Talking to My AI (Part #2) — Is it useful yet?
🌈 Abstract
The article discusses the author's experience of using a voice-powered AI assistant to help understand the complex topic of the double-slit experiment in quantum physics. The author explores the potential benefits and limitations of using voice interaction for learning and thinking through complex subjects, compared to traditional text-based methods.
🙋 Q&A
[01] Exploring Voice-Powered AI Assistants
1. What were the author's key observations and insights from using a voice-powered AI assistant for several weeks?
- The author found that talking to the AI assistant felt like chatting with an "attentive but fundamentally unimaginative best friend" - it could express opinions on any subject but couldn't have a serious, in-depth conversation.
- The author believes voice-powered AI will be transformative, as talking engages the brain differently than reading/writing and allows for a more natural flow of thought.
- The author found that voice interaction enabled faster, more natural course correction and the ability to hone in on the right level of detail, compared to text-based interactions.
- However, the AI assistant still had limitations, such as overusing certain verbal crutches and struggling to provide the desired level of detail on complex topics.
2. How did the author use the voice-powered AI assistant to try to better understand the double-slit experiment in quantum physics?
- The author tried two approaches: 1) an unprompted interaction with the Gemini 1.5 model, and 2) prompting the model to act as a physics expert and providing it with articles on the double-slit experiment.
- The author found that the voice interaction allowed for faster, more iterative questioning to fill gaps in understanding, compared to reading articles or text-based chatbots.
- However, the model still struggled to provide the right level of detail, often reverting to high-level analogies or repeated phrases when pressed for more specifics.
[02] Potential and Limitations of Voice-Powered AI for Learning
1. What were the key advantages the author identified for using voice-powered AI for learning and thinking through complex topics?
- Voice engagement activates the brain differently, allowing for a more nimble, flowing thought process compared to text-based interactions.
- Voice enables faster, more natural course correction, allowing the user to hone in on the right level of detail.
- "Talk to your data" becomes a more powerful use case with voice, as it changes the dynamic and the types of questions that can be asked.
2. What were the key limitations the author identified with the current state of voice-powered AI for this use case?
- The AI assistant tended to rely on verbal crutches and repeated phrases, which broke the illusion of a natural conversation.
- The model struggled to provide the desired level of detail, often reverting to high-level analogies or repeating previous responses when probed for more specifics.
- The model's knowledge was still limited by its training data, and adding more curated information helped but did not fully solve the issue.
3. What does the author conclude about the future potential of voice-powered AI for learning and thinking?
- As voice-powered AI assistants become more ubiquitous, their usefulness as thinking and brainstorming partners will become more apparent.
- However, realizing this potential will require continued development of methodologies and benchmarks specifically aligned with the nuanced, multi-step back and forth of voice interaction.