Aira, Google DeepMind and Guided Vision: How Blind Testers Trained Gemini

Google's new Guided Vision feature in Gemini Live was trained with help from Aira and its blind Trusted Testers. Everette Bacon and Lynette Frison of Aira explain how the partnership with Google DeepMind works, what blind users taught the AI, and what it means for Aira's future. Steven Scott and Shaun Preece open with vibe coding, including homemade Claude voice assistants that answer in about three seconds and people using AI to bring old Echo Show and Meta Portal hardware back to life. Brandon Bracey's BlindRSS reader is now on Android, and Steven unboxes a new 8-inch Amazon tablet. Everette Bacon, Chief of Blindness Initiatives at Aira, explains how Explorers swapped free minutes for sharing their sessions to train DeepMind's Astra model. Lynette Frison describes running the Trusted Tester program, where a human visual interpreter watches every AI session and steps in when the answer is unsafe or wrong. Early versions hallucinated and once told a tester to find a sighted person. Today's Guided Vision is, in Everette's words, a little too safe. The panel covers how blind users prompt differently, from turning a vitamin bottle to describing remote buttons by touch. They talk about medical and personal requests, why some people feel more at ease asking AI, and why finding dog poop is Aira's number one request. Everette is open about the business side: Google pays for the data, and Aira hasn't decided whether to charge for Aira AI. Everyone agrees that blocking image descriptions for adult content or swear words treats blind people like children. Aira's Build AI program has now expanded to the UK, Canada, Australia and New Zealand. Chapters 0:00 – Intro and vibe coding the TV 2:40 – Building a Claude voice assistant 6:29 – Reviving old tech, BlindRSS for Android and a new tablet 11:52 – Aira, Google DeepMind and Guided Vision 17:32 – Trusted Testers, hallucinations and human in the loop 25:52 – How blind users' prompts train the AI 28:36 – Comfort, privacy and sensitive requests 38:13 – The future of Aira AI and Guided Vision 44:25 – Image description limits and content filters 51:09 – Build AI expands and wrap-up Relevant Links Aira: https://aira.io/ Aira Build AI FAQ: https://aira.io/build-ai-faq/ Guided Vision in Gemini Live: https://blog.google/innovation-and-ai/products/gemini-app/guided-vision-gemini-live/ BlindRSS: https://github.com/serrebidev/BlindRSS Claude: https://claude.com/ ---- Follow on: YouTube: https://www.doubletaponair.com/youtube X (formerly Twitter): https://www.doubletaponair.com/x Instagram: https://www.doubletaponair.com/instagram TikTok: https://www.doubletaponair.com/tiktok Threads: https://www.doubletaponair.com/threads Facebook: https://www.doubletaponair.com/facebook LinkedIn: https://www.doubletaponair.com/linkedin Subscribe to the Podcast: Apple: https://www.doubletaponair.com/apple Spotify: https://www.doubletaponair.com/spotify RSS: https://www.doubletaponair.com/podcast iHeadRadio: https://www.doubletaponair.com/iheart About Double Tap Hosted by the insightful duo, Steven Scott and Shaun Preece, Double Tap is a treasure trove of information for anyone who's blind or partially sighted and has a passion for tech. Steven and Shaun not only demystify tech, but they also regularly feature interviews and welcome guests from the community, fostering an interactive and engaging environment. Tune in every day of the week, and you'll discover how technology can seamlessly integrate into your life, enhancing daily tasks and experiences, even if your sight is limited. "Double Tap" is a registered trademark of Double Tap Productions Inc. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.