Today we reflect about the concept of “tool”
As I said in past blogs a pilot is able to fly a plane, “limiting” his knowledge of it up to a certain level.
He does not have to know the behavior of the wing profiles in fluidynamics, the wind tunnel, such as the electrical circuits of its navigation radar. Programmers made software of the autopilot , and so on……
We call it “TOOL”. And pilot uses it.
In this case, there are a lot of other people who have been involved in design and construction and who know this set of arguments, and (eventually) can be asked about what they have built, in spite of how difficult it is.
This is called human progress. “Nani gigantum humeris insidentes”,
What would happen if an airplane pilot could not access this chain of knowledge, albeit partially? …..Perhaps made by a deep neural networks.
This is an extremely new situation especially when much of this chain of knowledge evolves mostly autonomously, without the possibility of human control.
Deep learning is more than an advance pattern recognizer of today. Or it might be.
THIS IS NOT A TOOL but an AUTONOMOUS INTELLIGENT ENTITYwith which we can potentially collaborate. Or compete.
And compete in a human specificity: intelligence.
Our Lego dear alien.
Very often this is much more effective than the human counterpart, and can pass very easily Turing test.
Faster than you. Probably smarter(?) than you.
But what if it behaves in an unpredictable way? Are we ready to accept and control many unpredictable behaviors that interact with our lives?
What if it creates damage? Who knows their internal structure and try to correct the malfunction? NOBODY.
We must be extremely aware of this global evolution scenarios like Internet . Today, any virus or cyber-attack can paralyze from the fridge to the banks. Today we can already lose control. And tomorrow?
Of course I do not want recycle Orwellian Big Brother , or seem like a paranoid preacher.
Now we are engage a bigger adventure and we all be aware of this. A game between our immortality and self distruction.
As I have previously written, the advent of search engines has created a fantastic specific knowledge and a big general “ignorance” of whole argument.
Do we know what we are really doing and be aware of all the implications?
BTW at present many neural networks can only be run by large equipment not available to any academic researcher.
Better IMO improve our human intelligence and “reenginering” web as global communication and learning tool, before dreaming about AI.
Life on earth is already a complicated plot of billions of hardly interacting brains with magmatic results. We do not even need artificial superhumans ones!
