# The Most INSANE AI News This Week! 🤯

## Метаданные

- **Канал:** Julian Goldie SEO
- **YouTube:** https://www.youtube.com/watch?v=QrKymGUdZ1s
- **Дата:** 08.03.2025
- **Длительность:** 22:50
- **Просмотры:** 6,377

## Описание

🚀 Get a FREE SEO strategy Session + Discount Now: https://go.juliangoldie.com/strategy-session 

Want to get more customers, make more profit & save 100s of hours with AI? Join me in the AI Profit Boardroom: https://go.juliangoldie.com/ai-profit-boardroom

🤯  Want more money, traffic and sales from SEO? Join the SEO Elite Circle👇
https://go.juliangoldie.com/register

🤖 Need AI Automation Services? Book an AI Discovery Session Here: https://juliangoldieaiautomation.com/

Click below for FREE access to  ✅ 50 FREE AI SEO TOOLS 🔥 200+ AI SEO Prompts! 📈 FREE AI SEO COMMUNITY with 2,000 SEOs ! 🚀 Free AI SEO Course 🏆 Plus TODAY's Video NOTES...
https://go.juliangoldie.com/chat-gpt-prompts

- Join our FREE AI SEO Accelerator here: https://www.facebook.com/groups/aiseomastermind
- Need consulting? Book a call with us here: https://link.juliangoldie.com/widget/bookings/seo-gameplanesov12

In this video, I will cover major AI developments including Deep Seek R2, OpenAI's agents, and other breakthroughs:

0:00 - Intro: Multiple AI advancements causing market disruption
0:41 - Deep Seek R2 release coming sooner than expected with improved capabilities
1:39 - Deep Seek's efficiency approach using less computing power
2:49 - Global impact of Deep Seek, including market disruption and geopolitical concerns
4:25 - Expected features and launch timeline for Deep Seek R2
5:15 - OpenAI's tiered AI agents ($2K-$20K monthly) and revenue projections
6:35 - AI agent capabilities including coding, troubleshooting, and development assistance
8:44 - Job replacement concerns, especially for entry-level positions
9:50 - Risk to OpenAI from more affordable competitors
10:17 - Alibaba's Qwen QWQ-32B model matching larger models with fewer parameters
12:42 - Mistral's new OCR system for document understanding and data extraction
16:33 - Wundersphere's Wave 4 update with improved coding features
17:11 - Amazon's upcoming Nova AI model with reasoning capabilities
17:57 - ChatGPT's new coding update for macOS with direct IDE integration
20:57 - Advice for preparing for AI advancement and promotional offers

## Содержание

### [0:00](https://www.youtube.com/watch?v=QrKymGUdZ1s) Intro: Multiple AI advancements causing market disruption

the most insane AI news this week the AI world just exploded this week deep seek R2 is coming way sooner than anyone expected and might crash the $1 trillion AI Market open ai's new $220,000 AI agents will replace entire teams of workers Alibaba dropped a model that matches top AI with just 5% of the parameters and mrr's new tech reads everything on a document even math equations hey if we haven't met already I'm the digital Avatar of Julian gold CEO of SEO agency Goldie agency whilst he's helping clients get more leads and customers I'm here to help you get the latest AI updates let's dive in before these companies change the world again

### [0:41](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=41s) Deep Seek R2 release coming sooner than expected with improved capabilities

we'll start with the Deep seek R2 release and trust me this is huge news for the AI world this Chinese AI model might just flip the entire industry upside down is coming way sooner than expected and the price is so cheap it's making big tech companies Panic according to Reuters deep seek was planning to release their R2 model in early may but now they want it out as soon as possible maybe even in the next few weeks this is a big deal because the first deep seek model R1 literally caused a one trillion dollar selloff in the global stock market when it launched that's trillion with a t it scared investors that much why because deep seek R1 showed that a small Chinese startup could make an AI model almost as good as the big American companies but at a fraction of the cost and R2 is going to be even better the R2 model will have better coding skills and can reason in more languages Beyond just English and Chinese they want to add Spanish and other languages to make it more accessible worldwide but what makes

