The Current Status of OpenAI
PLUS: SneakyPrompt can trick text-to-image models into unwanted content.
Casey Clifton
November 20, 2023
Good morning! Today is Monday, November 20th and we're covering the latest OpenAI leadership changes and Europe's push to gain an edge in AI through expanded supercomputing access. New to The Intelligence Age? Sign up here.
Drop us a line anytime at [email protected] with feedback.
News & Insights
Out For Good
Sam Altman, after being fired as CEO of OpenAI, was in discussions with the OpenAI board about a potential return. However, he will not be returning as CEO. Instead, he and former OpenAI Chairman Greg Brockman will join Microsoft, while Emmett Shear, former Twitch boss, will take over as the interim CEO. Altman's firing and the subsequent management changes were unexpected and led to significant internal upheaval. Some staff members and researchers, such as Szymon Sidor, have left OpenAI following these events. Microsoft, a major backer of OpenAI, is also contemplating taking a role on the board in light of these developments. Reuters, Time
Can Supercomputers Give Europe An Edge In The AI Race?
In an era where AI shapes the future, Europe's strategic move to democratize supercomputing could be a game-changer. Ursula von der Leyen, the Commission president, emphasized the significance: "Europe is a leader in supercomputing... And we need to put this power to use." By offering startups and SMEs unprecedented access to top-tier supercomputing resources, the EU aims to compress AI development timelines dramatically. This initiative seeks not only to bolster Europe's position in the global AI landscape but also to align emerging AI technologies with European values.
Amidst a competitive global scenario, the EU's push for technological sovereignty strikes a chord. Thierry Breton, internal markets chief, remarked, "Our European leadership in high-performance computing... will now translate in scientific, technological and business competitiveness." The Large AI Grand Challenge, with its up to four winners destined to release open-source models, encapsulates this ethos. But, will this open approach to AI development foster the next wave of innovation and ethical AI practice we hope to see on the global stage? siliconrepublic
Together with Kosmik
Transform Teamwork with Kosmik
Ever lost more time organizing information than actually using it? Meet Kosmik, our sponsor and your new workspace ally.
Capture and collaborate on notes, images, or websites—no more endless searching.
Operates offline, keeping your data secure and local.
An infinite canvas lets you browse the web and centralize resources with your team.
What sets Kosmik apart? It adapts to your workflow while bringing your team onto the same literal page.
Elevate your team's productivity and creativity. Learn more
Please support our sponsors!
Mastering AI
Top Tools
Want more tools?
We’ve curated over 3000 AI tools and organized them by occupation and use case to help you be more productive with AI. Access the full library here.
Practical AI
OpenAI’s Custom GPTs: Everything You Need To Know About How To Use Them And Build Your Own
OpenAI's latest offering, custom GPTs, elevates the potential for personalization in generative AI. Users can now fundamentally shape ChatGPT's interactions and access specific datasets. This isn't just a tweak; it's a transformative addition that empowers users to mold AI communication to their needs. Amidst OpenAI's internal shake-ups, custom GPTs emerge as a beacon for the company's future, suggesting a shift towards user-driven innovation.
Custom GPTs are not merely about preset options; they're about integration. By enabling APIs, these AI variants can tap into a myriad of web services, extending their capabilities. The experience might be buggy, and the speed may falter, especially with third-party APIs, but the potential is there. It’s analagous to the evolution of smartphone apps, and we might be at the start of a whole new breed of software. fastcompany
Around The World
More News
Menlo Ventures Secures $1.4 Billion to Boost Artificial Intelligence Investments Amidst a Surge in Tech Funding Activity.
Trending Today
the mission continues
— Sam Altman (@sama)
Nov 20, 2023
Research Review
Text-To-Image AI Models Can Be Tricked Into Generating Disturbing Images
Text-to-image AI models, while sophisticated and widely adopted, are not infallible. Recent research has exposed a vulnerability termed "SneakyPrompt," capable of circumventing the safety measures of models like Stable Diffusion and DALL-E 2. Researchers demonstrated that AI could be manipulated into producing content it's programmed to avoid. This revelation isn't just about ethical use; it underscores a significant security flaw.
"SneakyPrompt works by tweaking the AI's language recognition, turning innocuous tokens into requests for prohibited content," says Yinzhi Cao. The technique poses a stark reminder: our AI systems are only as robust as their safeguards. And while companies like Stability AI are developing new defenses, the arms race between safety measures and those intent on breaking them continues. techcrunch
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.