
Each week, we bring you 5 stories that resonated the most in our internal Slack channel #AI-news. We write the newsletter using various AI tools because we're an AI company and our marketing wants to move with the times too. ๐
Today you're reading the 56th issue in a row.
#1
OpenAI launches o3-mini ๐: A new rival for DeepSeek R1 ๐จ๐ณ!
OpenAI has introduced o3-mini ๐, a smaller and smarter AI model aiming to compete with DeepSeek R1, a new open-source model from China ๐จ๐ณ that has caused quite a stir. OpenAI is making o3-mini available to Plus, Team, and Pro subscribers ๐ณ, while free users will have only limited access โ.
This announcement comes at a time when DeepSeek R1 is challenging major players like OpenAI, Google, and Anthropic by offering a powerful model for free ๐ก. OpenAI claims that o3-mini is fine-tuned for math, science, and programming ๐ป and can tackle complex logical reasoning tasks by breaking them down into smaller parts ๐งฉ.
Additionally, OpenAI is actively recruiting PhD candidates ๐ to help train its models, focusing on creating scientific programming tasks to test AIโs problem-solving abilities. The company also plans to introduce new features โก, such as web search integration and function calling ๐, enhancing its modelsโ ability to work with external data and user-provided code ๐จโ๐ป.
ย
#2
Qwen 2.5-Max ๐ค: Alibabaโs answer to DeepSeek ๐ง?
Alibaba has introduced Qwen 2.5-Max, an open-source AI model, during the Lunar New Year celebrations, signaling that it takes AI seriously while competitors take a break ๐ด. This move follows the recent arrival of the DeepSeek model ๐, which disrupted the market and contributed to Nvidiaโs massive $593 billion valuation drop ๐ฐ.
Qwen 2.5-Max leverages a โmix of expertsโ architecture ๐ง , making it faster and more scalable ๐. It has been trained on over 20 trillion tokens and improved through human feedback ๐ฃ๏ธ. The model is available via Alibaba Cloud, allowing developers and businesses to easily integrate it into their applications ๐ป.
As interest in open-source AI models grows, this release strengthens Alibabaโs position in AI research while also raising concerns about security ๐ and potential misuse โ ๏ธ. More similar models are expected to be released soon โณ. What about you? Have you tried these new Chinese models yet ๐จ๐ณ?
ย
#3
EU AI Act is coming ๐: Which AI systems will be banned โ?
On February 2, 2025, the EU AI Act ๐ will take effect, banning AI systems with โunacceptable risksโ โ, such as social scoring, emotion recognition in workplaces and schools, and manipulative techniques. Companies that violate these rules face hefty fines ๐ฐ of up to 7% of their annual revenue.
However, the AI Office ๐ has yet to clarify which AI systems specifically fall under these bans, despite promising guidelines before the deadline โณ. Companies fear compliance chaos, while civil society groups criticize the rushed adoption of regulations โ ๏ธ.
Another hot topic is the enforcement of transparency and risk management rules for general-purpose AI models, such as ChatGPT-4 ๐. These rules will come into force in August 2025, but the AI Office is still finalizing its Code of Conduct, struggling to meet deadlines while addressing stakeholders' unresolved concerns.
With a small team and growing pressure, the AI Office needs to speed things up ๐. Without clear rules, the credibility of the AI Act and the EU's leadership in AI regulation ๐ could be at risk. Whatโs your take on this ๐ง?
ย
#4
IBM cuts costs ๐ฐ: Plans to save $3.5 billion ๐ต!
IBM plans to save $3.5 billion in 2025 ๐, primarily through layoffs โ and spending cuts across various areas. The company is shifting its focus to AI and digital transformation ๐ค as demand for these technologies soars ๐.
Even as profits decline, IBM is thriving in AI. The company announced that its AI business generated $5 billion ๐ต, as customers increasingly favor AI solutions over traditional IT services. While software revenues grew, consulting revenues declined, signaling a shift in client preferences ๐.
Investors reacted positively, sending IBMโs stock price up 9% ๐. However, employees should brace for more layoffs ๐, following similar trends in previous years. IBM is working to secure its position as a leader in AI solutions ๐ while keeping costs under control ๐.
ย
#5
"Human Authored" ๐: A new certification for AI-free books โ !
The Authors Guild of America has introduced the "Human Authored" certification โ to help readers identify books written without AI assistance. As AI-generated books flood online platforms ๐, this initiative ensures that readers can clearly distinguish human-created content ๐.
Certified books will be listed in a public database ๐ for easy discovery. Currently, only Guild members who have written books entirely on their own โ๏ธ can apply, but eligibility will gradually expand.
The use of minor AI tools, such as spell-checking, is allowed, but the core text must be 100% human-written. The goal isnโt to reject AI ๐ค but to bring clarity to a publishing world increasingly influenced by artificial intelligence ๐ข. Whatโs your take on this?