### [1:39](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=99s) Deep Seek's efficiency approach using less computing power

deep seek so special here's what's crazy they're doing more with less the R1 model was built with less powerful envidia chips than what companies like open Ai and anthropic use but it's still performed almost as well as models that cost billions to make deep seek only needs to improve by 4. 5 points on the benchmark test to become the best AI model in the world and R2 might just do that plus the pricing is insane deep seek R1 was 20 to 40 times cheaper than what open AI charges for similar models that forced open AI to lower their own prices just to stay competitive so how did a Chinese startup manage to compete with trillion doll American tech companies there are a few key factors first they focus on efficiency deep seek created something called the mixture of experts architecture this means the AI only activates the parts of the model it needs for each task which saves a ton of computing power second invested heavily in Computing infrastructure their parent company spent 1. 2 billion low about 185 million on two supercomputing clusters back in 2020 and 2021 one of these clusters called Firefly 2 uses around 10,000 Nvidia a800 chips for training AI models third they're constantly

### [2:49](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=169s) Global impact of Deep Seek, including market disruption and geopolitical concerns

innovating last week they hosted an open source week where they released five new tools flat Malay a GPU optimized decoding kernel for faster text generation deep a communication library for models like V3 and R1 deep gem a library for efficient Matrix multiplications all of these tools make AI development faster cheaper and more efficient and they're all hints at what's coming with the R2 model here's why this matters this isn't just Tech news this is world changing stuff if deep seek R2 is as good as experts think it will be it could completely disrupt the AI industry we're talking about an open source model that might perform as well as open AI GPT 40 or Claude 3. 7 on it but at a tiny fraction of the cost that means smaller companies can now afford to use Cutting Edge AI developers can build on top of it without expensive API fees it could force American companies to slash their prices even more this is democratizing AI in a way we haven't seen before it's taking power away from a few dominant players and spreading it around the world now let's talk about the global impact the release of deep seek R1 caused such a panic that some US senators have proposed jail terms of up to 20 years for Americans who download certain Chinese AI models that shows you how seriously they're taking this threat there's also been controversy about how deep seek got so many Advanced chips despite us export restrictions to China some Western companies claim deep seek might have 50,000 high-end Nvidia chips that are banned for export to China whether that's true or not what matters is that deep seek has found ways to do more with less they're innovating out of necessity and that's making them more efficient than the competition so what can we

### [4:25](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=265s) Expected features and launch timeline for Deep Seek R2

expect from Deep C Car 2 when it launches based on everything we know it will likely be on par with or better than open ai's GPT 4 row or Claude 3. 7 Sonet it will be dramatically cheaper possibly 2040 times cheaper than competing models it will support more languages it will have improved coding abilities it might even be more efficient than R1 thanks to their new tools the craziest part with the new tools they've released during open source week deep seek R2 might be even cheaper than R1 was that would be absolutely gamechanging when will it launch based on the ruers report deep seek wants to launch R2 as early as possible some experts think it might come after the Ching Ming Festival in China which ends on April 6th others think it might launch in mid April either way we're likely talking weeks not months also

### [5:15](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=315s) OpenAI's tiered AI agents ($2K-$20K monthly) and revenue projections

open ai's new $20,000 AI agents are about to change everything according to a leaked report by the information open AI is building PhD level AI that works 24/7 without breaks companies are betting billions on these AI workers replacing humans the information just reported on open AI plans to launch three tiers of AI agents and the prices are mind-blowing we're talking $2,000 a month for entry-level knowledge workers $10,000 for developer agents and whopping $20,000 for PhD level research agents let me break down what this means and why it matters to you right now open ai's Revenue was $1 billion in 2023 jumped to $3. 7 billion in 2024 and they're forecasting $12. 7 billion in 2025 that's not just growth that's explosion and the craziest part a huge chunk of that new money is coming from these AI agents in the leaked projection charts open AI showed three colorcoded revenue streams the red is chat GPT the blue is their API but there's a new green bar that's almost as big as their entire 2024 Revenue that green bar is all from these new AI agents open AI expects these agents to make up about 20 25% of their total revenue that's billions of dollars they think companies will spend on AI workers and they're not loan soft Bank Oracle And even the US government are betting big on this through the Stargate initiative now what

