- 23
- 8 313
Unrealtech - IT, BigTech, Chips, EV
เข้าร่วมเมื่อ 9 ม.ค. 2025
The latest Engineering/Tech/IT/New Technologies Shared by Dr. Error
Evidence of DeepSeek’s Data Theft… But Still a Partner? | Microsoft’s Stance
The Truth Behind DeepSeek’s Data Theft Allegations-Innovation or Infringement?
Recent allegations of data theft have set the deep learning industry abuzz. Did DeepSeek really use OpenAI’s data without permission? As mounting evidence suggests unauthorized usage, Microsoft has announced its collaboration with DeepSeek and pledged support for the DeepSeek-R1 model.
While OpenAI and Microsoft have confirmed that they are investigating DeepSeek’s alleged data theft, they are simultaneously recognizing DeepSeek’s technological capabilities and engaging in partnerships. So, is DeepSeek a data-stealing company or an innovative partner? And what does Microsoft’s seemingly contradictory stance really mean?
In this video, we take a deep dive into the full story of the DeepSeek data theft scandal, Microsoft’s decision to collaborate, and OpenAI’s response. Let’s uncover the truth behind this hot topic in the AI industry!
#Microsoft #DeepSeek #DataTheft #distillation
Written by Error
Edited by Jin Lee
AI dubbing ▶ Perso : perso.ai
#ESTsoft #Perso
📧 unrealtech2021@gmail.com
Recent allegations of data theft have set the deep learning industry abuzz. Did DeepSeek really use OpenAI’s data without permission? As mounting evidence suggests unauthorized usage, Microsoft has announced its collaboration with DeepSeek and pledged support for the DeepSeek-R1 model.
While OpenAI and Microsoft have confirmed that they are investigating DeepSeek’s alleged data theft, they are simultaneously recognizing DeepSeek’s technological capabilities and engaging in partnerships. So, is DeepSeek a data-stealing company or an innovative partner? And what does Microsoft’s seemingly contradictory stance really mean?
In this video, we take a deep dive into the full story of the DeepSeek data theft scandal, Microsoft’s decision to collaborate, and OpenAI’s response. Let’s uncover the truth behind this hot topic in the AI industry!
#Microsoft #DeepSeek #DataTheft #distillation
Written by Error
Edited by Jin Lee
AI dubbing ▶ Perso : perso.ai
#ESTsoft #Perso
📧 unrealtech2021@gmail.com
มุมมอง: 101
วีดีโอ
NVIDIA Already Knew… DeepSeek’s Low-Cost Optimization, Microsoft Pushes FP4! Blackwell is Ready!
มุมมอง 19220 ชั่วโมงที่ผ่านมา
DeepSeek’s FP8 revolutionized AI training by drastically reducing costs, but now Microsoft takes it to the next level with FP4. FP4 cuts memory usage by more than half and enables up to 5x faster computation on next-gen GPUs. NVIDIA has already prepared for this shift, and with its Blackwell GPUs, it is set to dominate the AI hardware market. This breakthrough will further reduce AI training co...
DeepSeek Shock: AI Leaders’ Thoughts and Forecasts Reveal the Answer
มุมมอง 271วันที่ผ่านมา
Andwelgonghak has extensively covered the DeepSeek phenomenon, but what do actual AI experts think? 🤔 In this video, we analyze insights from AI leaders such as Pat Gelsinger, Andrew Ng, the CEO of Perplexity, and NVIDIA engineers, based on their discussions on X. We explore diverse perspectives-whether DeepSeek is a game-changer for AI scaling, a revelation of reinforcement learning’s limitati...
DeepSeek R1’s Low Cost? The Truth Behind the Claims and Real Costs!
มุมมอง 974วันที่ผ่านมา
Is DeepSeek R1 really that cheap? The real debate is about V3, not R1. Cost comparisons often ignore key details, like H800-based calculations. R1 also requires a base model and extra reinforcement learning. These facts are rarely mentioned. We analyzed the research to uncover the truth. Plus, Google’s Gemini 2.0 Flash Thinking is even cheaper. Let’s break it all down. #DeepSeek #V3 #R1 #LowCos...
