
Summary
Written by a model from the transcript; not yet reviewed.
Alex Cranz and Nilay Patel review the current state of generative AI smart home voice assistants, focusing on Amazon's Alexa Plus and Google's Gemini for Home. They test both assistants on tasks like smart home control, calendar management, and complex multi-step queries, finding that while the new AI models allow for more conversational interactions and command chaining, they are notably slower, less deterministic, and frequently unreliable compared to older assistant generations. The discussion highlights the architectural bottlenecks of piecing LLMs over older backends, the frustration of inserted advertisements in the Amazon ecosystem, and the overwhelming surveillance implications of deeply integrated smart home camera networks.
- Gemini for Home appears to rely on the older Google Assistant backend, making smart home commands noticeably slower and less predictable. 5:31
- Alexa Plus can handle multi-step natural language requests but frequently injects unwanted upsells and product advertisements into the user experience. 14:23
- New LLM-powered voice assistants suffer from severe performance latency when processing compound smart home commands. 19:39
- Smart home camera integrations can track pets and human activity, raising significant privacy concerns about surveillance data collection. 36:14
Chapters
- 0:02Intro and Vergecast Overview
Hosts introduce the episode, guests, and preview the main topic of reviewing Gemini for Home and Alexa Plus.
- 3:28User Experience with Google Assistant and Gemini
Neelai shares his experience using Google Assistant and Gemini for Home, highlighting issues with smart‑home control and latency.
- 13:10Alexa Plus Conversational Flow and Sales Tactics
The hosts explore how Alexa Plus handles conversational requests, its tendency to sell products, and the impact on user trust.
- 19:16Smart‑Home Command Chaining and Speed Issues
Discussion of how both assistants handle multi‑step commands, the speed differences, and the frustration caused by slow responses.
- 21:16Calendar Management and Integration Problems
Comparison of calendar handling between Alexa and Google, noting gaps in work‑calendar integration and user frustration.
- 30:16Conclusion and Next Steps
Wrap‑up of the episode, acknowledging current limitations and expressing hope for future improvements from Amazon, Google, Apple, or OpenAI.
- 32:50Future of Smart‑Home AI and Surveillance Concerns
Hosts debate the potential of AI‑powered ambient experiences, the role of cameras, and privacy implications of data collection.
Highlights
- 5:31
Gemini backend architecture theoryinsight
Nilay explains his conspiracy theory that Gemini for Home is simply issuing commands through standard Google Assistant in the background because of how slow and clunky the smart home responses are.
- 14:23
Parental advice AI comparisonstory
Alex tests asking both Gemini and Alexa for parenting advice after dropping a son off at college, showing how Alexa immediately incorporates a product plug into its response.
- 19:39
Smart home speed degradationdata
Alex and Nilay discuss how moving to new LLM-powered voice assistants caused speed to take a massive hit, making simple home automation tasks frustratingly slow.
- 22:36
Google live chat breakdownstory
Nilay shares a disastrous test of the Gemini Live chat function where the assistant misidentified facts about a podcast segment and then became trapped in a loop talking to itself.
- 27:42
Apple Siri regression on iOSinsight
Nilay points out that the new Siri on iOS betas has degraded in smart home contexts, responding with conversational remarks instead of executing saved home automation scenes.
- 34:02
Milk run surveillance nightmarestory
Nilay recounts a real-life scenario where running out for milk late at night triggered a swirl of automated alerts to his wife in Denver, highlighting the creepy nature of smart surveillance.
Quotes
“The promise of all this AI is that, I mean, just think about what you just said.”
Unnamed speaker · 6:54 “Amazon just has a totally different revenue model.”
Unnamed speaker · 14:59 “Alexa does a much better job with that, which is surprising because Google is the calendar.”
Unnamed speaker · 20:37 “We can turn the lights on and off with our voice. That was a great value proposition.”
Unnamed speaker · 37:07
Sponsored segments
- American Express 2:18–2:30
- Midi 2:46–3:06
- Michelob Ultra 16:50–17:18
- Oracle 40:45–41:07
About this episode
For better or worse, Amazon Alexa and Google Home's smart voice assistants have been imbued with the powers of generative AI. The promise is that you can use natural language with your smart speaker to adjust your thermostat, lock the doors, and find out what your pets have been up to. The reality is smart home voice assistants seem to be going through an identity crisis. Senior reviewer Jennifer Pattison Tuohy invites Verge editor-in-chief Nilay Patel onto his own podcast to compare notes on life with Alexa Plus vs. Gemini for Home. Further reading: GTA VI keeps leaking ahead of its gameplay premiere I bought DJI’s banned camera — it was cheap and easy Does giving a camera wings dodge the FCC’s drone ban? Meta glasses are a workplace menace Google’s Pet Memory forgot who my cats are I tasked Alexa Plus with tackling my to-do list — it was hit or miss With Alexa Plus, Amazon finally reinvents its best product How AI broke the smart home in 2025 I let Gemini watch my family for the weekend — it got weird Google built a great smart speaker, but Gemini isn’t ready for it Jony Ive’s first OpenAI gadget is reportedly a hockey puck-sized smart speaker Subscribe to The Verge for unlimited access to theverge.com, subscriber-exclusive newsletters, and our ad-free podcast feed. We love hearing from you! Email your questions and thoughts to [email protected] or call us at 866-VERGE11. Learn more about your ad choices. Visit podcastchoices.com/adchoices
Transcript
7,349 words · 9 speakers
Topics
- Google Assistant
- Apple
- Siri
- Gemini for Home
- OpenAI
- Amazon
- Google Home
- Alexa Plus
- HomeKit
Page updated · transcript and summary are machine-generated