### [6:35](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=395s) AI agent capabilities including coding, troubleshooting, and development assistance

can these AI agents actually do these AI coding assistants can do way more than just write code now they can access your command line install packages manage your development environment and even troubleshoot issues on their own when they hit an error they don't give up they keep trying different approaches they can read documentation clone GitHub repositories make changes to code run tests and fix bugs they handle all the boring stuff like figuring out why dependency aren't working or why you're getting error messages what's impressive is how they can take over the entire workflow they don't just suggest code they set up the entire project environment install everything needed and make sure it's all working together the human only needs to step in for things the AI can't access like providing API keys or confirming what's happening on screen and yes they can absolutely help build million dooll apps and games these systems can write complex code design interfaces Implement features and fix bugs at a pace humans simply can't match they're already helping developers build valuable software products that would have taken much longer with human only teams think about what this means right now we have ai that can write code that's the basic chat models we've had work inside your development environment like cursor does run command line instructions and install packages debug and troubleshoot when things don't work the only thing missing is the ability to fully see and interact with your screen like a human but both open Ai and anthropic already have this technology open AI calls it operator and anthropic calls it computer use when you combine all these pieces you get something that can do everything a human programmer can do and that's exactly what these $10,000 a month developer agents will be now can they replace a senior developer with 10 years of experience probably not yet but could they replace entry-level developers or people just learning to code absolutely think about it this way on a scale from 1 to 100 where one is someone just starting to Learn Python and 100 is the best programmer in the world these AI agents might already be at level five or 10 that means they can already do more than someone just learning and they're getting better fast this is where things

### [8:44](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=524s) Job replacement concerns, especially for entry-level positions

get scary for some jobs if a company can pay $10,000 a month for an AI that works 24/7 never gets tired and constantly improves with updates why would they hire an entry-level developer at $8,000 a month plus benefits but here's a really big question is 2025 too early for this open AI thinks they can triple their revenue this year partly because of these agents that timeline seems ambitious but these companies move faster than anyone expects and that brings me to the most important part what this means for you and your job there are two numbers you need to watch what percentage of workers could these AI agents replace if it's 5% that's significant but manageable if it's 70% we're looking at massive disruption how much cheaper are these AI agents compared to human workers if they're only 5% cheaper adoption will be slow companies will weigh the risks carefully but if they're 95% cheaper that creates a stampede to replace workers as fast as possible the ideal scenario is slow adoption that gives us time to adjust the nightmare scenario is rapid replacement that creates massive unemployment almost overnight now here's

### [9:50](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=590s) Risk to OpenAI from more affordable competitors

the danger for open AI they're setting themselves up to be disrupted if a company like deep seat creates an open source agent that's 80% is good but costs 50 times less that's what most people will use the price will crash and this isn't just about developers these agents are targeting knowledge workers researchers and office administrators too the $22,000 tier is aimed at replacing High income knowledge workers the $20,000 tier is for PhD level research more breaking news this

### [10:17](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=617s) Alibaba's Qwen QWQ-32B model matching larger models with fewer parameters

time from quen which is run by Alibaba they've just released their qwq 32b model which is their new reasoning model with 32 billion parameters this is rivaling other models like deep seek 1 and it's completely open source so you can host it locally The Benchmark comparisons are super impressive KW crew 32b is performing well against deep seek R1 open ai1 mini and deep seek R1 distilled on various tests it's doing particularly well on the aim test falling just behind R1 on Live code bench but outperforming deep SE car1 and open AI 01 mini on live bench it also leads on BFC test which are basically for function calling if you want to get access the easiest way is to go to quen's website you can also get it on hugging face and it's available onma to run locally just download Olma go to qwq get the 32b model and run it inside the terminal command is simple run qwq the only thing I'll say about running it locally is that it's super slow even on a Mac M3 Pro instead you can navigate to chat. and select qwq 32b preview from the drop down in the top left this is their reasoning and coding model what's really cool is you can switch on thinking mode and use qbq inside there along with web search and artifacts this is super useful because combining qwq with artifacts lets you start doing some serious coding tasks one thing to note is that quen and also deep C car1 has been a little unreliable recently when I've tested it so this is a good alternative plus it's a free model with open AI 01 preview and tools like o03 mini and 03 preview you have to pay unless you just want to keep hitting the limits qwq is a free alternative the cool thing about qwq compared to deeps car1 is that you can connect the reasoning model to the internet without having to pay for a tool like perplexity to do that with deeps car1 the search feature often breaks directly in the chat and you usually have to use perplexity to get it working to recap alibaba's qwq 32b model is more efficient deep C car1 uses 671 billion parameters while qwq uses just 32 billion parameters it uses multi-stage RL learning which is pretty crazy it's a lot smaller but matches the same performance in logic math and coding it's open source and free for commercial and research uses under the Apache 2. 0 license and it has a higher context window with 131,000 tokens versus deep seek r1's 128,000

