Bạn có chắc chắn muốn xóa bài viết này không ?
Artificial Intelligence. Our newest submissive servants?
People, especially the laymen, usually fantasize about Artificial Intelligence (AI). They wildly dream up the scenario that AI bots do our household, order the foods for our diner, switch the light on when we come home from exhausted work. Our house is AI-driven, too. A thief has no chance to break in because his iris, his fingerprints and his "odor" does not match one of our family members. In short: in the very near future, let's say 2030, we will live in Eden and have plenty time to enjoy our life.
- Cooking? AI maid does it.
- Babysitting? AI maid again.
- Working? AI bot does the job.
- Gardening? AI bot again?
- Driving? AI car drives us to a place where we want to be. Ever heard of Telsa, an "autonomous" driving car?
- Jogging? AI smartwatch tells us when we should stop.
- Loving? Well, as long as we don't fall in love with an AI doll. In China, for example, love-doll can be ordered on-line (click HERE to see for yourself). But is "she" AI? I don't know...
To be honest, do you really believe in such a scenario? Obviously people usually tend to be positive and think positive, too. Probably none of us thinks about the aftereffect, the negative repercussion that might gravely effect us. Such a paradisiac environment with all smart AI submissive servants requires someone who creates, builds, programs, maintains and pays for the AI servants. We? No they. Who's the they? The AI bots themselves. They reproduce and become smarter by their own evolution. Huh? Well, the next heretic question is: if they could "reproduce themselves and become smarter by their own way" then why they should serve us as our submissive slaves? They could become our "masters". A damn bad idea, huh?
You may say the Joe starts to polemize again against AI. Well, with my MS graduation in Computer Science and PhD in Nuclear Physics plus more than 20 years IT programming experience (as System and Application Software developer) I believe I could somehow adjudicate on this AI thematic. couldn't I? First of all, let me give you my "AI-understanding" and the AI futuristic scenario in reality.
What is AI? The word Artificial Intelligence is as vague as "FuzzyLogic". AI is the greatest word that everyone has his/her own imagination and no one could contradict him or her. My understanding bases on the facts that AI bases on smart Search Algorithms (SA) and efficient Knowledge Repository (KR). AI reflects Human Intelligence (HI). Also: AI depends and relies on HI. SA was and is the main driving force of human civilization and progress. SA is no other thing than a sequence of optimization to avoid redundancy and superfluity. The finer SA was, the better the outcome would be. Of course, there're some rules to make sure that some deviations could be reduced. But the rules are set by HI. Faulty HI rules will lead to faulty AI. Or to be more dramatic: dumb AI.
to the Neandertaler
up to us, human. They, and later we, were always on the search for a better life. The finding of fire improved life dramatically. Grilled meat tasted much better, easily to chew. Chewy meat became "edible" for old men.
- We searched for a good dwell: from a natural cave to a makeshift home.
- We searched for a peaceful place to live: from Africa to Europe and now all over the continents. Even in the most far-flung islands in the remote Pacific.
- We searched for the most efficient killing technique: from a wooden club over a slingshot to an H-bomb.
We searched and are still searching. Searching is the driving force of human civilization. Today, we try to apply our Searching knowledge onto machine and dub our searching endeavor as Artificial Intelligence.
- A tic-tac-toe is "intelligent" if its moves base on Alpha-Beta SA.
- Path Finding would be hopeless without A-Star SA.
- Baby-Giant-Step SA is applied in public key cryptography.
- Google SearchEngine relies heavily on MRA -MapReduce (Search) Algorithm.
As we realize, without our own searching experiences there's no AI. Our search for a solution leads to AI. The reason is simple and obvious. SA requires a tedious work that could take some hours or days or even weeks. And that makes us boring. We become impatient and often give up. A machine does not have such a sense of boringness or tediousness. It works, works and works till it finds the end result or we stop it. An example in reality. The Mastermind IBM Watson computer.
As IBM proudly announced its most-touted AI machine Watson its marketing machine drummed up Watson's AI for business noisily: Build Your Cognitive Business with IBM. Cognitive Business. What a big word. To demonstrate how good and intelligent Watson was IBM let its prodigy compete against human in a "Jeopardy show". Of course, it won!
The hype was perfect, an army of greedy entrepreneurs speculated on Big Money and fell into the marketing trap of IBM. Especially, Watson was touted as the best for Medical Diagnosis (e.g. Cancer). Health Care is the biggest business in all aging rich nations like the US, Japan and the EU. And IBM knows that. Its slogan is highly pregnant: IBM Watson Health - Cognitive Healthcare Solutions.
Big hospitals started to run after Watson. Slowly the hype settled down, rational sobriety comes up. But the damage was already done. The world-renowned Texas University MD Anderson Cancer Treatment & Cancer Research Center canceled in February 2017 a promising, but troubled contract with IBM for its Watson platform. No reason was publicly given. Inofficially it was the unpredictable cost soaring: planed 2.4 mio US Dollar, reality: 39 mio US Dollar.
However, as I mentioned above about the efficiency of the underlying Knowledge Repository which was only fed by HI. And I believe that it was the true reason for an abrupt "cancelation" of Watson. According to Forbes.com
Although the audit took no position on Watson’s scientific basis or functional capabilities, it did describe challenges with assimilating Watson into the hospital setting,” said Charlie Schmidt, writing for the Journal of the National Cancer Institute. “Experts familiar with Watson’s applications in oncology describe problems with the system’s ability to digest written case reports, doctors’ notes, and other text-heavy information generated in medical care.
The sentence "problems with the system’s ability to digest written case reports" underlines the unspoken significance of KR and it was confirmed as following
And weakness there are. Watson requires many months of laborious training, as experts must feed vast quantities of well-organized data into the platform for it to be able to draw any useful conclusions. And then it can only draw conclusions based upon the body of data, or ‘corpus’ (plural: ‘corpora’) that it has been trained on.
The ‘well-organized’ requirement is especially challenging for Watson, as unprepared data sets are typically insufficient. As a result, Watson customers must hire teams of expert consultants to prepare the data sets, a time-consuming and extraordinarily expensive process.
Watson also cannot make connections across different corpora, and thus is unable to glean even basic insights outside each corpus. For example, training Watson on oncology will give it no insights into heart disease – a failing that dramatically limits its use in clinical settings.
Again, as I said "AI depends and relies on HI." Anyone who heretically believes that AI can reproduce itself and becomes more and more intelligent by "deep-learning" either belies him or herself, or is simply an ignorant. HI is and always the master of AI. And Earnest Sohn confirmed that.
A team of Booz Allen Hamilton experts and an MD blogging for Health Affairs explained this challenge. “Human intelligence outperforms machine-learning applications in complex decision making routinely required during the course of care, because machines do not yet possess mature capabilities for perceiving, reasoning, or explaining,” explained Ernest Sohn, a chief data scientist in Booz Allen’s Data Solutions and Machine Intelligence group...“Moreover, despite significant progress, even state-of-the-art machine-learning algorithms often cannot deliver sufficient sensitivity, specificity, and precision (that is, positive predictive value) required for clinical decision making.”