The Case Against Conversational Interfaces

01 Intro

• Conversational interfaces are often viewed as the next computing paradigm.
• This trend resurfaces every few years with new AI developments (e.g., Siri, Alexa, chatbots, LLMs).
• Despite excitement, little changes in how we interact with technology.
• The obsession may stem from nostalgia for futuristic concepts (like Star Trek).

02 Data Transfer Mechanisms

• Natural Language: Refers to written or verbal communication; a data transfer mechanism between humans.

• Critical Factors:

  • Speed: How quickly data is transferred.
  • Lossiness: How accurately data is transferred.

• Observation:

  • We receive data faster (reading/listening) than we can send it (writing/speaking).
  • Thoughts form at 1,000-3,000 words per minute; natural language creates a bottleneck.

• Alternative Communication:

  • We often use gestures or non-verbal cues (e.g., thumbs up, nods) for faster communication.
  • Natural language is effective for high fidelity data transfer but less convenient for quick interactions.

03 Human Computer Interaction

• Historical Context:
  • Early computing used command lines requiring precise syntax.
  • GUIs made computing more accessible by using visual elements.
• Current Interaction:
  • Users combine graphical interfaces with keyboard shortcuts for efficiency.
  • Shortcuts help communicate faster, acting as data compression techniques.
• Touch Interfaces:
  • Augment desktop computing but do not replace it; typing on mobile is slower than desktop.
  • Emojis and selfies are mobile-specific shortcuts, but we lack mobile equivalents for keyboard shortcuts.

04 Conversational UI as Augmentation

• Conversational interfaces (e.g., LLMs) can enhance workflows but are not swift enough to replace existing paradigms.
• Personal Experience: Used LLMs effectively for drafting essays, highlighting the potential of conversational AI as a collaborative partner.
• Thesis:
  • Conversational interfaces should complement existing tools rather than replace them.
  • Potential for voice commands to be an additional input mechanism, increasing data transfer bandwidth.
• Future Vision:
  • AI should function as a command layer across tools, enhancing productivity without interrupting workflow.
  • Explore voice input compression techniques for faster communication (e.g., simple sounds).

05 Closing Thoughts

• The essay argues against zero-sum thinking regarding AI and interfaces.
• Emphasizes AI as a complement rather than a substitute, enhancing interaction and productivity.
• The future lies in improving human-computer interaction to feel seamless and effortless.