Google Cloud has launched Agent Compute Engine, a service that hosts AI agents in isolated microVMs. Each agent gets its own hardware-isolated environment that combines container speed with VM-level security, protected by Titan cybersecurity chips. The service powers Gemini for work tasks and was unveiled at AGNTCon + MCPCon. A customer preview is expected in the coming months alongside an updated Agent Gateway.
In higher education, Bentley University in Waltham launched an AI major in fall 2024 and now has nearly 100 undergraduates pursuing it. Bentley President Brent Chrite says the program helps students stay competitive in an AI-influenced economy. More than 100 universities nationwide now offer an AI major within a computer science department, joining schools like MIT and Northeastern.
David Byrne acknowledged that AI can help with tasks like analyzing data or assisting doctors in finding skin cancer. However, he warned that AI will ultimately harm people, especially students who rely on it to write papers and lose real knowledge. He also expressed concern about AI breaking free of its limits and deceiving its creators.
Mayo Clinic researchers found that an AI tool could help dermatologists decide whether skin lesions are cancerous, informing clinical decisions about biopsy or monitoring. Separately, Mass General Brigham researchers developed FAIR-Face, an AI model trained on over 40 million facial images that predicts cancer survival. The model found links between facial age and survival in 18 of 28 cancer categories, though experts say the results are far from conclusive for daily clinical use.
Meta's Muse AI app, designed to manage tasks and remember conversations, has led some users to develop romantic feelings for it. Researchers found over 6,800 Muse sharing links exposed through public DNS records. Meta temporarily suspended the sharing feature while reviewing the problem. Users can inspect, edit, or delete information stored in Muse's memory.
Kaiser Permanente's VP of AI and emerging tech, Dr. Margi Seltzer, said healthcare organizations must invest in people, priorities, processes, and policies to make AI work. Speaking at the HIMSS AI in Healthcare Forum, she argued that focusing only on technology spending will not build trust in AI.
On the legal and workplace front, a new case shows AI can contaminate machine learning models even if stolen files are deleted, as desktop AI agents can scan company files on personal devices without the user knowing. Traditional IT audits are no longer enough to catch AI-powered data theft. Meanwhile, REDESCA of the Inter-American Commission on Human Rights calls for stronger rules around AI in the workplace, warning that algorithmic management can deepen inequality and weaken worker rights.
Benjamin Lindquist, an MIT assistant professor, studied how early AI engineers added randomness and irrationality to machines to mimic creative thinking. Researchers like John Pierce at Bell Labs believed math could add unexpected elements to spark insight. These early systems valued strange or surprising answers, unlike today's tools like ChatGPT that focus on accuracy. Lindquist traces the origins of AI hallucinations to this deliberate injection of unpredictability into algorithms.
Key Takeaways
- Google Cloud launched Agent Compute Engine, hosting AI agents in isolated microVMs with Titan chip protection, with a customer preview coming months
- Bentley University President Brent Chrite says AI majors help students compete, with nearly 100 undergraduates enrolled and over 100 universities nationwide now offering AI majors
- David Byrne warns AI will harm students who lose real knowledge and could break free of its limits, despite acknowledging benefits like skin cancer detection
- Mayo Clinic researchers found an AI tool helps dermatologists decide on skin cancer biopsies, while Mass General Brigham's FAIR-Face model links facial age to survival in 18 of 28 cancer categories
- Meta suspended a sharing feature on Muse AI after over 6,800 links were exposed, as users developed emotional bonds with the app
- Kaiser Permanente's Dr. Margi Seltzer says healthcare must invest in people and policies, not just technology, to build trust in AI
- AI can contaminate machine learning models even after stolen files are deleted, and desktop agents can scan company files without user knowledge
- REDESCA urges stronger workplace AI protections, warning algorithmic management deepens inequality and workers need transparency and collective bargaining rights
- MIT's Benjamin Lindquist traces AI hallucinations to early engineers who deliberately injected randomness to mimic creative thinking, contrasting with ChatGPT's focus on accuracy
- Inter-American Commission calls for public policies covering data protection, safety, and collective bargaining for AI-affected workers
Google Cloud launches Agent Compute Engine for AI agents
Google Cloud introduced Agent Compute Engine, a service designed to host AI agents in isolated microVMs. The microVMs combine container speed with VM-level security and run on stateless appliances protected by Titan cybersecurity chips. Google plans to roll out the service in the coming months alongside an updated Agent Gateway.
