Welcome to Daily Zaps, your regularly-scheduled dose of AI news ⚡️ 

Here’s what we got for ya today:

• 🏎️ Whitehouse orders faster adoption of AI

• 🎥 China ahead of US in AI video

• 🖼️ 50x faster AI image generation

• 🤖 Watermarking AI-generated text

Let’s get right into it!

Sign Up | Advertise | Tools

GOVERNMENT

Whitehouse orders faster adoption of AI

President Biden has signed a national security memorandum directing the Pentagon and U.S. intelligence agencies to adopt artificial intelligence (AI) more rapidly as a matter of national security. The memo aims to maintain U.S. leadership in AI while protecting against threats from adversaries like China.

It outlines guidelines to ensure AI is used in ways that respect civil liberties and human rights, though some civil liberties advocates argue the safeguards are insufficient. The memo also addresses supply chain security and economic espionage, stressing the importance of quickly deploying AI to protect U.S. interests and maintain its competitive edge globally.

Link to full memorandum

WORLD NEWS

China ahead of US in AI video

Chinese AI video-generation apps, like Hailuo, have gained popularity in the U.S., with short clips of Gordon Ramsay going viral on platforms like TikTok and Instagram. Developed by Chinese startup MiniMax, Hailuo is competing with U.S. companies like Runway and Luma AI. These Chinese apps are entering the U.S. market as AI video generation is still in its early stages, providing an opportunity for Chinese companies to establish a strong presence.

Although Chinese firms face challenges, such as censorship concerns, their apps are attracting users with competitive pricing and high-quality output.

STARTUPS

50x faster AI image generation with OpenAI model

Diffusion models have transformed generative AI by enabling the creation of realistic images, 3D models, audio, and video, but they are slow because they require many steps to generate samples. A new method, called sCM (simplified continuous-time consistency models), speeds up this process by reducing it to just two steps while maintaining high-quality results. This method achieves a 50x speed improvement compared to diffusion models, while still producing samples of similar quality.

BIG TECH

Watermarking AI-generated text

Google has made SynthID Text, a tool for watermarking and detecting AI-generated text, generally available to developers and businesses. It can be downloaded from Hugging Face and Google’s Responsible GenAI Toolkit. SynthID Text works by adjusting the probability distribution of tokens in AI-generated text, creating a watermark that can be detected even if the text is paraphrased or modified. Although integrated with Google's Gemini models, it struggles with short text, translations, or factual responses.

Deepmind blog post about SynthID.