"Essentially all models are wrong, but some are useful"
Last week, during the ESI conference on complex systems at TU/e, the presentation of Professor Edward A. Lee (University of California Berkeley) on "Intelligence & Computation" helped me pose a couple of questions to myself.
What exactly are we trying to re-build with Artificial Intelligence?
Since -as the Professor highlighted- the most important parts of human intelligence are either not computational or cannot be replicated.
- Consciousness and sub-consciousness
- Language imperfectly encodes our thoughts
- Most truths in our head never get expressed
- Communicated ideas are never perfect
- Timing (sense of time & definition of time)
- Chemistry
- Embodiment
- Non determinism
- Chaos (endless feedback loops)
Thus, few realizations/questions came along:
- Is AI just grasping the intelligence of our current time?
?Since human DNA is not only evolutionary, but also some parts of it contain non-transmittable information.
- If intelligence is the next challenge for system complexity, shouldn’t we be really mindful on how we define it, and therefore use it within any given complex system?
Thinking back, I realized that he had already provided an -initial- answer to my questions through a scientific/philosophical quote.
“Essentially all models are wrong, but some are useful”
George E. P. Box - Norman Richard Draper
Visual Artist
5 年technical art
Maak leven, werken & organisaties beter & leuker! 5xA: ??Ambition ??Aim ?Achieve ??Appreciate ??Amplify!
5 年Interesting thought. The everlasting perfection of imperfection, which brings us the dynamics of evolution of nature, shows there is no unbiassed truth. With AI and with all models we use to transfer ideas, thoughts, insights, et cetera, we need to consider its limitations. All models are reductions of complexity, used by us to help us understand. That’s why they are usefull, yet imperfect.