OpenAI's Superhuman AI Safeguard

PLUS: Microsoft unveils Phi-2 language model that packs power into a small package.

Welcome! Today is Friday, December 15th and in today's news we have updates on initiatives to align advanced AI systems safely, as well as new Intel processors optimized for artificial intelligence capabilities across devices. New to The Intelligence Age? Sign up here.
Drop us a line anytime at [email protected] with feedback.

News & Insights

OpenAI Thinks Superhuman AI Is Coming — And Wants To Build Tools To Control It

OpenAI's Superalignment team, led by co-founder Ilya Sutskever, is tackling the formidable challenge of developing methods to regulate AI systems that surpass human intelligence. At NeurIPS, team members Collin Burns, Pavel Izmailov, and Leopold Aschenbrenner conveyed the urgency of aligning AI with human intentions, an area where current strategies may fall short as AI sophistication grows. Their approach involves guiding advanced AI models with weaker ones to promote desired behaviors and mitigate risks, an innovative strategy reflecting the need for robust governance frameworks in an era of rapid AI advancements.

With superintelligent AI on the horizon, OpenAI's initiative underscores the criticality of aligning such systems with human values. They're backing this up with a $10 million grant program, inviting global research contributions and planning a 2025 conference to highlight progress. Despite concerns of premature focus and potential commercial interests as noted by Eric Schmidt's involvement, OpenAI commits to openly sharing superalignment research, emphasizing their dedication to AI safety and benefit for humanity. techcrunch

Intel Puts Artificial Intelligence First With Processor Announcements

Intel's latest advancements signify a strategic pivot to prioritize AI capabilities across the spectrum of computing devices. The newly unveiled Intel Core Ultra mobile processor family, with its on-chip AI accelerator, promises to revolutionize AI tasks for end-users by greatly improving power efficiency and performance. This chipset is a game-changer for graphic designers and smart factories, providing the muscle to run demanding generative AI applications locally.

Meanwhile, the 5th Gen Intel Xeon processors, set to be available next year, are poised to become the backbone of data centers seeking AI acceleration. These processors are tailored to optimize the performance of large language models, with up to 42% higher inference and fine-tuning capabilities. This leap in efficiency underlines Intel's commitment to embedding AI into the fabric of every PC, a vision echoed by CEO Pat Gelsinger's remarks on technology being a "neutral" force that must be shaped for the greater good. techrepublic

Together with The AI Solopreneur

Grow your business or personal brand on autopilot

If you’ve ever wanted to build an audience for your business or personal brand, AI Audience Accelerator will help you—in a fraction of the time.

  • Create one week's worth of engaging Twitter, LinkedIn, and newsletter content in one hour.

  • Learn how to create content that builds your authority and generates leads.

  • Spend less time on the content treadmill and more time focusing on growing your business.

Please support our sponsors!

Mastering AI

Top Tools

  • Taskade: All-in-one productivity tool with AI-powered chat, tasks, and collaboration. (website) (sponsored)

  • Polymath Robotics: Safe, autonomous navigation for industrial vehicles in controlled environments. (website)

  • WritePanda: AI-powered writing assistant for ads, documents, podcast transcription & more. (website)

Practical AI

Around The World

More News

  • OpenAI's Chief Scientist Ilya Sutskever outlines a strategic approach to regulate and control super-intelligent AI, ensuring its alignment with human values and safety.

  • Nearmap Employs Advanced Aerial Imagery and Artificial Intelligence to Revolutionize Risk Assessment and Pricing for Australian Insurance Industry.

  • IBM Offers Free 10-Hour Training in AI Fundamentals to Bolster Tech Skills Across Industries.

Too many AI tools, not enough time? We write a guide about 1 top AI tool every week and email it to our premium subscribers. You’ll also get access the full tools database, where you can filter by industry and occupation, plus our AI use case library. Sign up here

Trending Today

Research Review

Microsoft Unveils Phi-2, A Small Language Model That Packs Power

Microsoft's Phi-2, a small language model (SLM), is breaking ground with its compact yet robust capabilities. Surpassing peers like Meta's Llama-2 and even Google's Nano 2, Phi-2 stands out in the generative AI space not just for its size but for its adeptness at tasks typically reserved for heftier models. "State-of-the-art performance" isn't a term thrown around lightly, yet Phi-2 earns this with less than 13 billion parameters, challenging the assumption that bigger is always better.

The strategic training approach of Phi-2, using high-quality textbook data supplemented by select web content, exemplifies Microsoft's intent to refine SLMs. Their goal is clear: to rival the performance of larger scale models without the cumbersome computational demands. This pivot towards efficiency resonates in an industry where the balance of power and practicality is paramount. "A question remains whether such emergent abilities can be achieved at a smaller scale using strategic choices for training," Microsoft reflects, hinting at a future where AI's footprint might shrink, but its impact does not. zdnet

Written by Isaac R. Ward, Casey Clifton, and Alex Brogan.

Send us feedback at [email protected] and help us provide the best coverage of Artificial Intelligence possible.

Interested in reaching smart business leaders like you? To become an Intelligence Age partner, apply here.

Join the conversation

or to participate.