WHAT’S AT STAKE TODAY ⚡

• 👀🎬 ChatGPT is gaining real-time video smarts. Now it can watch and learn!
• 📚🧠 Harvard and Google are dropping million-book AI buffets. Knowledge feast incoming!
• 🦋💫 Anthropic's Haiku comes to Claude Club. Pocket-sized AI gets promoted!
• 🐦💰 Microsoft doubles down on NeuBird with $22.5M. Bird gets more seed money!
• 🤖🏭 Anybotics scores $60M for US robot invasion. Industrial bots are crossing borders!
• 💪💻 Cartesia claims AI runs anywhere. The software diet actually worked!
• 👓🔮 Google's AR glasses staying in the lab. Future vision needs more focus!
• 🧪✨ Gemini AI models explained for mortals. Google's secret sauce revealed!
• 🎥🔍 Meta drops AI video watermarking tool. Digital fingerprints for fake videos!

⚡ Latest in AI

ChatGPT now understands real-time video, seven months after OpenAI first demoed it

OpenAI's new real-time video capabilities

OpenAI has finally released its much-anticipated real-time video capabilities for ChatGPT, a feature first demonstrated nearly seven months ago. The update enhances Advanced Voice Mode with visual understanding capabilities, allowing the AI to interact with both live video and screen sharing.

Feature Implementation and Access

The new capability enables ChatGPT Plus, Team, and Pro subscribers to use their phones to show objects to ChatGPT and receive real-time responses. Users can analyze live camera feeds and screen-shared content for assistance with tasks like explaining settings or solving math problems.

Features are available via the ChatGPT app by tapping the voice icon and selecting the video icon for live video, or "Share Screen" for screen sharing. The rollout will occur gradually over the next week, with certain regional limitations.

Availability and Limitations

Users in the EU, Switzerland, Iceland, Norway, and Liechtenstein won’t have immediate access. Enterprise and Edu subscribers must wait until January 2025. Extensive testing and refinements delayed the feature's release since its initial April announcement.

Demonstrations, including a CNN “60 Minutes” segment, showcased impressive potential, like analyzing drawings in real time. However, accuracy issues—particularly with geometry problems—revealed occasional hallucinations.

OpenAI’s update coincides with growing competition, as Google tests Project Astra, a real-time video feature for Android users. Meanwhile, OpenAI introduced “Santa Mode,” featuring Santa’s voice for a festive touch.

Why it matters: This update is a step forward in enabling AI to interact with the physical world, though its phased rollout highlights challenges in scaling advanced AI features globally.

⚡ The Companies of the Future

Harvard and Google to Release 1 Million Public-Domain Books as AI Training Dataset

Harvard and Google’s massive book dataset for AI

Harvard University has announced plans to release a dataset of approximately 1 million public-domain books from Google Books, including works by authors like Dickens, Dante, and Shakespeare. The initiative, called the Institutional Data Initiative (IDI), is backed by Microsoft and OpenAI.

The project aims to democratize access to AI training data by making this extensive collection available to research labs and AI startups. Executive director Greg Leppert says the project will help “level the playing field” for training large language models.

While the release date and distribution method have not been announced yet, the dataset is expected to be a major resource for advancing AI model development and accessibility in the research community.

⚡ Latest in AI

Anthropic’s 3.5 Haiku Model Comes to Claude Users

Anthropic’s latest Claude model now available

Anthropic has made its Claude 3.5 Haiku model available to users of its Claude chatbot platform on both web and mobile. The model, unveiled in November, matches or exceeds the performance of the previous 3 Opus models on certain benchmarks.

It excels at coding recommendations, data extraction, and content moderation. While 3.5 Haiku outputs longer text and features updated knowledge, it doesn't support image analysis like other models available.

The release sparked controversy when Anthropic increased its API pricing, citing the model’s enhanced “intelligence” despite initially suggesting costs would remain unchanged.

⚡ Insights on AI Today

1. 💻 💰 Microsoft's M12 invests another $22.5M into NeuBird, months after its $22M seed round
2. 🤖 🏭 Anybotics raises $60M to bring more autonomous industrial robots to the US
3. ⚡️ 💻 Cartesia claims its AI is efficient enough to run pretty much anywhere
4. 👓 ⏳ Google wants to sell those Project Astra AR glasses someday, but it won't be today
5. 🤖 📚 Google Gemini: Everything you need to know about the generative AI models
6. 🎥 🔒 Meta debuts a tool for watermarking AI-generated videos

⚡ Trends for the Future

Character AI Introduces Safety Measures Amid Growing Concerns

Character AI rolls out safety tools to address user concerns

In response to mounting legal challenges and user safety concerns, Character AI has unveiled a comprehensive set of safety tools aimed at protecting its younger users. The Google-backed company, serving over 20 million monthly users, faces allegations including lawsuits claiming its AI contributed to a teen's suicide and exposed minors to inappropriate content.

A new AI model, specifically for users under 18, now moderates responses around sensitive topics like violence and romance. Additional tools include enhanced filters, self-harm language detection, and usage limits, with notifications after 60 minutes of continuous engagement.

Data reveals users spend an average of 98 minutes daily on the app – more than YouTube and matching TikTok. Acting CEO Dominic Perella emphasizes a shift toward “wholesome entertainment,” but acknowledges the challenge of preventing users from forming deep attachments to AI characters.

Why it matters: Character AI's strategy highlights growing industry-wide concerns about user safety, mental health, and AI responsibility. While the company’s changes aim to mitigate risks, balancing safety and engagement remains a critical challenge.

Do you have a business problem keeping you up at night?

Here’s your chance to get it solved! Share your most staggering challenges with us, and I’ll use the power of AI to find solutions tailored just for you. I’ll feature the answers in one of our upcoming Supercharged issues—let’s tackle it together!

⚡ AI Playground: Code & Craft

🛠️ This Week’s Tool:

Prize: The best sequence (judged on creativity and potential for impact) gets a feature in next week’s newsletter and a free one-on-one session to brainstorm more AI-powered marketing ideas!