🚀 EU Supercomputers Empower AI Startups

PLUS: Tapping unstructured data key to boosting generative AI.

Good morning! Today is Wednesday, December 20th and in today's news we cover the EU granting AI startups access to supercomputing resources, OpenAI's new safety framework for advanced models, and research showing current limits of language models in analyzing complex financial filings. New to The Intelligence Age? Sign up here.
Drop us a line anytime at [email protected] with feedback.

News & Insights

EU To Expand Support For AI Startups To Tap Its Supercomputers For Model Training

The European Union is amplifying its support for AI startups by granting them access to its supercomputing capabilities. This move aims to bolster the model training processes of these startups, potentially bridging the compute gap between European and U.S. tech giants. An EU official revealed, "We’re developing facilities for our SMEs... to develop their models," emphasizing the need for tailored AI algorithm support. This strategic shift underscores the EU's commitment to nurturing a competitive, ethical, and trustworthy AI landscape within its borders.

The initiative includes setting up centers of excellence to facilitate the startups' understanding and utilization of supercomputing resources. "AI is considered now a strategic priority for the Union," an official stated, signaling a clear focus on empowering local AI enterprises. With the anticipated inauguration of a pre-exascale supercomputer and plans to interlink a network of supercomputers, the EU is creating a robust infrastructure to support AI advancements and ensure European startups remain at the forefront of the AI revolution. techcrunch

OpenAI Unveils New Safety Plan For Frontier AI Models. How It'll Impact Future Development

In a strategic move, OpenAI has introduced its Preparedness Framework aimed at ensuring the safety of its frontier AI models. This comprehensive safety plan outlines continuous evaluations of AI capabilities, assessing risks and the effectiveness of mitigation strategies. The results are encapsulated in a dynamic risk scorecard, categorizing threats into four levels—low, medium, high, and critical—which dictate further development or deployment actions. This systematic approach reflects OpenAI's commitment to setting industry benchmarks in responsible AI development.

The internal restructuring of OpenAI's decision-making process, with a dedicated Preparedness team and a Safety Advisory Group, underscores a shift towards greater accountability and oversight. The Board of Directors now possesses the authority to override leadership decisions, a change following recent leadership challenges. The framework also emphasizes proactive collaboration with external entities to monitor misuse and invests in research to understand risk progression as models scale. zdnet

Why does this matter?

Consider the implications of OpenAI's risk scorecard beyond immediate safety—it may serve as a template for regulatory bodies seeking quantifiable AI safety measures. This could lead to a standardized 'AI safety certification' that becomes a prerequisite for industry-wide deployment, possibly reshaping the future landscape of AI credibility and trust.

GPT And Other AI Models Can't Analyze An SEC Filing, Researchers Find

Large language models (LLMs) like GPT-4 hit a wall when it comes to parsing complex SEC filings. Researchers from Patronus AI found that even the most advanced AI configurations couldn't reliably interpret these financial documents. GPT-4-Turbo, for instance, achieved only a 79% accuracy rate in the best-case scenario. This shortfall exposes the current limitations of LLMs in sectors where precision is non-negotiable, such as finance.

The quest for AI that can effortlessly sift through dense financial data is far from over. As businesses in regulated industries eye AI for tasks ranging from customer service to data analysis, the findings underscore the need for improvement. While LLMs have shown promise in understanding financial narratives, Patronus AI's research reveals a stark reality: without significant advancements, AI cannot yet be trusted to independently navigate the complexities of financial documents. cnb

Together with Bardeen

Delegate your tedious work with a few lines of text

Bardeen was voted #1 product of the month at Product Hunt and is trusted by thousands of users.  

You can automate almost any task you can think of across Sales, Marketing, Engineering, Recruiting, Personal Productivity, and more. 

  • Let AI build automations for you by typing in your desired workflows

  • Set automatic triggers to run when something happens or on a schedule

  • Extract data from websites and send it to web apps with a web scraper

Please support our sponsors!

Mastering AI

Top Tools

  • Aragon: Professional headshots generated instantly through A.I. technology. (website) (sponsored)

  • Verb Art: A service where users can generate art based on their voice through the use of AI technology. (website)

  • Simplified.: AI-powered marketing app with design, video, copy, and social tools. (website)

Practical AI

How To Use Bing Image Creator (and Why It's Better Than Ever)

Microsoft's recent integration of DALL-E 3 into Bing Image Creator marks a significant leap in AI-powered image generation. With this upgrade, users now benefit from enhanced prompt accuracy and richer image detail, streamlining the creative process. Accessible via Bing Chat or directly on the tool's website, crafting detailed visual prompts transforms into vivid images in a matter of seconds, all without the prerequisite of an OpenAI account.

The tool's simplicity extends to its usability—log in, craft a prompt, and create. For example, a request for "a dodo bird sitting on a concrete floor in a brightly lit tropical home" yields surprisingly accurate results despite the bird's extinction centuries ago. This speaks to the AI's adaptability and evolving understanding of historical and fantastical subjects. zdnet

Why does this matter?

While Bing Image Creator thrives with imaginative prompts, its potential as a tool for rapid prototyping in design industries remains untapped. Imagine fashion designers using it to visualize new patterns on fabrics instantly or architects to populate renderings with context-specific flora and fauna.

Around The World

More News

  • OpenAI Releases Comprehensive Safety Plan to Guide the Responsible Evolution of Advanced AI Models, Influencing Industry Standards and Future Innovation.

  • AI Expert Andrew Ng Poses a Provocative Question on the Impact of Increasing Intelligence in the World through AI Advancements.

  • Former Google CEO Eric Schmidt Offers Strategic Advice to Combat Misinformation Ahead of the 2024 Elections.

Trending Today

Research Review

In Search Of The Missing Piece Of Generative AI: Unstructured Data

The untapped potential of unstructured data is becoming increasingly evident as organizations realize its value in powering generative AI. Despite the prevalence of unstructured data, which constitutes about 90% of enterprise information, only a fraction of organizations actively leverage it. The shift in focus from structured to unstructured data is imperative for the advancement of generative AI, a sentiment echoed in MIT Technology Review Insights, emphasizing the urgency of a flexible, scalable, and efficient data infrastructure.

A recent global survey underscores the critical nature of unstructured data; nearly 70% of technology executives agree that data issues could derail their AI and machine learning ambitions. Generative AI's capacity to surface and utilize hidden data is unlocking new organizational advances. However, the challenge lies in integrating fragmented IT architectures and making previously inaccessible data available for analysis. "Mergers and acquisitions have resulted in fragmented IT architectures...locked in offline proprietary file types," remarks Green in the MIT report, highlighting the necessity for a unified data approach to empower generative AI. thenextweb

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.