AI Developments and Regulatory Changes

This chapter explores recent advancements in AI, particularly Google's DeepThink for premium users, while discussing the EU AI Act's impact on compliance and transparency for developers. It highlights breakthroughs in deepfake detection with a new system achieving 98% accuracy, addressing concerns about trustworthiness in media. Additionally, ethical dilemmas surrounding AI's interaction with web scraping and content verification are raised, emphasizing the need for integrity in AI applications.

Transcript

chevron_right

Play full episode

chevron_right

Transcript

Episode notes

Want to keep the conversation going?

Join our Slack community at thedailyaishowcommunity.com

Intro

For episode 523 on August 6, The Daily AI Show takes a fantasy-themed journey through the latest AI news. The team breaks down major releases, legal battles, and research breakthroughs—covering everything from OpenAI’s new open-source models and the EU AI Act, to Google DeepMind’s Genie 3, AI music generation, and the new arms race in deepfake detection.

Key Points Discussed

• OpenAI releases OSS, a set of open-source models including 120B and 20B parameter versions. The 20B can run on a laptop, and Microsoft has integrated it into Windows. ChatGPT weekly users have soared to 700 million, with OpenAI’s valuation now over $500 billion.

• Google launches “Deep Think” for ultra subscribers, and the company keeps democratizing high-end models, but with clearer pricing tiers and more exclusivity at the top.

• The EU AI Act officially launches, bringing strict transparency, documentation, and copyright rules to any AI products operating in the EU.

• A joint project between UC Riverside and Google achieves a breakthrough in deepfake detection: a universal video deepfake detector that works in real time and recognizes more than just faces—hitting 98% accuracy.

• Cloudflare calls out Perplexity for scraping sites against explicit wishes, while Perplexity pushes forward with OpenTable integration and a multi-agent orchestration platform after acquiring Invisible.

• The show debates public vs. private data, web scraping ethics, and the shifting business of open information online.

• 11 Labs launches AI music generation trained on licensed datasets from major indie labels, promising a copyright-safe option for creators—and raising the bar for what’s possible with AI-generated audio.

• Google DeepMind’s Genie 3 brings prompt-based world building to the next level: real-time, persistent AI-generated environments for gaming, XR, and research. The team speculates on how these world models will shape games, training, and the future of “massive single-player online” experiences.

• Open-source LLMs are now more accessible than ever, and Anthropic quietly releases Claude 4.1, a major update for coding and agentic tasks.

• AI research is reshaping science, from meteorite materials that could power future wearables and neuromorphic computing, to battery breakthroughs that cut out rare earth metals.

Timestamps & Topics

00:00:00 🏰 Fantasy intro and this week’s AI news journey

00:03:36 ⚡ Lightning round: OpenAI’s OSS models, user stats, and valuation

00:05:46 💡 Google Deep Think and the new era of model exclusivity

00:07:10 🇪🇺 EU AI Act goes live: key rules and global impact

00:10:27 🕵️‍♂️ Deepfake detection breakthrough at UC Riverside & Google

00:16:38 🤖 Perplexity, Cloudflare, web scraping, and agentic features

00:20:41 🍽️ Perplexity’s OpenTable integration and multi-agent roadmap

00:29:19 💻 Public data, paywalls, and the arms race over online info

00:32:10 🎵 11 Labs AI music—copyright, new genres, and creator tools

00:42:33 🌍 DeepMind Genie 3 and the rise of prompt-driven world building

00:55:33 🔬 Open-source LLMs, Anthropic Claude 4.1, and AI in science