Episode Summary
Executive Summary: In this episode of This Week in Startups, Jason Calacanis and Sandeep Madra explore and demo several recent AI innovations. They cover Hume AI's video-based emotion/sentiment analysis, chip leaks regarding the iPhone 16 and A18 chip, and the game-development platform Forever AI, which uses AI to remix games. They also discuss new ChatGPT features that allow integrated actions with various GPTs, and Hugging Face's new open-source assistants ecosystem. Additionally, the duo uses 11 Labs text-to-speech GPT within ChatGPT to generate a short story audiobook. The conversation expands into broader societal themes, including content authentication, ethics around AI tracking, and the debate over anonymity and doxxing on social media platforms.
Main Topics: AI Emotion and Attention Analysis (Priority: 4/5): Sandeep demos Hume AI, which analyzes video clips to predict call quality, attention, and mood, classifying traits like self-confidence using models trained on 80 videos. Jason questions the accuracy and implications for coaching, sales, and potential dystopian uses in surveillance. Generative AI in Gaming: Forever AI (Priority: 5/5): The hosts demo Forever AI, a platform that lets users remix casual games by editing code with natural language prompts (e.g., 'make the ship shoot dual missiles') and replacing visuals with AI-generated images. They compare it to YouTube's democratization of video, predicting a similar explosion for individual game creators. ChatGPT's Integrated GPT Ecosystem (Priority: 4/5): A new ChatGPT feature allows users to invoke specialized GPTs (like @kayak) directly from the command line to get flight options, with plans for broader functionality. Jason and Sandeep see this as a step toward a super app, though they note limitations like needing prior discovery of GPTs and lack of multi-GPT chaining. Audio Book Creation via 11 Labs and ChatGPT (Priority: 3/5): Using 11 Labs text-to-speech GPT within ChatGPT, the duo generates a two-paragraph story and converts it to an audiobook in 'Jarvis' voice. The seamless process prompts discussion on AI-generated content and the potential for 'organic' labeling. Hugging Face Opens Ecosystem for GPTs (Priority: 4/5): Hugging Face launches an open-source 'assistants' ecosystem, allowing creators to build GPT-like tools using models like Llama or Mistral, and deploy them under any backend. This challenges OpenAI's closed model and fosters greater accessibility and model diversity. Anonymity, Doxxing, and Platform Rules on Social Media (Priority: 3/5): The hosts debate the ethics of exposing anonymous accounts on X/Twitter, discussing doxxing definitions, Elon Musk's stance, journalistic practices, and the use of AI to protect or unmask identities. They agree on the tension between anonymity and accountability for large accounts. Societal Impact of AI on Content and Identity (Priority: 3/5): Jason reflects on generational shifts from pre- to post-AI, drawing parallels to how internet and streaming changed media consumption. They speculate about 'AI-free' labeling for human-made content and compare analog experiences (e.g., live music, classic cars) to AI-generated content.
Key Arguments: Hume AI's model trained on only 80 videos is too limited for reliable conclusions, especially with extreme confidence scores like 99.39% (Jason). Forever AI fundamentally democratizes game creation, enabling individuals to produce and publish games without traditional coding or teams (Sandeep). ChatGPT's @-mentions for GPTs are useful but still limited by discovery and single-GPT constraints; the ideal would be chained multi-GPT interactions (Jason). The 11 Labs audiobook demo demonstrates powerful integration, but lacks seamless publishing to stores; licensing and residuals for voice actors remain issues (Jason). Hugging Face's open assistants ecosystem is a strong counter to OpenAI's closed model, improving openness and competition (Sandeep). Anonymous accounts with large followings cannot claim an absolute right to anonymity; platforms set their rules, and individuals choose the risks (Jason). AI can both unmask anonymous users (by stylometry or pattern analysis) and protect them (by rewriting content to remove personal tics) (Sandeep).
Data Points: Confidence Score: 99.39% - Hume AI's model classified Jason as 'self-confident' with 99.39% confidence based on a 30-second video clip. Training Data Size: 80 videos - The specific Hume AI model used (self-confident vs. self-doubting) was trained on only 80 videos. GPTs Available at Launch: 3 million - OpenAI reported 3 million GPTs were available at the launch of the GPT store. iPhone 16 Expected Launch: September 2025 - Analyst Dan Nystedt leaks that iPhone 16 will launch with A18 chip and generative AI. Content Filter Blocked Attempt: 1 - An attempt to generate 'Donald Trump' visual in Forever AI was blocked by content filters.
Pivotal Quotes: "It's 2024. There are no more excuses for an ugly website. Stop settling for okay and start using Squarespace and be awesome." — Jason Calacanis: During a sponsor segment, Jason emphasizes modern design standards and the value of a polished online presence. "Now that's going to fundamentally change it. So you think about Roblox, Epic Games, and everything that comes in and around it, I think we're going to see a massive explosion." — Sandeep Madra: While discussing Forever AI's potential to democratize game development, Sandeep predicts a transformative explosion in the gaming creator economy. "I think people are going to start saying this is organic, 100% human. Like, literally, the label, 100% human, no AI inside. Like, I know this sounds crazy and dystopian, but I think people are going to label content like this is organic." — Jason Calacanis: Reflecting on AI-generated audiobooks, Jason anticipates a market for 'AI-free' content, paralleling organic food labeling.
Implications: These demos show AI is rapidly transitioning from passive analytics to active creation and interaction. For entrepreneurs, the key differentiator will be training data, brand, and unique features, not just AI integration. The rise of open-source ecosystems like Hugging Face and creative platforms like Forever AI will further democratize content production. Ethical and regulatory discussions around anonymity, surveillance, and authentic 'human-made' content will intensify as these tools become mainstream.
About This Week in Startups
Jason Calacanis covers startups, tech, markets, media, and all the hottest topics in business and technology. He also interviews the world’s greatest founders, operators, investors, and innovators.