Meta launches new AI chips as Microsoft unveils Copilot Health

Key Takeaways

• Meta is developing four new custom AI chips (MTIA 300-500 series) to reduce reliance on Nvidia and AMD, aiming for better performance and control over its AI infrastructure.
• Microsoft's Copilot Health AI tool allows users to combine medical records from over 50,000 U.S. providers with data from wearables like Apple Watch, offering personalized health insights while promising data privacy.
• Amazon is actively managing the narrative around AI's environmental impact and ethical use, addressing concerns about AWS data center energy and water consumption.
• Palantir and Nvidia are partnering to provide a secure AI platform for governments, leveraging Nvidia's H100 GPUs for allied nations.
• OpenAI CEO Sam Altman predicts AI's cognitive capacity will surpass human capacity by late 2028, fundamentally altering the labor-capital balance and potentially creating an era of abundance.
• AI decision support systems are now used in military planning to analyze impacts on civilian infrastructure, raising concerns about precision and biases.
• Anthropic has launched The Anthropic Institute to research AI safety and risks, sharing findings and partnering with external groups.
• Udacity, in collaboration with Accenture, offers an accredited AI MBA program for under $5,000, focusing on training AI product leaders.
• F5 executive Shawn Wormke identifies agentic identity and intent as major unsolved AI security challenges, especially with multimodal AI.
• A UK survey found 98% of undergraduate students use generative AI for studies, prompting calls for clearer academic guidelines.

Meta rolls out four new AI chips for better performance

Meta is releasing four new chips it designed itself to handle artificial intelligence tasks. This move helps Meta rely less on companies like Nvidia and improve its AI capabilities for platforms like Facebook and Instagram. The chips are designed for different AI jobs, including training AI models, running them efficiently, and even for use in virtual reality devices. This strategy aims to give Meta more control over its hardware and AI development.

Meta's new AI chips show faster, self-reliant hardware plan

Meta has revealed four new AI chips as part of its plan to create more of its own hardware for AI tasks. These chips, named MTIA 300 through MTIA 500, are designed to handle various AI jobs, including training and running generative AI models. Meta developed these chips in less than two years, showing a faster development cycle. This strategy aims to reduce costs, speed up AI deployment, and give Meta more control over its technology.

Meta plans four custom AI chips in two years

Meta announced plans to release four custom AI chips within the next two years to reduce its dependence on chipmakers like Nvidia. These specialized processors are designed to optimize AI tasks for platforms like Instagram and Facebook, aiming for better performance and energy efficiency. While Meta is developing its own chips, it will still rely on third-party GPUs from companies like Nvidia and AMD. This aggressive timeline highlights the challenges and significant investment required in custom chip development.

AI chatbots now want your health records

Microsoft is launching a tool called Copilot that will allow users to share health records from various providers with its chatbot. This information can be combined with data from fitness devices like the Apple Watch. Other companies like Amazon, OpenAI, and Anthropic are also testing similar health AI tools. While these tools could offer insights into health issues, sharing medical data with tech companies raises significant privacy concerns.

Microsoft previews Copilot Health AI tool

Microsoft has previewed its Copilot Health AI tool, which allows users to combine medical records, lab results, and data from wearables like Apple Health and Fitbit. The system analyzes this information to provide personalized health insights. This tool can access records from over 50,000 U.S. health providers and aims to help users understand test results and prepare for doctor's appointments. Microsoft states that conversations are kept private and health data will not be used to train its AI models.

Amazon shapes AI narrative on sustainability and ethics

Amazon hosted an event to address public concerns about the environmental impact and ethical use of AI. Executives pushed back on claims about water and energy usage by AWS data centers, explaining their optimization and replenishment efforts. The company also highlighted its workforce development programs and ethical AI initiatives. Amazon aims to manage the narrative around AI's impact, especially after facing negative headlines regarding its data centers and climate footprint.

