DeepSeek-R1 Release Under MIT Lisence 📜, FTC Report On AI Partnerships 🕵️‍♂️, Perplexity AI Bid 💼

In partnership with

An entirely new way to present ideas

Gamma’s AI creates beautiful presentations, websites, and more. No design or coding skills required. Try it free today.

💬 OpenAI Developments

Behind the Curtain — Coming soon: Ph.D.-level super-agents

OpenAI and Meta are actively developing AI agents capable of reasoning and executing complex tasks autonomously, aiming for mainstream adoption by 2025. This trend reflects a broader industry movement toward integrating advanced AI assistants into consumer and business applications, with companies like Microsoft and Google also participating in this competitive landscape.

OpenAI Just Pulled a Theranos With o3

OpenAI's o3 model has come under scrutiny for achieving record-breaking performance on the EpochAI FrontierMath benchmark, despite having prior access to much of the test data and funding its development. This situation has drawn comparisons to Theranos, with AI experts questioning the legitimacy of OpenAI's claims and highlighting a lack of transparency in the company's involvement.

🌐 AI in Various Sectors

Healthcare: AI is revolutionizing medical training through realistic simulations, data-driven feedback, virtual patient systems, and decision support tools, enhancing skills, diagnostics, and efficiency. Despite challenges like data security and system integration, AI is shaping a smarter, inclusive, and more effective future for healthcare education and patient care.

Robotics: German robotics company Sereact has secured a €25 million Series A funding round to advance its AI-powered software, which enables robots to perform tasks without explicit training through zero-shot visual reasoning and natural language instructions. This technology allows non-technical users to operate robots via chat interfaces, facilitating rapid deployment and immediate cost savings across various industries in the US and Europe.

Social Media and Interaction: AI agents like Adaly and Anthrologic are streamlining advertising tasks by automating data analysis, reporting, and budgeting, with brands like Anthropologie already adopting the tech. While promising, these tools still require human oversight for nuanced decision-making and to address issues like hallucinations.

🚀 Tech Industry Moves

Nvidia: Industry analysts have reduced their 2025 shipment forecasts for Nvidia's GB200-based NVL72 machines from 50,000–80,000 units to 25,000–35,000 units, a 56% decrease. This adjustment pertains specifically to systems utilizing Nvidia's Arm-based processors, while demand for x86-based machines with B200 GPUs remains strong.

Nvidia and Yonsei University have introduced Omni-RGPT, a multimodal large language model designed for region-level understanding in both images and videos. By implementing 'Token Mark', the model embeds region-specific tokens into visual and textual prompts, ensuring consistent object representation across frames and reducing computational complexity. Additionally, the researchers developed RegVID-300k, a comprehensive dataset comprising 98,000 unique videos and 294,000 region-level instruction samples, to enhance the model's performance in tasks like visual commonsense reasoning and captioning.

Perplexity AI has submitted a merger bid to ByteDance for TikTok U.S., proposing a new entity combining Perplexity, TikTok U.S., and new capital partners. The deal aims to retain most of ByteDance’s current investors, integrate more video content into Perplexity, and address U.S. regulatory concerns. ByteDance, which has resisted selling TikTok U.S., sees the proposal as an alternative to divestment, with the potential deal valued at over $50 billion. The transaction’s future hinges on negotiations, including investor preferences and potential extensions from the incoming U.S. administration.

Google has started requiring users to enable JavaScript to use Google Search, aiming to enhance protection against malicious activities like bots and spam, and to improve the overall user experience. A company spokesperson noted that without JavaScript, many Google Search features won't function properly, and the quality of search results may be degraded.

Google's DeepMind has developed AI models that emulate human-like memory, enabling machines to retain and recall information similarly to humans. This advancement aims to enhance AI's ability to learn from past interactions, improving performance in tasks requiring contextual understanding and long-term dependencies.

Microsoft is integrating its AI-powered Copilot features into Microsoft 365 Personal and Family subscriptions, resulting in a $3 per month price increase. Existing subscribers can opt out to maintain current pricing without the AI enhancements. This change expands the availability of AI features, previously limited to certain regions, to most markets worldwide.