### [12:42](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=762s) Mistral's new OCR system for document understanding and data extraction

tokens mrol also just dropped something huge they're calling it the world's best document understanding API throughout history advancements in information retrieval have driven human progress from hieroglyphs to the printing press to digitization each leap has made human knowledge more accessible today we're at the edge of the next big leap unlocking the collective intelligence of all digitized information about 90% of the world's organizational data is stored as documents and to tap into this potential mistr has introduced mistr OCR this is an optical character recognition API that sets a new standard in document understanding unlike other models mistal OCR understands each element of documents media text tables equations with Incredible accuracy it takes images and PDFs as input and extracts content in an ordered inter Le text and images format this makes mistr OCR ideal for use with a rag system taking multimodal documents like slides or complex PDFs as input they've made Mistral OCR the default model for document understanding across millions of users on leche and are releasing the API Mistral OCR latest at 1,000 pages per dollar and approximately double the pages per dollar with batch inference the API is available today on their developer site La platform and coming soon to their cloud and inference Partners as well as on premises here are the highlights state-of-the-art understanding of complex documents natively multilingual and multimodal top tier benchmarks fastest in its category Dockers prompt structured output selectively available to self host for organizations dealing with sensitive information Mistral OCR excels in understanding complex document elements including interleafed imagery mathematical Expressions tables and Advanced layouts such as latex formatting the model enables deeper understanding of Rich documents such as scientific papers with charts graphs equations and figures in Benchmark tests Mistral OCR consistently outperforms other leading OCR models it extracts embedded images from documents along with text which other llms don't do on their internal text only test set containing various publication papers and PDFs from the web mistra OCR scored 94. 89% overall compared to Google Document AI at 83. 4 2% Azure OCR at 89. 5 2% and GPT 40 at 9. 77% what's really impressive is that mistal OCR is natively multilingual it can pass understand and transcribe thousands of scripts fonts and languages across all continents this is crucial for both Global organizations and local businesses in benchmarks by language Mistral OCR scored 99. 09% for Russian 99. 20% for French 9755 for Hindi and 97. 1% for Chinese outperforming Azure OCR Google doc Ai and Gemini in all cases and mrol OCR is fast being lighter weight than most models in the category it performs significantly faster processing up to 2,000 pages per minute on a single node the use cases for this are massive digitizing scientific research leading research institutions are using Mistral OCR to convert scientific papers into AI ready formats accelerating scientific workflows preserving historical documents organizations are using it to digitize historical documents and artifacts streamlining customer service departments are exploring Mist OCR to transform documentation and manuals into index knowledge making literature across design education legal Etc AI ready mistal OCR is helping companies convert technical literature engineering drawings lecture notes presentations regulatory filings and more into IND indexed answer ready formats additionally wind surf just

### [16:33](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=993s) Wundersphere's Wave 4 update with improved coding features

released their Wave 4 update and it's packed with new features that make coding with AI much smoother they've added previews that let you iterate on your apps rapidly by easily sending elements and console errors back to Cascade as context they've added tab to import linter integration suggested actions and more what's cool is the previews feature solves the problem of having to manually explain what looks wrong or copying console errors instead it adds listeners and ux so that Cascade can be aware of exactly how you want to iterate you can point and click on the component you want to make changes to and there's a button to automatically pull console errors meanwhile Amazon is

### [17:11](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=1031s) Amazon's upcoming Nova AI model with reasoning capabilities

planning to launch its own AI model with Advanced reasoning capabilities according to Business Insider the offering is tentatively scheduled to launch by June under the Nova brand a group of generative AI models Amazon unveiled late last year Amazon wants the new model to take a hybrid reasoning approach that provides quick answers and more complex extended thinking within a single system reasoning models have recently become the next Frontier in AI they often work more slowly but can also tackle tougher Problems by trying multiple Solutions and backtracking Via Chain of Thought techniques companies including Google open Ai and anthropic have released their own reasoning models recently while deep seek drew a lot of attention for building a similar offering more efficiently one of Amazon's priorities is to make its no reasoning model more price efficient than competitors chat GPT has just

