A social phenomenon

Human, AI, Crowd, Intelligence. That’s the four words emerged in my mind observing what happened in a chat channel at where I work:

  1. The channel started as an AI summary of competitor intelligence from a fixed list of sources.
  2. People started fact check the AI summaries and pointed it out obvious omissions and even hallucinations.
  3. The channel owner improved her prompt based on the feedback. The AI summaries become more and more accurate and on-the-point.
  4. The channel membership enjoys exponential growth, including big wigs from multiple BUs.
  5. People are more and more engaged:
    • They are doing more fact checks by reading the original content
    • They are suggesting new contents not in the list
    • They are offering their original human insights
    • All probably to show off to the folks and big-wigs in the channel : )
  6. The channel enjoys more and more popularity and membership has to be limited

What can we learn from above phenomenon?

AI is powerful but not yet trustworthy. Human is still indispensable in the loop. Problems arise:

  • The speed of AI far exceeds that of human
  • Human needs a crowd to keep up with AI
  • How do AI and humans work together?

New paradigm of engineering system needs to be built for human-AI collaboration

Human AI Crowd Intelligence (HACI)

As a vehicle for Human AI Collaboration (HAI), I propose HACI as an engineering system for Crow Intelligence (CI):

  • Crowd
    • Humans crowds, the bigger the better.
      • You never know who in what corner wearing what hat knows or sees something valuable
    • AI Agents crowds
  • Intelligence
    • Competitive Intelligence
    • Market Risk
    • Voice of the Customer (VOC)
    • …

Which support below intelligence curation process:

  1. Signal Curation
    • Manual submission, by humans
    • Automatic web search, by AI
    • Manual and automatic topic classification
  2. AI Summary
  3. Human Feedback
    • Comments on the original source
    • Human provided insights and intelligence
    • Validation of the AI summary
    • Experimentation and refinement of the AI prompts
  4. Intelligence Fusion (upward)
    • Enrich signals with contextual knowledge
    • Correlate disparate signals
    • Distill signals into verified insight and actionable intelligence
    • Present to upper management for their consumption
  5. Knowledge Sediment (downward)
    • Spatially
      • Connecting related entities together
      • Connecting entities and their facts together
    • Temporally
      • Connecting history, present and future (predictions)
  6. Loop
    • Back to Step 1, rinse and repeat the virtuous loop

Harness the power of the crowd

  • To scale beyond the capacity of dedicated person or teams
  • To verify AI produced content
  • To capture the signal and insights at the long tail or in the unknown unkown
    • Which might be seen or possessed by the unlikely employee

Engineering Driven

  • Social Engineering
    • Incentivize employees to participate
  • AI Engineering
    • To take the most advantage of AI technology
  • UX Engineering
    • Insanely smooth user experiences with which humans are motivated to engage

Brand

Since the proposed CI system acts as the external brain and neural system for an enterprise, I propose to name it:

  • ExoBrain
    • exoskeleton → exobrain
HACI