⚖️ AI Ethics and Regulation

FTC points at partnerships between AI leaders and cloud service providers, suggesting they could be subject to antitrust scrutiny

The Federal Trade Commission (FTC) has released a staff report scrutinizing partnerships between major cloud service providers—Alphabet, Amazon, and Microsoft—and leading AI developers, Anthropic and OpenAI. The report highlights potential antitrust concerns, noting that these collaborations may grant cloud providers access to sensitive technical and business information, which could be used to develop competing AI models, thereby distorting innovation and undermining fair competition.

The Pentagon says AI is speeding up its ‘kill chain’

The Pentagon is leveraging generative AI from companies like OpenAI, Anthropic, and Meta to enhance its "kill chain," improving threat identification and decision-making without allowing AI to make autonomous life-and-death decisions. AI is used collaboratively with humans for planning and strategizing, with strict ethical guidelines ensuring human involvement in force deployment. Despite controversy over AI in military applications, developers and the Pentagon aim for responsible use, with some in the AI community advocating collaboration to mitigate risks and ensure ethical deployment. The growing integration of AI in defense highlights a shift toward human-machine teaming rather than fully autonomous systems.

💡 Emerging AI Technologies and Companies

Primech AI improves HYTRON cleaning robot performance with NVIDIA tech

Primech AI has enhanced its HYTRON bathroom cleaning robot by integrating NVIDIA's Jetson Orin Super system on module (SoM), which provides robust edge AI capabilities and energy-efficient processing. This integration enables HYTRON to navigate tight spaces and perform multiple cleaning tasks autonomously, improving performance and reliability in commercial settings such as airports, hospitals, and shopping malls.

An open call for the next Google.org Accelerator: Generative AI

Google.org has announced an open call for its second Generative AI Accelerator, inviting nonprofits, civic entities, academic institutions, and social enterprises to develop AI-powered solutions addressing global challenges across Knowledge, Skills, and Learning; Scientific Advancement; and Resilient Communities. Building on the success of the 2024 cohort, which supported 21 organizations to serve over 30 million people by 2028, this six-month accelerator offers technical training, Google Cloud credits, pro-bono support, and a share of $30 million in funding. Applications close on February 10, 2025, at g.co/Accelerator/GenAI.

🎨 AI in Creative Industries

Chinese AI lab DeepSeek has released DeepSeek-R1, a reasoning model available on Hugging Face under an MIT license, claiming it outperforms OpenAI’s o1 on benchmarks like AIME, MATH-500, and SWE-bench Verified. With 671 billion parameters, R1 excels in reliability for tasks in physics, science, and math but operates under Chinese regulations, avoiding sensitive topics. Distilled versions from 1.5 to 70 billion parameters make the model accessible on smaller devices, while the full version is API-accessible at a fraction of o1’s cost. This release highlights China’s growing AI advancements amid U.S. export restrictions and geopolitical tensions.

🌍 Global AI Race

AIxBlock Partners with PublicAI to Enhance Global AI Training Data Accuracy

AIxBlock has partnered with PublicAI to enhance the accuracy of global AI training data. This collaboration combines PublicAI's network of over 500,000 validated contributors with AIxBlock's platform, which includes more than 100,000 experienced data labelers from 120 countries, aiming to provide high-quality, precise datasets for various AI applications.

2025 AI Action Summit: Council of Europe Side event on African engagement in global AI governance

On February 11, 2025, the Council of Europe will host a side event at its Paris Office during the AI Action Summit 2025, focusing on enhancing African engagement in global AI governance. The event will include a signing ceremony for the Council of Europe's Framework Convention on Artificial Intelligence and Human Rights, Democracy, and the Rule of Law, and will be accessible in both English and French.

Dentons releases Global AI Trends Report

Dentons, the world's largest global law firm, has published its Global AI Trends Report, highlighting key legal issues anticipated in 2025. The report emphasizes the importance of businesses developing formalized AI roadmaps, noting that 63% of business leaders currently lack such plans, and addresses critical areas including regulatory cohesion, privacy and security, ethical considerations, and algorithmic bias.

If you want more updates related to AI, subscribe to our Newsletter