On An Alien Planet
A good way to determine what the first thinking machine will look like is to think about what it will need to do to adapt and survive on an alien planet.
An ‘alien planet’ is any new environment - the deep sea Mariana Trench was alien to him, yet James Cameron, an evolved primate with land-based mammalian training, could engineer his way down to it.
At the extremes, one may think of a yet undiscovered planet that exists in no internet corpus of today. Something right out of Netflix’s Alien Worlds.
This is a data question, an energy question, and a hardware assembly question, rolled into one.
How does it learn from very little data? How does it operate with very little energy? How does it reorganize/repair from very few building blocks?
Miss answers to any one of these questions, and the system dies by default. The human brain is able to deal all of them, hence one instantiation of it - James Cameron - can make it to the Mariana Trench.
Today’s machine learning systems answer none of these questions. Such is the state of modern AI.
Lest it may seem daunting, it is no extraordinary ask. Mere 6 year olds adapt to such alien worlds every day, every week. Playing and mastering new and alien video games with commonplace ease that would stump any machine.
Indeed each new frontier is like a deep sea trench.
And for each new explorer, man, but moreso machine, each unanswered question is a fatal festering wound.