DeepSeek Personal Data Stored on Chinese Servers? Fact Check
มุมมอง 19วันที่ผ่านมา
#deepseek #china #ai #openai #data #privacy #chatGPT #shorts
DeepSeek’s Surprise Open-Source Release! 🚀 The Ultimate Image AI That Surpasses OpenAI’s DALL-E 3
มุมมอง 2672 วันที่ผ่านมา
Amidst the global shock caused by DeepSeek-R1, DeepSeek is making waves, seizing the moment with Janus-Pro, a cutting-edge multimodal AI model for understanding and image generation. Unlike traditional AI models, Janus-Pro optimizes performance by separating image understanding and generation processes. With enhanced training strategies and an expanded dataset, it outperforms even large-scale m...
DeepSeek Personal Data, Device Info, Phone Numbers Leaked & Stored on Chinese Servers? Fact Check
มุมมอง 1812 วันที่ผ่านมา
Using DeepSeek services may expose personal data, usage patterns, and even keyboard inputs. Many people use it for convenience, but the security risks are serious. In fact, not just DeepSeek-major AI companies also utilize personal data to some extent. Upon reviewing DeepSeek’s privacy policy, it appears that data is stored on Chinese servers, raising concerns. However, AI models running locall...
Open-Source ATTACK | DeepSeek-R1reduced model retains performance even after an 80% size reduction
มุมมอง 1.2K2 วันที่ผ่านมา
DeepSeek’s AI model is still massive, but the open-source community has already found a way to solve this problem. DeepSeek-R1, which delivers performance similar to OpenAI’s O1, originally had a size of 720GB. However, the Unsloth team successfully reduced it to just 131GB. This isn’t just about shrinking the size-it has been optimized to maintain performance while being lightweight and fast. ...
Was DeepSeek’s Reinforcement Learning-Based Performance Improvement Also Used by OpenAI?
มุมมอง 3323 วันที่ผ่านมา
We have confirmed that DeepSeek R1’s reinforcement learning-based model shares significant similarities with what OpenAI’s Dr. Hyung Won Chung presented at last year’s MIT seminar. A comparison between DeepSeek’s operating principles and Dr. Chung’s presentation reveals notable commonalities. This raises the question of whether DeepSeek replicated OpenAI’s technology or if this is simply a natu...
You’re Getting It Wrong… AI Big Tech is Collapsing? China’s DeepSeek Controversy | Paper Analysis
มุมมอง 2.6K4 วันที่ผ่านมา
There’s a lot of buzz around China’s DeepSeek R1, but few have actually analyzed the core content of its research paper. In this video, we break down the key aspects of the DeepSeek R1 paper, examining its real performance and limitations. We also explore the potential impact of this technology on the AI industry and its future outlook. The relationship between open-source ecosystems, global AI...
How did LCD rule displays for 20 years? From beating CRTs to evolving into mini LED and QLED
มุมมอง 1.2K4 วันที่ผ่านมา
How did we go from the bulky CRT monitors of the past to today’s sleek and slim LCDs? #CRT #LCD #Display #Monitor #QLEDTV Written by Error Edited by jhLee unrealtech2021@gmail.com
Why IonQ and ion traps in quantum computing? | Quantum error correction principles | IonQ vs Rigetti
มุมมอง 396 วันที่ผ่านมา
The quantum computing revolution aims to surpass the limits of classical computers! Among the key players, IonQ stands out with its ion-trap technology. But how does it compare to other approaches, such as superconducting qubits used by Google, IBM, and Rigetti, or photonic qubits? We’ll dive into the principles, advantages, and challenges of the ion-trap method. From the basics of quantum comp...
Transformers 2.0? Google unveils 'Titans,' a new AI model that overcomes previous shortcomings
มุมมอง 6937 วันที่ผ่านมา
This video introduces Google’s next-generation AI model, ‘Titans.’ Inspired by the human memory system, ‘Titans’ presents a groundbreaking approach beyond Transformers, enabling AI to handle complex tasks with fewer computational resources. It signals a shift in the AI chip competition and paves the way for a new paradigm in AI technology. The video also delves into a concrete vision for on-dev...
The Principle Behind Video Compression That Shrinks 99.9%
มุมมอง 4439 วันที่ผ่านมา
We live in an era where videos are essential anytime, anywhere. Without video compression technology, uploading or downloading videos would take an entire day. Let’s explore the principles behind the video compression technology that makes it all possible! #MPEG #MotionVector #GOP Written by Error Edited by jeeholee unrealtech2021@gmail.com
Xi, Jinping is the biggest threat to civilization.
