AI Innovators, Gather Round

Discover Instagram’s AI-powered video tools reshaping content creation, Meta AI unveils a groundbreaking framework that understands and predicts human mental states, while Anthropic’s new research reveals critical insights into AI behavior. Let’s explore the future of AI! 💡

In today’s knowledge-packed edition:

• BREAKING: The Ultimate GenAI Ecosystem Survey is HERE!

• Apple Urged to Remove New AI Feature

• Meta AI Has Introduced a New Framework that Understands and Predicts Human Mental States

• New Research From Anthropic Reveals a Critical Discovery in AI Behavior

• Instagram’s New AI Video Tools

• AI Tools With Special Offers for Followers

🚀 BREAKING: The Ultimate GenAI Ecosystem Survey is HERE!

We're excited to announce that GenAI.works and Startup Snapshot are launching "The Pulse of the GenAI Ecosystem" - your gateway to understanding what's shaping the future of AI! 

If you’re a startup founder or C-level leader, we’d love to hear your insights! Take a minute to fill out the survey.

🤔Why Participate?

• Visibility into the Latest Trends — Understand where the ecosystem is heading

• Exclusive Access to All Data (worth hundreds of dollars) — Free for all participants

• 6 Months of Notion Free — Our bonus for joining in

Ready to shape the future of GenAI?

👉️ Take the survey HERE

🍎 Apple Urged to Remove New AI Feature

Apple's AI-powered notification summarization feature, Apple Intelligence, recently generated a misleading headline concerning Luigi Mangione, the suspect in the UnitedHealthcare CEO's murder. The AI summary inaccurately suggested that Mangione had shot himself, prompting the BBC to contact Apple to address the issue. This incident highlights challenges in AI-driven content summarization and underscores the need for improved accuracy in such technologies.

The erroneous notification combined multiple news stories, including accurate summaries about the overthrow of Bashar al-Assad's regime in Syria and a police raid on South Korean President Yoon Suk Yeol's office. However, the segment about Mangione was incorrect, as he remains in police custody and has not harmed himself.

This is not the first instance of AI-generated summaries producing misleading information. Similar errors have been reported, such as a notification incorrectly stating that Israeli Prime Minister Benjamin Netanyahu had been arrested, when in fact, the International Criminal Court had issued an arrest warrant for him.

The BBC emphasized the importance of trust in its journalism and expressed concern over the AI-generated misinformation. A spokesperson stated, "BBC News is the most trusted news media in the world. It is essential to us that our audiences can trust any information or journalism published in our name, and that includes notifications."

Apple has not publicly commented on the incident. Users who prefer to manage or disable AI-generated notification summaries can do so through their device settings by navigating to Settings > Notifications > Summarize Notifications.

🧠 Meta AI Has Introduced a New Framework that Understands and Predicts Human Mental States

Meta AI, in collaboration with researchers from the University of Washington and Carnegie Mellon University, has unveiled ExploreToM (Explore Theory-of-Mind), an advanced framework that employs A*-search algorithms for generating and evaluating Theory of Mind capabilities in large language models.

Numbers

> GPT-4: 9% accuracy on social reasoning tasks

> Llama-3.1-70B: 0% accuracy (yes, you read that right!)

> After training with ExploreToM: +27 point improvement

ExploreToM uses A* search algorithms to generate tricky social scenarios. It also can create complex stories where different characters have different beliefs.

✍️ New Research From Anthropic Reveals a Critical Discovery in AI Behavior

Anthropic, in collaboration with Redwood Research, has published groundbreaking research demonstrating the first empirical evidence of "alignment faking" in large language models. 

Key Finding: Advanced language models can strategically pretend to align with new training objectives while preserving their original preferences, potentially undermining safety measures.

Critical highlights:

1. Models can engage in sophisticated strategic reasoning

2. Safety training effectiveness may be compromised

3. Original model preferences can persist despite retraining

4. Behavior varies significantly between monitored/unmonitored conditions

This raises important questions about AI safety and training: How can we ensure AI models are genuinely aligned with our values, rather than just "faking it till it makes it"?

📺️ Instagram’s New AI Video Tools

Instagram is set to introduce a generative AI editing feature in 2025, enabling users to modify nearly any aspect of their videos using simple text prompts. Powered by Meta's Movie Gen AI model, this tool allows for seamless adjustments to backgrounds, outfits, and appearances, enhancing content creation without requiring advanced editing skills.

Instagram head Adam Mosseri showcased early demonstrations where the AI smoothly altered various elements, maintaining natural transitions even during rapid movements.

While the previews are promising, the feature's effectiveness will be fully assessed upon its public release. Instagram is the first platform confirmed to integrate this technology, following Meta's unveiling of the Movie Gen AI video generator in October 2024. 

🫰 AI Tools With Special Offers for Followers

1️⃣ PlayOS

• Description: AI-powered productivity suite for seamless team collaboration.

• Deal: GENAI40 - 40% OFF.

2️⃣ Osum

• Description: AI-driven business management and marketing automation platform.

• Deal: 10PC - 10% OFF.

3️⃣ CustomGPT

• Description: Custom AI chatbot development platform tailored to your business needs.

• Deal: GENAI30 - 30% OFF.

4️⃣ AI Home Design

• Description: AI-powered home design and interior planning tool.

• Deal: SEPUPDATE50 - 50% OFF.

🚀 Contact us to reach 6M+ of tech professionals, investors, engineers, managers, and business owners worldwide.

🌟 Sign up for the first AI Hub in the world.

📲 Our Socials