Understanding Artificial Intelligence

Neil D. Lawrence

New Court, St Swithin’s Lane

Henry Ford’s Faster Horse

Embodiment Factors

bits/min billions 2,000
billion
calculations/s
~100 a billion
embodiment 20 minutes 5 billion years

Heider and Simmel (1944)

Human Communication

For sale: baby shoes, never worn

Evolved Relationship with Information

New Flow of Information

Evolved Relationship

Evolved Relationship

\[\text{reflect} \Longleftrightarrow \text{reflex}\]

The Great AI Fallacy

If AI isn’t a tool for us, then we are the tool of AI.

Accelerate Program: Empower the User

AutoAI: Resolve Intellectual Debt

Data Trusts: Empower People through their Data

AI@Cam

Conclusions

  • Humans view intelligence as embodied, but AI is not embodied.
  • Critical decisions are dependent on human nuance to reconcile the tension between \(p\)-Fairness and \(n\)-Fairness.
  • This implies that AI should only ever be seen as a tool of humans.
  • Much of current technology makes us a tool of the AI.

Conclusions Contd

  • Three interventions to address this.
    • Empower the humans to better use the tools.
    • Develop better standards for creating AI systems.
    • Develop data intermediaries to allow citizens to have a voice in how their data is used.

Thanks!

References

Heider, F., Simmel, M., 1944. An experimental study of apparent behavior. The American Journal of Psychology 57, 243–259. https://doi.org/10.2307/1416950
Lawrence, N.D., 2017. Living together: Mind and machine intelligence. arXiv.