31 | AI: Can I Recognize You from a Picture? (1)

Our Menu :)

• I-XRAY: The AI Glasses That Reveal Anyone’s Personal Details

• Meta Ups the Ante with Movie Gen: Revolutionizing Text-to-Video Generation

• Glean who?

Jobs and Training

• Job Listings this week

• Andrew Ng’s AI Python for Beginners.

Read time: 5 Minutes

WEEKLY DIGEST 
I-XRAY: The AI Glasses That Reveal Anyone’s Personal Details 

A few weeks ago, Meta unveiled an intriguing feature for its new Ray-Ban Meta glasses: the ability to gather information about your surroundings. This includes innocuous tasks like identifying flowers. However, two Harvard students recently demonstrated how easily these smart glasses can turn into a privacy nightmare.

Anhphu Nguyen and Caine Ardayfio developed an app called I-XRAY that transforms the Ray-Bans into a powerful doxxing machine. This app can reveal personal details such as names, addresses, and phone numbers just by looking at someone through the glasses.

How Does It Work?

The Ray-Bans can record up to three minutes of video, with a privacy light that's barely noticeable—akin to a firefly in broad daylight. This video is streamed to Instagram, where an AI monitors the feed. I-XRAY then uses PimEyes, a facial recognition tool, to match these faces to public images and employs AI to dig up personal details from public databases.

The duo's demo had strangers alarmed when they realized how easily identifiable they were from public online information. They tested the app on unsuspecting subway riders, convincingly pretending they knew them using details from I-XRAY. (More like I-SCARY.)

You can see the app in action here: Watch the Demo.

Fortunately, Nguyen and Ardayfio are not releasing the app. It's just a demonstration of how easy this is to do. But if two students can create this, you can bet someone else will, too.

Big Tech's Role

While big tech companies like Meta might not be the ones developing such invasive apps, they are certainly involved in other ways. Meta has been evasive about whether they're using your glasses' pics to train their AI. When asked, they simply said, “We're not publicly discussing that.”

Protecting Your Privacy

If you're concerned about your privacy, the Harvard duo shared tips on how to scrub yourself from facial recognition databases. You can find these tips in this Google Doc.

Main Services to Opt Out Of

Here are the main services they suggest you opt out of (with links to directly opt out):

1. PimEyes - Opt Out

2. FaceCheck.ID - Opt Out

3. FastPeopleSearch - Opt Out

4. Instant Checkmate - Opt Out

Stay informed and take steps to protect your privacy in this increasingly digital world.

WEEKLY DIGEST
Meta Ups the Ante with Movie Gen: Revolutionizing Text-to-Video Generation

Meta has introduced Movie Gen, a groundbreaking series of four systems that generate videos with consistent characters, alter imagery, and add matching sound effects and music. This innovative technology will be available on Instagram in 2025, but you can already view and listen to examples here. The team has detailed their work in an extensive 92-page paper.

Key Features

• Generated Videos: Movie Gen Video outputs 256 frames (up to 16 seconds) at 1920x1080 resolution. It uses flow matching to produce detailed imagery from text prompts.

• Consistent Characters: Given an image of a face, Movie Gen Video generates a video depicting a person with that face, with varied poses and expressions.

• Altered Clips: The system can alter videos by changing backgrounds, adding objects, or restoring altered clips to their original content.

• Synthetic Soundtracks: Movie Gen Audio generates sound effects and music for video clips up to 30 seconds long, starting from pure noise and extending audio as needed.

Performance Highlights

• Overall Video Quality: Movie Gen outperformed competitors like Runway Gen3, Sora, and Kling 1.5 in qualitative evaluations.

• Character Generation: Achieved a net win rate of 64.74 percent versus ID-Animator.

• Soundtrack Generation: Outperformed various video-to-audio models with a net win rate between 32 percent and 85 percent.

• Video Alteration: Beat all competitors more than 70 percent of the time in the TGVE+ dataset.

WEEKLY DIGEST
Glean who?

Rumors in valley suggest that Sam Altman, the CEO of OpenAI, has identified five AI startups that OpenAI's recent investors should avoid backing:

🛑 Anthropic
🛑 xAI (Elon Musk)
🛑 Safe Superintelligence (Ilya Sutskever)
🛑 Perplexity
🛑 Glean

While the first four are well-known, Glean AI stands out as a lesser-known but formidable competitor. Let's dive into what makes Glean a potential future winner in the enterprise AI space.

Glean AI: The Work AI Platform

Glean AI, founded by a team of former Google search engineers and industry veterans, is a Work AI platform connected to enterprise data. It allows users to find, create, and automate anything, making it a powerful tool for enhancing workplace productivity.

• Revenue Growth: Glean's annualized revenue grew to $39 million in January 2024, marking a 4x growth from the $10 million recorded in 2023.

• Funding: In September 2024, Glean raised over $260M in a Series E, valuing the company at $4.6 billion.

• Founders: Arvind Jain (CEO), T.R. Vishwanath (CTO, formerly at Meta and Microsoft), Tony Gentilcore (Product Engineering, formerly at Google), and Piyush Prahladka (now CEO of his own stealth startup).

Glean's Suite of Products

• Work AI Platform: A unified search tool that aggregates information from various applications, allowing users to access critical data efficiently.

• AI-Powered Workspace Search: Enhances workplace productivity by providing enterprise-grade AI and search capabilities.

• Customer Testimonials and Stories: Real-world examples of how Glean is transforming workplaces.

The Future of Enterprise AI Search

Glean specializes in enterprise-grade AI and search capabilities that have the potential to scale well. The startup focuses on enhancing workplace productivity by providing a unified search tool that aggregates information from various applications.

Arvind Jain, co-founder and CEO of Glean, has a vision to expand human potential to do extraordinary work. The trajectory of Glean AI follows an expected path, and I expect it to become one of the first winners in the pure-play RAG (Retrieval Augmented Generation) category.

Conclusion

As the enterprise AI search space becomes more consolidated, Glean AI is poised to emerge as a leader. Its innovative approach to enhancing workplace productivity and its impressive revenue growth trajectory make it a startup to watch.

GEN AI STARTUP

Waymo is seeking a Backend Software Engineer for their Simulation Infrastructure team.

Mistral AI is seeking a Machine Learning Engineer to join their team.

Databricks is hiring a Sr. Corporate Communications Manager to join their team.

xAI has an opening for an Office Manager to oversee daily operations.

JOBS & TRAININGS
Andrew Ng’s AI Python for Beginners.

Working With Your Own Data and Documents in Python

• Learn how to handle both structured and unstructured data using Python, from reading text files to processing data in CSV format for tasks like vacation planning.

• Use Python and AI to process large volumes of text, such as extracting relevant details from food critics' journals or other documents, and generate useful insights.

• Turn repetitive code blocks into reusable Python functions to make your code more modular and efficient.

• Combine data processing techniques with AI to generate detailed travel itineraries, integrating multiple data sources.

Thank you, see you next week!