### [17:57](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=1077s) ChatGPT's new coding update for macOS with direct IDE integration

released a new AI coding update for Mac OS that can now edit directly in Ides like Visual Studio code or wind surf this is available to plus pro and team users to use it you need to download the chat GPT app to your desktop then open whatever IDE you're using like Visual Studio code which is free to download with both the chat GPT app and your code editor open you can click a button in chat GPT that lets you run with Visual Studio code it integrates with vs codee directly and you can also use it with windsurf when you've got your code open you can go back to chat GPT and edit that code for example you could say make this look more modern and chat GPT will attach your code and let you edit it right there you can select your code inside chat GPT give it instructions like make this look more modern and then click Auto apply code to automatically make the changes happen instantly in your code editor setting it up is super easy just open the chat GPT Chat bar by pressing option and space or by clicking the chat GPT menu bar icon you can also click the work with apps icon to manually work with apps from the chat GPT window it doesn't just work with coding it also works with text editing apps like Apple notes notion text edit and more when you're making code edits you don't need to copy and paste anymore you just ask chat GPT to edit open files directly when you ask for an edit chat GPT generates a diff that you can review and apply and there's an option to automatically apply edits as well one of the coolest features is that you can edit code in terminal use in advanced voice mode to do this you click the wave icon in the chat bar in the main window when you're in the advanced voice session the work with apps button gives you control of the app you're working with is this better than what was already out there honestly I think Claude 37 is still better for coding but if you're on a budget chat GPT might be better because you're using the app version which doesn't cost API credits you can switch between different models in chat gpt2 for example chat GPT 4. 5 usually feels more human it's not great for writing Big Blocks of content but for more creative tasks where it's just a few lines is quite easy to use if you're using AI tools for your business or personal projects this is huge news it means the cost of using Advanced AI will drop dramatically you'll have more options Beyond just open AI anthropic and Google open source models will continue to improve rapidly the tools you use every day will get better and cheaper this is especially important if you're trying to use AI to Gra your business the lower the costs the more you can experiment and integrate AI into your workflows the race to build the best AI models is heating up and these developments could be major turning points they show that you don't need billions of dollars or unlimited resources to compete at the highest level of AI development this is good news for everyone except maybe the big tech companies who have been charging premium prices for their AI models competition drives Innovation and lowest prices and that's exactly what we're seeing so what can you do to prepare

### [20:57](https://www.youtube.com/watch?v=QrKymGUdZ1s&t=1257s) Advice for preparing for AI advancement and promotional offers

first start using these tools yourself the people who know how to work with AI will replace the people who don't second focus on skills that AI still struggles with creative problem solving human interaction and strategic thinking thanks so much for watching if you want to save hundreds of hours with AI and make more money with AI to scale your business check out the AI profit boardroom the link is in the comments and description this comes with a community of over 268 members all interested in scaling their business with are just like you on top of that you also get access to our classroom with all of my best automations and courses on email and content automation social media video automations AI agents and workflows web scraping AI CEO automations and more you also get weekly Q& A calls where you can jump on and ask any questions you have Additionally you can just post in the community anytime you get stuck and we're constantly updating this with new content every single day if you want to get a free one to1 SEO strategy session that shows you how we take websites from 0 to 145,000 visitors per month and generate hundreds of thousands of dollars in sales feel free to grab that link in the comments and description on this free SEO acceleration session you'll get a custom tailored link building plan so you can generate more leads sales and profits from your website you'll be able to ask our worldclass experts any questions you have one to one and you'll learn the best link building strategy custom tailored for your website Julian reads every comment so make sure to drop your thoughts below what do you think about all these AI updates are you excited concerned let us know and as always if you enjoyed this video smash that like button subscribe for more AI updates and hit the notification Bell so you don't miss out on all the latest AI news thanks for watching and I'll see you in the next one

---
*Источник: https://ekstraktznaniy.ru/video/6753*