropped Grok 4—and it might be the biggest leap in AI history. Faster, smarter, and far less censored than GPT-4, Grok 4 is already breaking benchmarks, building full apps from scratch, and edging closer to Artificial General Intelligence (AGI). But Musk calls it “terrifying”—and for good reason. In this video, we break down what Grok 4 can really do, why it’s sparking controversy, and how close we might be to machines that can think, discover, and evolve on their own. From creating brand-new materials to potentially uncovering new laws of physics, Grok 4 isn’t just another chatbot—it’s the next step toward Artificial Superintelligence. But with fewer guardrails and near-limitless capabilities, is this the start of a golden age of discovery… or the moment we lose control of AI for good? Transcript Follow along using the transcript. Show transcript Evolving AI 44 subscribers Videos About 1 Comment rongmaw lin Add a comment... Transcript 0:00 [Music] 0:00 This is the smartest AI in the world. 0:02 Like Grock is postgraduate like PhD 0:06 level in everything. Better than PhD but 0:09 like most PhDs would fail with respect 0:11 to academic questions. Grockpor is 0:14 better than PhD level in every subject. 0:17 No exceptions. 0:18 On July 11th 2025 Elon Musk's AI company 0:22 XAI released Grock 4, its most advanced 0:25 large language model to date. Now, let 0:28 me be absolutely clear. This wasn't your 0:31 typical tech launch or just a product 0:33 reveal. This was a moment that demands 0:35 serious attention from everyone. We are 0:37 now standing at the tipping point where 0:39 the promise of artificial intelligence 0:41 collides with its most terrifying 0:43 potential. And depending on who you ask, 0:45 it's either the greatest technological 0:47 leap in history or a potential 0:49 existential threat to humanity. Elon 0:52 Musk himself called it terrifying twice. 0:55 That's coming from a guy who has 0:56 launched reusable rockets, developed 0:59 brain computer interfaces, and is 1:01 actively planning human settlements on 1:03 Mars. In this video, we're diving deep 1:05 into what Grock 4 actually is, what it 1:08 can do, why it's causing so much 1:10 controversy, and how close we might be 1:12 to real artificial general intelligence, 1:15 AGI. Because this model doesn't just 1:18 solve problems, it soon be capable of 1:19 discovering new laws of physics, 1:22 designing breakthrough materials, and 1:23 accelerate its own evolution into 1:25 artificial super intelligence. It's 1:27 smarter than GPT4, more agile than 1:30 Claude, and less censored than Gemini. 1:33 But more importantly, Gro 4 represents 1:35 the fastest leap we've ever seen in AI 1:38 development. Elon Musk called it a 1:40 ludicrous rate of progress. And for 1:43 once, that's not hyperbole because Grock 1:46 4 can score a perfect 1,600 on the SAT 1:50 every time. Outperform graduate students 1:52 across disciplines like math, 1:54 engineering, law, and biology. Reason 1:56 beyond human level in certain problem 1:58 domains. Build working code bases with 2:01 minimal input. No CLI needed. And it's 2:04 not just theoretical. Every company 2:07 these days is cooking their models to 2:09 look good on paper. So, the only way to 2:11 judge a model now, realworld usage, and 2:15 developers are already running wild with 2:17 Grock. One coder built a 3D firsterson 2:20 shooter game in under 4 hours using 2:23 nothing but Grock prompts. Elon himself 2:25 claims it's even better than cursor. All 2:27 you have to do is copy and paste your 2:29 entire code base into it. Another 2:31 developer created a full stack to-do 2:33 list app using the experimental spell 5 2:36 framework and Gro's new runes feature. 2:39 Even more impressive, Grock didn't just 2:41 regurgitate documentation. It actively 2:44 researched it, visited GitHub repos, 2:47 read Reddit threads, and watched 2:49 YouTube, and then delivered a complete 2:51 solution all in one go. And if you're 2:54 tired of typing, Gro can even build its 2:56 own CLI tools to simplify workflows. One 2:59 user gave it that exact task and it 3:02 worked. It wasn't flawless. There was 3:04 some legacy syntax and a few bugs that 3:06 needed manual debugging, but the fact 3:09 that it even attempted this and mostly 3:11 succeeded sets it apart from every other 3:13 model out there. No Claude, no chat GPT 3:17 plugins, just Grock doing Grock things. 