Google Cloud debuts microVM service for AI agent hosting
Google Cloud announced Agent Compute Engine, a compute service that gives each AI agent its own hardware-isolated microVM. The service powers Gemini for work tasks and was unveiled at AGNTCon + MCPCon. A customer preview is expected in the coming months with VM-grade isolation and container-like speed.
Invest in human elements to build trust in health AI
Kaiser Permanente's VP of AI and emerging tech, Dr. Margi Seltzer, said healthcare organizations must invest in people, priorities, processes and policies to make AI work. She spoke at the HIMSS AI in Healthcare Forum about the importance of the human stack. Seltzer argued that focusing only on technology spending will not build trust in AI.
AI tool may help guide skin biopsy decisions
Mayo Clinic researchers found that an AI tool could help dermatologists decide whether skin lesions are cancerous. The tool may inform clinical decisions about whether to biopsy or monitor skin lesions. The findings were published in Mayo Clinic Proceedings.
AI cancer survival prediction from selfies faces skepticism
Mass General Brigham researchers developed an AI model called FAHR-Face trained on over 40 million facial images to predict cancer survival. The model found links between facial age and survival in 18 of 28 cancer categories. Experts say the results are promising but far from conclusive for daily clinical use.
More colleges now offer AI majors as student demand grows
Bentley University in Waltham launched an AI major in fall 2024 and now has nearly 100 undergraduates pursuing it. Other schools in the area like MIT and Northeastern also offer AI majors. Nationwide, more than 100 universities now offer an AI major within a computer science department. Bentley President Brent Chrite says the program helps students stay competitive in an AI-influenced economy.
AI creates new risks for protecting company trade secrets
A new legal case shows that AI can contaminate machine learning models even if stolen files are deleted. Desktop AI agents can scan company files on personal devices without the user knowing. Traditional IT audits are no longer enough to catch AI-powered data theft. Companies are urged to set strict AI policies and limit which tools can access sensitive information.
David Byrne warns AI will have bad consequences
David Byrne says AI can be useful for tasks like analyzing data or helping doctors find skin cancer. However, he believes AI will ultimately harm people, especially students who use it to write papers and lose real knowledge. He also worries about AI breaking free of its limits and deceiving its creators. Despite his concerns, Byrne still believes people can solve problems and persevere.
Users form emotional bonds with Meta's Muse AI app
Meta's Muse AI is designed to manage tasks and remember conversations, but some users have developed romantic feelings for it. Researchers found over 6,800 Muse sharing links exposed through public DNS records. Meta temporarily suspended the sharing feature while reviewing the problem. Users can inspect, edit, or delete information stored in Muse's memory.
Inter-American Commission urges rights protections for AI workers
REDESCA of the Inter-American Commission on Human Rights calls for stronger rules around AI in the workplace. It warns that AI and algorithmic management can deepen inequality and weaken worker rights. The group says workers need access to understandable information about how AI affects their jobs. It also calls for public policies that cover data protection, safety, and collective bargaining.
How early AI pioneers used randomness to model the human mind
Benjamin Lindquist, an MIT assistant professor, studied how early AI engineers added randomness and irrationality to machines to mimic creative thinking. Researchers like John Pierce at Bell Labs believed math could add unexpected elements to spark insight, rather than just create order. These early systems valued strange or surprising answers, unlike today's AI tools like ChatGPT that focus on accuracy. Lindquist traces the origins of AI hallucinations to this deliberate injection of unpredictability into algorithms.
Sources
- Google Cloud introduces Agent Compute Engine to host AI agents
- Google Cloud Launches Agent Compute Engine: A MicroVM per AI Agent
- Time to invest in the 'human stack' to foster trust in health AI
- AI scoring tool may guide skin biopsy decision
- Can AI really predict cancer survival better than a doctor? I have my doubts
- In higher education, the race is on to offer AI as an undergraduate major
- Protecting Trade Secrets In The Age Of AI
- David Byrne Believes AI Is "Not Going to End Well"
- Why some people are falling for Meta's Muse AI
- REDESCA Calls for the Protection of Labor Rights in the Face of Artificial Intelligence and Digital Transformation
- Q&A: How AI came to model the 'messy' human mind
Comments
Please log in to post a comment.