"FIRST OFF GUYS!!!! the deepSeek v3 Technical report - WHICH GUYS REMEMBER i've showed case NUMEROUS amount of times in a earlier videos was published on dec 27th, 2024" Makes it sound like you knew this before the market crashed. When your first video was the day the market crashed. brainless sheep.
No.
형 AI 스피치 티 나나봐..
Gemini is cheaper if you don't value your privacy.
OpenAI is coping and seething RN
Great video! It looks like it will be helpful for studying English.
So good !!!!
Such an unpleasant voice and tone can't listen. By the way, "mini LED" and QLED technology still uses LCD panels, so they are LCDs. The only fundamentally different panel tech is OLED and micro-LED (which can finally be called LED tech, as they use miniature LEDs as pixels, as opposed to modern LCDs which use LEDs only for the backlight).
Would be an enjoyable video ifyou werent screaming the information at me 😂
북한 동포를 위한 함경도사투리 버전 채널은 없나여?ㅠㅠ
영어 어떤 프로그램 쓰신 건가요?
와우! 혹시 영어 더빙 어떤 소프트웨어 쓰시는 지 알려주실 수 있으신가요? 🙂
I love your english channel!! and you are good at Tech & english😊
아 몰랑
이 ai 영어잘하네
역기서도 활약하시네요.
Thank you! 🐰🐼🐻❄ Love it!
we could really do without these ai read wikipedia articles.
Informational video but stop yelling ffs
great explanation but ya... no need to yell lol most of us use headphones
truly interesting video. I haven't found any other channel that explains the model in such detail. Just one thing, you shouldn't use AI text to speech. Try to voice over the videos on your own. Even if you have a Korean accent it would still be better.
this entire video feels like it’s AI generated
It is. Including the god awful generated voice.
It is
Thank you for the video. The stock market reaction is majorly overdone. And, this will lead to a much bigger AI boom.
I would disagree. Companies like Meta, Microsoft and Google spent billions on Nvidia's AI hardware and engineers and what Deepseek shows that they don't need to do that. In the next year or 2, you will see cheaper AI accelerators from China and Chinese tech companies using Deepseek's open source code and have a much lower cost for their AI services and products. Nvidia's hardware is just overpriced and now there is no need for their latest and greatest hardware anymore.
@pugster73 I think it's being reported now that DeepSeek spent $500M on NVDA chips.
super informative, i like and subscribe so you can afford a proper mic
Your voice is annoying like hell.
IMO this will cause good competition and bring prices down in this tech as we have just found out the American companies have been telling their investors we need more money and higher quality hardware to run AI, but this was false so what else could they be telling us non truths about in tech solar comes to mind they are overpriced in the west but in the east they are cheap over all i look at this a plus because it will bring prices down to the retail customers
Thank you caller, I hope some day we can get you in the studio so we can hear you properly.
With Deepseek R1 Nvidia GPU will become cheaper and affordable
scam altman just lost for a long time in a really bad way.
China China China The game is on and it is not over until it is over yet this must be a in your face slam dunk for Sam ALTMAN the greedy lol
Why did I read this in President Trumps voice lmaoooooo
DeepSeek is the true "Open AI"
So true lol Perhaps Sam ALTMAN's middle name might be Greedy or Dishonest
Constantly referring to countering China is silly; the fundamental logic is false. Why DeepSeek released R1 is to allow the world to collaborate. Talk about countering China is wrong. Why didn't any Korean company come up with their AI, instead of following the U.S lead? Faster and bigger computer will not lead to more powerful AI; the human brain needs only 13 watts and it weighs 3 pounds.
DeepSeek may have exposed a scam. Companies overexpending resources. More can be done with less.
I need to see your mouth moving Good content though
lol
That was nice and informative, but... why are you screaming at me? I AM listening.
오~~ 네이티브 ㅎㅎ
Amazing!
형, 내가 영어가 짧아 구독이랑 좋아요만 누르는 거 이해해 줘~
thanks bro!!
기계음이 그래도 좀 있네요 잘봤어요
speaking very fast but i can hear it clearly
But how do u even reach the back side? The weafers are ~300μm thick, while the front+back does't look like 10μm to at best 50μm thick. Grinding down that much on a wafer with ~1ft diameter sounds nearly impossible. Are they slicking/dicing the 30cm wide weafer to more managable size before engineering the back side?
As I know, they use BG (back grinding) CMP process to obtain appropriate thickness.
@@범고래-z8q Thnx.
XPU canggih juga ya
최고다...이게 공대남이지