There are no Nexus 6 for Blade Runners to snub out. Andrew Martin is nowhere to be seen. The software Hal 9000 has not emerged. Why have AI’s promises fallen so short? When will we see these dreams realized? As the questions mount, many are losing hope in the idea that man will ever create the AIs of our childhood sci-fi dreams. Yet as I write this little blurb a new round of funding is pouring into the AI Industry. The dream is alive once again. Technology has made its Mooreful leap and armed with new tools our brightest minds have convinced the money to fund their quest for this prized new land once again.
The following is a brief run down of 6 failures that have plagued our previous attempts at AGI / AI.
Failure 1: Scope
Projects fail to define exactly what they are aiming to create. There is a substantial difference between creating an AI that will emulate human behavior and one that is a new intelligence of its own kind. An AGI is a general intelligence that is supposed to be able to apply knowledge and learning from one area to another. A traditional narrow chess AI may be able to beat Gary Kasparov in a game on the checkered board, but can this same program beat him in a game of Stratego? It probably can’t even play the game of Stratego because Deep Blue can’t play games that it was not programmed to play. You can’t speak or communicate via language to Deep Blue or his buddy Jr., thus you can’t even begin to attempt to teach him a new game. These narrow-AIs were never taught to learn anything outside the parameters of their narrow programming. As complex and impressive as they are, they simply were programmed to play chess. That was what they were intended to do. Scope-wise, this really isn’t any more of an Artificial Intelligence than your Start->Run->Calculator is. While more elaborate narrow-AIs could definitely be applied to benefit humanity, these hardly are going to get us to the holy grail of AGI.
Failure 2: Man Tries To Create AI In Their Own Personal Image.
Inherent to most approaches towards creating AIs is our limited awareness of self. Before one can duplicate or even emulate their ‘self’, they must be aware of what and who they are. How many people do you know would you consider to be enlightened individuals? How many of them would you consider to have a very strong, brutal, and honest awareness of exactly who and what they were on various scientific, psychological, physiological, and philosophical levels? Now how many of these people are in the field of AI leading a project to create an AGI? Most techs are…… um…. techs. Yes, we have brilliant techs, but where are our cross-discipline geniuses that can function and thrive in multiple domains? When attempting to create an AI/AGI that can emulate a human one can only operate from the level of their own awareness. You can’t duplicate or try to create something which is outside your scope of awareness. We create in our own image. We can stumble upon new creations by happenstance, as history has demonstrated with many of our great inventions; but AGI will not be solved by accident. Humanity is not a million monkeys with an infinite amount of time…
Failure 3: Targeting An End Product Leads To Closed-mindedness And A Narrow Approach To The Real Challenge.
Many projects have an end product in mind. They want a program that does xyz, so they design their program to do xyz. As soon as any additional need arises, or success begets new hopes in a deeper narrow-AI, the lacking design and ultimate failure becomes painfully clear.
Failure 4: Corporate Restrictions Encourage Short-cuts To Meet Deadlines.
Any project forced to make demonstrations or meet corporate deadlines will not create AGI. Compromises and short-cuts are the AGI devil.
…Premature optimization is the root of all evil – Donald Knuth
Failure 5: The BLACK SWAN Awaits.
Assuming Moore’s Law to be true, works, only as long as Moore’s Law is true. Since the time of OIL our technology has obeyed this generous ‘law’. It requires an ample supply of energy and if/when energy is no longer as prevalent or available for us humans to use the Law will cease to be a Law. If/When it breaks so will all the predictions and hopes that were predicated upon its perpetual truthfulness.
Failure 6: The Ship Sails Where The Captain Directs.
The minefields of Pride, Arrogance, Stubbornness, Old-Mindsets, and Ego await. To successfully create an AGI the project will have to successfully avoid these human disasters.
2011……….. The Result? Over-hyped promises that have yet to be realized.