DeepSeek has been accused of hijacking ChatGPT. OpenAI says they have proof that a Chinese group trained their model using American work. DeepSeek claims to be ChatGPT and follows OpenAI's rules. An official response is awaited.
| #AI#Leader |
theguardian
| 2025/01/29
Scale AI’s CEO, Alexandr Wang, highlights key aspects of the U.S.-China AI competition in light of the new DeepSeek model. He points out that Chinese labs possess more H100 Nvidia chips than commonly believed, despite export restrictions. According to Wang, DeepSeek alone operates around 50,000 H100s, a fact that cannot be openly acknowledged due to U.S. export controls.
| #AI#Leader |
Scale AI CEO Alexandr Wang
| 2025/01/23
OpenAI has introduced Operator, an advanced AI designed to automatically handle various online tasks, such as filling out forms, booking travel, and ordering groceries. You give it a task, and it independently navigates the web to complete it.
| #AI#Leader |
OpenAI
| 2025/01/23
The Samsung Galaxy S25 series has been revealed. It looks very similar to last year’s models but includes some internal upgrades. New AI features have been introduced, such as a smart assistant and improved text selection. Overall, it’s a modest update.
My interpretation: Reviews remain less enthusiastic about the usability of AI features in smartphones. It’s not that there is no future for them, but rather a question of how truly transformative they are for users in the near future.
| #AI#Device |
Marques Brownlee
| 2025/01/22
Hamilton Ulmer, former Mozilla employee and a seasoned expert in Exploratory Data Analysis (EDA), highlights a persistent challenge in analytics: 80% of the effort goes into cleaning messy, unreliable data, while most tools prioritize the final visualization layer. This often-overlooked “first mile” of data work is critical for transforming raw data into a trustworthy foundation for meaningful insights.
| #Analysts |
Hamilton Ulmer
| 2025/01/22
OpenAI and the U.S. administration have announced Project Stargate, an ambitious $500 billion initiative to develop advanced AI systems within the United States. Backed by major companies like OpenAI and SoftBank, the project aims to solidify the country’s leadership in AI technology. The first phase, with an initial investment of $100 billion, will launch in Texas.
| #AI#Leader |
OpenAI
| 2025/01/21
According to Christophe Blefari, the 2025 Data trends highlight the importance of the Semantic Layer in the Data Warehouse. It helps reduce discrepancies in figures between BI tools. Iceberg facilitates the connection of multiple Data Lakes, thereby enhancing flexibility.
| #Analytics |
Christophe Blefari
| 2025/01/20
On January 20, DeepSeek released DeepSeek-R1, a powerful open-source AI model that competes with GPT-4o in math, coding, and reasoning—but at a much lower cost. Instead of relying on massive amounts of data like OpenAI, it learns by trial and error, breaking problems into smaller steps using Chain of Thought (CoT) reasoning. This helps it solve math problems better than GPT-4o and Claude 3.5 Sonnet while using fewer resources. DeepSeek-R1 proves that top AI models don’t need huge datasets to perform well. Its launch shook the tech world, showing that advanced AI can be built much cheaper. As a result, tech stocks dropped, and NVIDIA lost big. Even Donald Trump reacted, calling it a “wake-up call” for the industry.
| #AI#Leader |
Reuters
| 2025/01/20
Many junior data analysts struggle to get hired, says Phuong Nguyen. The key? Build a personal project. Showcasing your skills through real work helps recruiters see your potential. 🚀
| #Analysts |
Phuong Nguyen
| 2025/01/19
As the Ex-CPO of BlaBlaCar points out, even the best algorithm won’t matter if you can’t make it work for users. The real challenge is bridging the gap between technical excellence and user experience.
| #Marketing |
Rémi Guyot
| 2025/01/15
Elon Musk's chatbot, Grok, received upgrades in H2 2024. It now has better reasoning and can understand images. Users can access new features like web search and PDF support. Grok's image generator shows less restriction, making it interesting for users.
| #AI |
X , Elon Musk
| 2025/01/13
ChatGPT’s new Advanced Voice with Vision feature allows you to interact using your voice and video, making it more engaging and fun—like FaceTiming ChatGPT or showing it the world around you, reminiscent of the movie Her.