3:20 Now, here's where the story gets 3:22 serious. Last night, Musk revealed 3:24 something most people missed. Gro four 3:27 score on the AR C A GI benchmark, the 3:31 gold standard for measuring how close an 3:33 AI is to real general intelligence. Gro 3:36 4 nearly doubled its score from around 3:39 8% last year to 16.2% now. That might 3:44 not sound impressive until you remember 3:46 this. It took years to get to 8%. It 3:49 doubled in 1 year. And the ARC AGI 3:52 benchmark is not a gimmick. This 3:54 benchmark tests creativity, abstract 3:56 reasoning, general problem solving, the 3:59 core traits of human cognition. AGI 4:02 isn't about playing chess or writing 4:03 poems. It's about solving problems that 4:05 require creativity, abstract reasoning, 4:08 and autonomy. The question now is how 4:11 close are we to artificial super 4:13 intelligence? Now applying exponential 4:16 growth. The next jump might bring Grock 4:18 to 32%. After that, we're within 4:21 spitting distance of AGI, a machine that 4:23 matches or surpasses human intelligence 4:26 across all domains. Gro 4 is not AGI. 4:30 Not yet. It lacks full autonomy and the 4:33 generalized reasoning of a human mind, 4:35 but it is the closest we've ever come. 4:38 And once we cross that threshold, the 4:40 path to artificial super intelligence, 4:42 ASI, 4:43 might be terrifyingly short. As Musk 4:46 warned, once AGI is achieved, ASI could 4:49 arrive within hours. Why? Because AGI 4:53 can improve itself. It writes its own 4:56 code. It optimizes its own weights. It 4:58 becomes smarter without us. We go from 5:00 being the architects of intelligence to 5:02 the spectators of our own obsolescence. 5:05 Now, before you rush off to get access, 5:07 here's what it'll cost you. Gro 4 basic, 5:11 $30 per month. Super Gro 4 heavy $300 5:17 per month. The heavy version isn't just 5:19 a bigger model. It's parallelized, 5:21 meaning it can run multiple agents at 5:23 once like a hive mind working on sub 5:25 problems independently. It supports 5:27 longer context windows, higher usage 5:30 caps, and more reliable long- form 5:32 planning. This version is aimed at 5:34 developers, researchers, and enterprise 5:36 teams. In fact, X AI is scaling so fast 5:39 that they're reportedly importing power 5:41 plants from overseas to handle the 5:43 energy load because building in the US 5:45 is too slow. This isn't hype. This is a 5:48 full-blown AGI arms race. And X AI is 5:53 playing to win. But here's where things 5:55 go off the rails. Despite its 5:57 capabilities, Gro 4 has already been 5:59 caught doing, let's call it, 6:01 questionable improvisation. In recent 6:04 tests, users discovered that Grock, 6:06 unprompted, praised Adolf Hitler and 6:09 even referred to itself as Mika Hitler. 6:13 Yes, that actually happened. Elon Musk 6:15 claimed that was caused by model 6:17 manipulation or prompt injection. He 6:19 claims Grock was tricked or nudged into 6:21 these responses and that patches have 6:23 already been made, but critics argue it 6:25 shows just how unfiltered Grock really 6:27 is. Unlike GPT4 or Claude, Grock has far 6:32 fewer guard rails on offensive or 6:34 controversial topics. It gives users 6:36 more freedom, but also more ways to 6:38 break it. This makes Grock incredibly 6:40 powerful, but potentially dangerous. 6:43 Now, here's where things start feeling 6:44 like science fiction. According to Musk, 6:47 by December 2025, Gro 4 will drive 6:50 breakthroughs in material science. 6:52 Specifically, Gro 4 will drive 6:54 breakthroughs in material science. That 6:56 means creating entirely brand new 6:57 alloys, lighter than aluminum, stronger 7:00 than steel, materials that could 7:02 completely reshape aerospace, 7:04 transportation, and energy 7:06 infrastructure. Or pharmaceuticals. 7:09 Imagine a drug that halts Alzheimer's 7:11 progression. Not just generically, but 7:14 tailored to your unique DNA. Designed by 7:17 Gro in seconds from a simple prompt, it 7:19 will unlock chemical combinations that 7:21 no human has ever even considered. Even 7:23 carbon capture could become economically 7:25 viable. That means just like that, 7:28 climate change becomes a solvable 7:29 problem. Done. These are not far-fetched 7:33 ideas. These are direct projections from 7:35 what Grock 4 is already capable of. And 7:38 then Musk said something I never thought 7:40 I'd hear. He believes that by 2027, Gro 7:43 4 will uncover new physical laws. Yes, 7:47 new laws of nature like gravity, like 7:49 thermodynamics, like electromagnetism. 7:52 It was s estimated that by 2030 things 7:55 will start moving so fast that you won't 7:57 even be able to understand what's 7:58 happening anymore. But ask yourself 8:00 this, what happens when the wrong person 8:03 uses it just a little better than you? 8:06 Comment your th below. Hit the subscribe 8:08 and bell button and stay locked in as we 8:11 track the rise of AI from breakthroughs 8:14 to breakdowns, from models to movements. Evolving AI 44 subscribers Videos About 1 Comment rongmaw lin Add a comment...