OpenAI Vision and OpenAI Gobi

Welcome to Daily Zaps — Here’s what we got for you today: 🙌

OpenAI Vision and OpenAI Gobi

OpenAI is preparing to launch GPT-4's image understanding feature, called "GPT-Vision," in response to Google's upcoming multimodal language model, Gemini. This feature, demonstrated during GPT-4's unveiling, allows the model to interpret images.

OpenAI aims to make image understanding more widely available, potentially integrating it with DALL-E 3. These developments could be announced at OpenAI's developer conference on November 6. OpenAI is also considering a new multimodal AI model called "Gobi," which could become GPT-5 in the future.

38TB of data accidentally exposed by Microsoft AI researchers

Microsoft's AI research team accidentally exposed 38 terabytes of private data, including secrets, private keys, passwords, and over 30,000 internal Microsoft Teams messages. They shared these files using an Azure feature called SAS tokens, which inadvertently granted access to the entire storage account. While this incident was not a direct exposure to the public, it highlighted the security risks associated with SAS tokens and the need for proper governance and monitoring.

Microsoft’s account of this issue. “To reiterate, the information that was exposed consisted of information unique to two former Microsoft employees and these former employees’ workstations. No customer data was exposed, and no other Microsoft services were put at risk because of this issue. Customers do not need to take any additional action to remain secure.”

Together With Our Partners

Instantly build an AI chatbot with your data

Enter a website URL or upload any PDF file to create a ChatGPT powered custom chatbot in 2 minutes.

Here’s how it works:

• 📖 Toss in your company website, PDFs, or even a YouTube video, and watch as Dropchat works its magic and turns your content into an AI powered sidekick.

• 💬 Didn’t read the material? No problem, chat it up and Dropchat will give you all the answers.

• 📝 Want to challenge yourself? Create interactive flashcards for a self-quiz session.

• 🔗 Share the chatbot via code snippet or URL with students, employees, customers, and partners.

• 🌐 Add the chatbot to websites for 24/7 AI powered sales, marketing, and tech support.

Get started with Dropchat.

Google launches Bard Extensions

Google is expanding its Bard AI tool to integrate with core apps and services like Gmail, Docs, Google Drive, Google Maps, YouTube, Flights, and Hotels. Bard extensions allow users to ask questions and request information from these services within a single conversation.

Bard extensions are opt-in, and Google promises that personal information won't be used for ads or to train Bard. The tool also offers a verification feature and collaborative chat sharing.

UK’s 100+ page report on A.I. consumer protection and antitrust

The U.K. government’s Competition and Markets Authority (CMA) has released a 100+ page report outlining proposed seven guiding principles for the responsible development and use of foundation models (FMs), which are versatile AI systems with broad capabilities.

1. Accountability – FM developers and deployers are accountable for outputs provided to consumers.

2. Access – ongoing ready access to key inputs, without unnecessary restrictions.

3. Diversity – sustained diversity of business models, including both open and closed.

4. Choice – sufficient choice for businesses so they can decide how to use FMs.

5. Flexibility – having the flexibility to switch and/or use multiple FMs according to need.

6. Fair dealing – no anti-competitive conduct including anti-competitive self-preferencing, tying or bundling.

7. Transparency – consumers and businesses are given information about the risks and limitations of FM-generated content so they can make informed choices.

The CMA emphasizes that effective development and use of FMs can lead to innovation, economic growth, and benefits for consumers and businesses. However, it also warns of potential harms if competition is weak or consumer protection laws are not followed, including misinformation and market power consolidation.

Full pdf report here