🌐 International AI Security Guidelines Published

PLUS: DeepMind's AI proposes 380,000 new stable materials.

Good morning. Today is Thursday, November 30th and today we're looking at new international AI security guidelines and a technique to improve AI answer quality using the latest online data. New to The Intelligence Age? Sign up here.
Drop us a line anytime at [email protected] with feedback.

News & Insights

New AI Security Guidelines Published By NCSC, CISA & More International Agencies

AI systems are evolving at an unprecedented pace. Security, however, cannot play catch-up. The NCSC and CISA, along with 16 other countries, have established guidelines ensuring AI security from the ground up. These Guidelines for Secure AI System Development detail a life cycle approach: secure design, development, deployment, and operation. The consensus is clear: security must be integrated, not an afterthought.

The guidelines address all AI system types and stakeholders. They align with the principles set forth at the AI Safety Summit. Lindy Cameron of the NCSC states, "AI is developing at a phenomenal pace...there is a need for concerted international action." This action is now embodied in guidelines that foster a secure, global digital landscape. Read the guidelines.

How Google And OpenAI Prompted GPT-4 To Deliver More Timely Answers

Generative AI's ability to deliver current information has often hit a wall with data cut-offs. OpenAI's GPT-4, for instance, originally drew from a knowledge pool only as recent as 2021. Google and OpenAI's FreshLLM project uses a new prompting method, FreshPrompt, to blend top Google search results into GPT-4's input, enabling it to respond with up-to-the-minute data. This method has propelled GPT-4's accuracy from a dismal 8% to an impressive 70.2% for recent facts.

In testing the boundaries of AI's knowledge with real-world, current facts, 600 FreshQA questions were crafted. The questions ranged from static to dynamic, challenging AI to sift through 'fresh knowledge' efficiently. The FreshPrompt technique not only improved GPT-4's accuracy dramatically but also reduced outdated or incorrect answers. "This dramatically diminishes the presence of outdated and hallucinated answers," note the researchers. zdnet

Together with Attention

Your Attention, Please

Sales teams face challenges in understanding deal losses and effective strategies. Attention's Generalized Insights provides a solution using AI to analyze sales interactions.

It delivers clear, actionable insights from calls and emails, helping to pinpoint why deals are lost and how to replicate your top sales reps.

This platform transforms complex qualitative data from your customer touchpoints into understandable reports, enhancing your team performance.

Learn more here.

Please support our sponsors!

Mastering AI

Top Tools

  • AI Audience Accelerator: If you’ve ever wanted to build an audience, AI Audience Accelerator will help you—in a fraction of time time. (website) (sponsored)

  • Meeple.ai: AI-powered one-pagers to improve follow-up and boost sales. (website)

  • ListingCopy.ai: AI creates ad copy for real estate professionals, saving time. (website)

Practical AI

Around The World

More News

  • AWS Unveils Neptune Analytics, Integrating Vector Search with Graph Data for Enhanced Query Performance and Data Insight.

  • Google DeepMind's Advanced AI Tool Pioneers the Discovery of Over 700 Novel Materials with Potential Applications in Technology and Industry.

Trending Today

Research Review

An AI Dreamed Up 380,000 New Materials. The Next Challenge Is Making Them

Google DeepMind's recent feat, dreaming up 380,000 stable material candidates, marks a significant leap in material science, pushing the boundaries of what's possible. The AI, named GNoME, used data from the Materials Project to propose new crystal structures, a staggering number that far exceeds traditional discovery methods. But here's the catch: validation. Synthesizing these materials in a lab is the real test, determining their practicality and potential for future applications.

The process involves a methodical dance between AIs and robots — a synergy of digital intelligence and physical precision. DeepMind's methodology sidesteps human bias, offering a more diverse range of materials than ever before. Yet, the challenge remains in the translation from theory to tangible. Of the 380,000, a fraction will be feasible for synthesis, given constraints like rarity of elements or extreme production conditions. crunchbase

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.