My interpretation: In my opinion, the implications for customer relationships are enormous. Imagine chatting with a “smart mirror” in a fitting room for personalized fashion advice, or having a shopping assistant that sees which products you’re considering and offers tailored suggestions based on your preferences.
| #AI#Marketing#Visio |
OpenAI
| 2025/01/12
At CES 2025, a wave of exciting innovations was unveiled, including Roborock’s Saros Z70 robot vacuum. Equipped with the groundbreaking OmniGrip arm, this advanced device can pick up small objects like socks and tissues, taking autonomous home cleaning to the next level.
| #Robots |
CNET
| 2025/01/10
At CES 2025, the Unity Tree stands out. Nearby, the W area showcases exciting robots with wheels. It's a lively atmosphere filled with tech innovations, making it a fun party for all visitors.
| #Robots |
Unity Tree
| 2025/01/07
Check out this fascinating roundup from WEF of the fastest-growing vs. declining jobs by 2030—it's like a tale of two industries. The growth column reads like an AI conference agenda (think: data warehouses, IoT specialists, ML specialists), while the decline column... well, it's basically a list of things ChatGPT already does really well.
| #Work |
World Economic Forum
| 2025/01/07
NVIDIA has unveiled Project DIGITS, a personal AI supercomputer, starting at $3k, designed to deliver an impressive 1 petaflop of AI performance. This compact yet powerful device enables AI researchers, data scientists, and students to develop and fine-tune large AI models directly from their desks.
My interpretation: NVIDIA is aiming to expand the market by bringing AI computing to personal devices, increasing the number of potential customers for its chips.
| #Computing |
NVIDIA
| 2025/01/06
AI is a powerful tool for rapid prototyping. In Andrew Ng’s latest newsletter, he notes that AI suits prototyping since prototypes don’t need high reliability or full integration. This enables quick, effective tests to assess value.
My interpretation: I agree. Today, even those with little coding experience can build dashboards or apps using LLMs, which generate code or guide them step by step through development.
| #AI |
Deeplearning
| 2025/01/01
Anne-Claire Baschet, Chief Data & AI Officer at Mirakl, highlights strategies fueling the French unicorn’s success following its $555 million fundraising in early 2024. She stresses the importance of understanding user needs, fostering collaboration between data, product, and engineering teams, and leveraging generative AI to streamline processes. Clear role definitions and effective recruitment are key to scaling teams, while self-service tools empower employees to utilize AI efficiently. Balancing machine learning and generative AI, making data-driven decisions, and continuously iterating based on feedback are crucial for impactful product outcomes, fostering a culture of improvement and ensuring Mirakl’s leadership in marketplace innovation.
| #Organisation |
Mirakl, Anne-Claire Baschet
| 2025/01/01
According to the 2024 State of Analytics Engineering study by dbt Labs, most analytics engineers in North America earn over $100K annually.
| #Analytics |
dbt Labs
| 2025/01/01
Sora is a powerful AI tool from OpenAI that creates videos quickly from text. Marques Brownlee presents how users can generate realistic or cartoon-style videos easily. While it's impressive, it sometimes has flaws in motion, making it clear that the videos are AI-generated.
| #AI#Video |
OpenAI
| 2025/01/01
Elon Musk shares his thoughts on the biggest engineering mistakes and his 5-step optimization process. He warns that smart engineers often focus on fixing things that shouldn’t exist in the first place. It’s crucial to question the problems we choose to tackle rather than simply accepting them as given.
| #Leader#Organisation#Productivity |
Elon Musk
| 2025/01/01
In Philippine call centers, AI enhances efficiency by monitoring calls, scoring agents, providing real-time assistance, analyzing sentiment, and optimizing speech. It increases call volume, automates quality control, consolidates tasks, and reduces workforce needs. AI also accelerates data review and content creation.
| #Marketing |
restofworld.org
| 2025/01/01
You can train your own AI model easily. It can be faster and cheaper than using big models. Break your problem down, use simple code, and gather good data. With the right tools, you can create powerful solutions.
My interpretation: Maybe you can, but we see that leading services go beyond basic functionality—they integrate video, voice, agents, and task management. I think it will take time before DIY models become mainstream.
| #AI |
Steve (Builder.io)
| 2025/01/01
Grok, xAI’s AI chatbot, is now free for all X users, with limits for non-Premium users. Launched as a humor-driven assistant, it competes with ChatGPT and DeepSeek. xAI, backed by $6 billion in funding, is testing a free version and considering a standalone app to boost engagement.
My intepretation: It will be interesting to see how widely Grok is adopted, given the fierce competition from ChatGPT, DeepSeek, and others.
| #AI#Leader |
theverge
| 2025/01/01
Meta is investing $10 billion in an AI data center in northeast Louisiana, creating 500 direct and 1,000 indirect jobs. The 4-million-square-foot facility will run on 100% renewable energy. At its peak, 5,000 workers will be involved. Meta will also invest $200 million in local infrastructure and partner with schools for STEAM education, further boosting economic growth.
My interpretation: The race between titans is just beginning.
| #AI#Leader |
www.opportunitylouisiana.gov
| 2025/01/01
OpenAI’s o3 has achieved a Codeforces rating of 2727, placing it among the top 175 human coders worldwide. This milestone showcases AI’s growing dominance in problem-solving and competitive programming.
My interpretation: I’m using o3 for coding this newsletter, and I can say that this model has reduced my coding time by a factor of 10.
| #AI#Leader |
Deedy Das
| 2025/01/01
OpenAI o3 is 2727 on Codeforces which is equivalent to the #175 best human competitive coder on the planet.