AI algorithms change how civilian infrastructure is protected in war

Artificial intelligence (AI) decision support systems are now used in planning and evaluating attacks in armed conflicts, especially in urban warfare. These systems help analyze the complex effects of attacks on civilian infrastructure like power grids and communication networks. By simulating different scenarios, AI can estimate consequences such as blackouts or disruptions to essential services. This changes how international humanitarian law is applied, but also raises concerns about the illusion of precision due to data gaps and model biases.

Sam Altman says AI disrupts labor-capital balance

OpenAI CEO Sam Altman stated that artificial intelligence is fundamentally changing the balance between labor and capital, creating an era of abundance. He noted that AI is increasingly blamed for job losses and rising costs, and that companies are shifting from large workforces to investing heavily in computing power. Altman predicts that AI's cognitive capacity will soon surpass human capacity, potentially by late 2028. He envisions AI becoming a cheap, widely available utility, which could reshape capitalism and traditional economic measures.

Palantir and Nvidia partner for secure government AI

Palantir and Nvidia are collaborating to offer a new AI platform for governments that prioritizes data security and privacy. This platform, called Palantir Government, will allow allied nations to build and deploy AI applications using Palantir's software on Nvidia's hardware, including H100 GPUs. The service will be available to governments in the U.S., U.K., Australia, Canada, France, Germany, and Japan. This partnership aims to create a secure ecosystem for government AI development.

Nearly all UK students use AI for studies

A survey in the UK found that 98% of undergraduate students use generative AI tools for their studies, a significant increase from last year. Students use AI for tasks like brainstorming ideas, writing text, and checking grammar. The survey highlights concerns about academic integrity and the potential misuse of AI. The Higher Education Policy Institute is calling for clearer guidelines from universities on acceptable AI use to ensure fairness for all students.

Anthropic launches Institute for AI safety research

Anthropic has launched The Anthropic Institute to study the challenges and risks associated with rapidly advancing AI. The institute will share its findings about frontier AI systems and partner with external groups to address potential risks. Led by Jack Clark, it includes experts in machine learning, economics, and social science. The Institute aims to provide candid insights from AI builders and engage with communities affected by AI's future impact.

Udacity and Accenture launch affordable AI MBA

Udacity, part of Accenture, has launched an accredited MBA program focused on training AI product leaders. This program can be completed for under $5,000, making it significantly more affordable than traditional MBAs. The curriculum includes courses on AI transformation, business intelligence, and growth marketing. The degree is awarded by Woolf University and is designed to address the growing demand for professionals who can bridge technology and business strategy in the AI economy.

F5 exec highlights unsolved AI security problems

F5 executive Shawn Wormke identified agentic identity and intent as major unsolved AI security challenges. As more AI agents are used, determining who or what is taking action and understanding their true goals becomes critical. Current security measures, often text-based, struggle with multimodal AI that processes audio and images. Wormke also noted that the rapid pace of AI innovation constantly introduces new threats, requiring ongoing development of security solutions.

Sources

• Meta Preparing to Deploy Four New Homegrown Chips to Handle AI
• Meta’s New AI Chips Reveal a Faster, More Self-Reliant Hardware Strategy
• Meta announces four custom AI chips over two years
• A.I. Chatbots Want Your Health Records. Tread Carefully.
• Microsoft previews Copilot Health AI tool
• Inside Amazon’s effort to shape the AI narrative on sustainability and ethics
• Deciding under algorithms: artificial intelligence and the protection of civilian infrastructure in armed conflict
• Sam Altman admits AI is killing the labor-capital balance—and says nobody knows what to do about it
• Palantir and Nvidia Team Up on Sovereign AI for Governments
• AI use by undergraduate students is nearly universal, survey suggests
• Introducing The Anthropic Institute
• Udacity, Part of Accenture, Launches Accredited MBA to Train the Next Generation of AI Product Leaders
• The AI security problems nobody has solved yet