NI = Natural/ Native Intelligence… the basic human intelligence
AI = Artificial Intelligence
AI is generating enormous focus and interest. A ton of new ideas from image recognition, art style transfers, poem generators to self-driven cars. I am personally building a lot of these apps and am awestruck by the possibilities. AI Deep learning rocks!!! The focus is so pervasive that people are questioning what will happen to the future of jobs. What jobs will get impacted? What will no longer remain? The common thread in all this is that AI can dramatically change the way jobs are done… hence the risk.
Shouldn't we be asking a different question?
Let’s step back and look at our NI.
Works differently: Think about this… How many hours did you drive your car to get a license? Using NI of-course. Google’s Waymo self-driven car has driven 3 million miles across 8 years and still is not pervasive on roads. Using AI in this case.
Human intelligence has its limits: It has taken us 100 years to measure gravitational waves that Newton postulated and we still don’t know whats out there… dark matter, dark energy etc. 96% of our “dark” universe is unknown and in the 4% of matter, we have explored only a few planets. This is one example that indicates the existence of an NI learning rate and limits.
This begs a question. Can AI be refocused differently? In a way that can advance humanity in the true discovery sense and not with efficiency as the focus. AI should advance NI and solve issues that NI is limited to solve
Here are a few examples:
Problems that have remained unsolved despite centuries of NI: Ex — Malaria eradication
Augment NI where it is fundamentally limited: Ex — Humans can only see visible spectrum of light while the universe has a lot more to offer… AI as a mechanism to augment what we see, hear etc including enabling the disabled
True unknowns that have tested NI limits: Ex — Discovering dark matter and black holes, time to put humans on Mars… break the inherent lead times that seem to exist in these
To do this one wonders if the current loop of “get data and learn” in a supervised way is most efficient or the focus should shift the way NI does… unsupervised… connect the dots