Daily AI Newsletter: New AR/AI glasses and an AI video generator for Snapchat

In today’s email:

😍 OpenAI is increasing the limits for o1
🇺🇸 California Passes Election ‘Deepfake’ Laws, Forcing Social Media Companies to Take Action
👀 LLMs Will Always Hallucinate, and We Need to Live With This
🧰 9 new AI-powered tools and resources. Make sure to check the online version for the full list of tools.

Snap announced new AR/AI glasses and an AI video generator.

Snap, the parent company of Snapchat, is expanding its AI tools for creators by introducing several new features at its annual Snap Partner Summit. One of the key announcements was Easy Lens, a tool that allows users to create Snapchat Lenses by simply describing them in plain English. The tool, available in Snap's developer platform Lens Studio, uses generative AI to turn text descriptions into 3D effects, characters, and transformations for photos and videos. Additionally, Snap introduced Body Morph, a tool for creating 3D characters and outfits from text or images, along with new image generators and hundreds of new Bitmoji animations.

Snap also introduced the fifth generation of Spectacles, their new standalone AR glasses powered by the innovative Snap OS. These lightweight, see-through glasses allow users to experience AR Lenses in the real world while interacting with friends in new ways. The Spectacles Developer Program offers developers access to the hardware and Snap OS, with tools like the new Spectacles app enabling features such as using your phone as a controller or spectator mode for friends without Spectacles. Spectacles are packed with cutting-edge hardware, including four cameras, a Snap Spatial Engine for hand tracking, and proprietary LCoS micro-projectors that produce sharp images. Their advanced waveguides allow users to see AR visuals without the need for complex calibrations, and the dual Snapdragon processors provide efficient performance with reduced power consumption. The glasses also feature a 46-degree diagonal field of view and automatic tinting for both indoor and outdoor environments.

Snap also announced a new AI-powered video-generation tool, currently in beta for select creators, that will allow users to generate videos from text prompts, with plans to extend to image prompts in the future. All AI-generated content will feature watermarks to ensure transparency when AI is used. This move sets Snap ahead of competitors like TikTok and Instagram, which have yet to release similar AI video-generation features, though Snap faces competition from companies like Runway, OpenAI, and Adobe, who also plan to release their own AI video tools.

Snap CEO Evan Spiegel highlighted the transformative impact of generative AI on the platform, emphasizing its role in driving creator growth. The number of creators publishing Lenses on Snapchat has risen to 375,000, with over 4 million Lenses now available across various platforms. Snap continues to innovate with AI, aiming to enhance user experiences and provide creators with more powerful tools to bring their ideas to life.

Finally, a developer tool that makes digging around for the answer a thing of the past.

Other developer tools can’t tell you how your codebase works and why. Unblocked can. We augment your source code with context from Slack, Confluence, Jira (and more), so your team gets helpful answers without waiting on or interrupting a coworker.

Unblocked gives you:

Instant answers about every aspect of your codebase
Relevant, historical context for any file open in an IDE
A Slack integration that answers questions from your coworkers so you don’t have to

⌛Engineers who use Unblocked save an hour or more a day.

“My team loves that they can get answers and solutions immediately, which means less time spent in confusion and more time creating.” David Knell - Distinguished Platform Architect at Drata

Get started for free

Cold War spy satellites and AI detect ancient underground aqueducts

Archaeologists are using artificial intelligence (AI) and Cold War-era US spy satellite imagery to rediscover ancient underground aqueducts, known as qanats, that helped civilizations thrive in some of the world's most arid regions. These qanats, which are up to 3000 years old, were designed to transport water from mountains or highlands to desert areas, enabling agriculture and settlement in otherwise inhospitable environments. Many of these aqueducts, which stretch across North Africa and the Middle East, have been lost over time.

A team led by Hector Orengo from the Catalan Institute of Classical Archaeology in Spain trained an AI model to detect the vertical shafts of qanats by analyzing black-and-white satellite images taken by US spy satellites between 1959 and 1986. The AI was further trained using artificial images of aqueducts, enhancing its ability to identify these ancient structures. When tested in areas with known qanats, such as Afghanistan, Iran, and Morocco, the AI correctly identified over 88% of qanat locations and 62% of all underground aqueducts.

While it remains uncertain if the technique can be applied to more complex landscapes beyond desert regions, the research has shown the potential of AI in archaeology. As Mehrnoush Soroush from the University of Chicago points out, even without 100% accuracy, the method allows archaeologists to narrow down large areas for closer study, marking a significant advancement in locating ancient water systems.

California Passes Election ‘Deepfake’ Laws, Forcing Social Media Companies to Take Action

California has enacted three new laws targeting the use of AI-generated "deepfakes" in elections. Signed by Governor Gavin Newsom, these laws require social media platforms to moderate and label deceptive election-related deepfakes, particularly those aimed at candidates or election outcomes. The laws aim to curb the spread of misinformation during the election period and include requirements for social media companies to take down or label these AI-generated fakes within 72 hours of a complaint. Though these regulations only apply to California, they could serve as a model for other states seeking to regulate AI-generated content in politics.

One law, effective immediately, bans the sharing of deceptive deepfakes in the 120 days leading up to an election and 60 days afterward. Another, taking effect in January, will require political ads using AI-generated content to carry labels indicating that the content is manipulated. These new laws mark California’s continued efforts to lead in AI regulation, following other states that have also implemented rules around deepfakes, particularly in elections and pornography.

While these regulations aim to protect voters from being misled, they are expected to face legal challenges. Digital rights groups, such as the Electronic Frontier Foundation, argue that enforcing these laws could infringe on free speech, as differentiating between deceptive speech and satire may be difficult. Despite concerns, supporters argue that the laws focus on transparency and do not impose financial penalties, potentially avoiding conflict with federal laws that protect online platforms.

Other stuff

OpenAI is increasing the limits for o1
Google will begin flagging AI-generated images in Search later this year
OpenAI is opening a subsidiary company in Paris
LLMs Will Always Hallucinate, and We Need to Live With This
Hume AI launched Empathic Voice Interface 2
AI can't cross this line and we don't know why.
Carpentopod: A walking table project
Mistral launches a free tier for developers to test its AI models

Automate Calls and Boost Conversions with AI Voice Assistants

Set up an AI receptionist (on 24/7) or an outbound lead qualifier (#Speedtolead). Book appointments, transfer calls, and extract info seamlessly. Integrates with HubSpot, GohighLevel, and more. Deploy in minutes, no coding needed.

Try AI Phone Calling

Win 1 year free Superpower ChatGPT Pro🤑