This week, we are attending two events: the Inaugural Convening of the International Network of AI Safety Institutes and the Conference on Frontier AI Safety Frameworks, where important discussions will take place about how to manage the next generation of highly capable AI technologies. A key topic will be Frontier AI Safety Frameworks, a tool for managing large-scale, public safety and national security AI risks. Microsoft and 15 other AI labs committed to publish their own frameworks ahead of the French AI Action Summit next year, and we are on track to do so. As we make progress, we wanted to highlight four areas where further progress is needed: ▪ Developing evaluation best practices. As scientists have acknowledged, evaluating advanced models is challenging given their general-purpose nature and the impact of deployment context. While early best practices for evaluations are developing, more work is needed to develop scientifically valid techniques. Multi-disciplinary collaboration must be accelerated to help develop more effective methods for adversarial testing and systematic measurement, as well as consistent approaches to sharing information about evaluation results. ▪ Advancing holistic risk assessment. More holistic approaches are needed to assess the risks a model may pose when integrated into a system and deployed in a real-world context. This involves considerations of system-level mitigations as well as societal and institutional factors external to the system that can impact whether and how a hazard materializes and the scale of its impact. A model with offensive cyber capabilities, for example, will have a different risk profile if deployed in an environment with broadly implemented cybersecurity safeguards, as compared to the situation one where no such safeguards exist. ▪ Assessing marginal risks and benefits. Potential risks should be weighed alongside a model’s benefits. In both cases, a marginal assessment can be helpful to shed light on how they compare with currently available tools or information. This helps to inform a risk-benefit analysis of how the model may be used, as well as model release and deployment decisions. ▪ Cohering across frameworks. While Frontier Safety Frameworks play an important role in managing public safety and security risks, they must integrate with other governance motions, for example the EU AI Act Code of Practice for general purpose models and reporting for the Hiroshima AI Process International Code of Conduct. At Microsoft, our Responsible AI Standard will continue to serve as the foundation for how we operationalize our responsibilities and govern AI risk. Much work is needed to build and strengthen the risk management techniques that will enable the next generations of AI technologies to be deployed in ways that are useful and trustworthy. We look forward to learning from and engaging with others at this week’s events.
Microsoft News and Stories
Technology, Information and Internet
Redmond, Washington 12,321 followers
About us
All the latest company news and stories about people doing extraordinary things with technology
- Website
-
https://news.microsoft.com
External link for Microsoft News and Stories
- Industry
- Technology, Information and Internet
- Company size
- 10,001+ employees
- Headquarters
- Redmond, Washington
Updates
-
President and COO, Strategic Missions and Technologies, Onalytica's Who's Who in AI - Influential Voices
Huge congratulations to our collaborators at Pacific Northwest National Laboratory, who just won HPCwire’s Annual Readers’ and Editors’ Choice Awards for our work together in sustainable battery material innovation using Azure Elements. As we continue to expand our partnership, I look forward to exploring new applications of HPC and AI to accelerate discovery across other scientific disciplines. Bravo! Brian Marble Tom Tabor Louis Terminello #AI4Science #HPC #AI #AzureElements
-
-
Check out this article by Jeana Jorgensen and Ryan Currie on why AI skilling is essential and how Microsoft is helping people and organizations thrive in the AI economy. With over 23 million people across 200+ countries already benefiting from Microsoft’s skilling initiatives, the impact is clear. From government officials to students, business users to technology professionals, these programs are building confidence in AI fluency and opening doors to new opportunities. Highly recommend giving this a read: https://lnkd.in/g9UmtU46 #AI #DigitalSkills #MicrosoftIgnite
-
-
Generative AI is transforming the way we work and innovate, with the potential to boost the U.S. economy by $3.8 trillion by 2038. A new white paper from Accenture and Microsoft explores the current state of gen AI and how it is fueling advancements across industries—from healthcare to travel to clean energy. Innovations like open-source models and reduced input costs are making generative AI more accessible and delivering real-world benefits like higher wages, lower consumer costs, and new products and services across industries. To fully realize these benefits, we must invest in a skilled workforce, scalable infrastructure, and trusted policy frameworks. Read more about how generative AI is empowering people, businesses, and society to achieve more: https://lnkd.in/g5Vyf8iF #GenerativeAI #ResponsibleAI #DigitalSecurity
-
-
As we wrap up another successful #MSIgnite, I wanted to dedicate November’s edition of Heart of Security to discussing a few of the groundbreaking innovations announced earlier this week that made this year’s event extraordinary. It was such a pleasure connecting with the wonderful people and partners at the event, including this month’s spotlight Tara Ragan. Tara is an outstanding compliance partner from Lighthouse who shared valuable insights on the evolving landscape of data security and compliance, and the role diversity plays in driving meaningful innovation. I’d love to hear how your perspectives on security, compliance, and governance are shifting in this dynamic era and how you imagine it will continue to change! Thank you for reading Heart of Security!💜
Securing our future from Microsoft Ignite
Vasu Jakkal on LinkedIn
-
Find out how AI agents can tackle tasks either with you or for you, from acting as a virtual project manager to handling more complex assignments. Check out our post about what’s behind agents and Microsoft’s plans for evolving them. https://lnkd.in/gtR9dTqX
AI agents — what they are, and how they'll change the way we work - Source
news.microsoft.com
-
The world’s creating more data than it can store. In “Ideas,” researchers reflect on their journey to develop a system for storing archival data in synthetic DNA, including the open-source release of their most powerful algorithm for DNA error correction. https://lnkd.in/em4H6Fiw
-
-
Microsoft Flight Simulator 2024 is bringing huge enhancements to the series, including incredible new details to the on-the-ground experience. Players are now able to exit their aircraft and see ground detail down to tiny stones and individual blades of grass, all modeled in 3D. That detail includes the impacts of takeoff, flight, and landing physics on the environment. If you land in a field, you might see indentations in the soil and dirt kicked up by your aircraft's landing gear. If you’re flying a helicopter, you could observe how the speed of the rotor impacts the intensity of a tree branch blowing. “We can basically get every rock on Earth now; it’s unbelievable,” says Head of Microsoft Flight Simulator Jorg Neumann. “We updated the aerial imagery for the entire world, and we set up a team of specialists that gather digital elevation maps from across the planet. The result of these last four years of work is that we now have an unprecedented level of ground detail.” Congrats to Microsoft Flight Simulator 2024 on an incredible new installment!
-
-
BiomedParse reimagines medical image analysis, integrating advanced AI to capture complex insights across imaging types—a step forward for diagnostics and precision medicine. https://lnkd.in/evBjgdeD
-
-
This. Is. An. Xbox. 🎮💻📱📺 As we celebrate 23 wonderful years and pay homage to the original Xbox, we’re grateful for our amazing community and how far we’ve come together - evolving with players to offer more choice and flexibility in how and where they play. Today, Xbox is more than a console - it’s any device where you can jump into your favorite Xbox games - at home, on the go, anywhere, and across your devices. 💫 🔗 Find out more: https://lnkd.in/gApE8m_j