Transcendence and Analytics: What If the End of Mankind as We Know It Were a Good Thing?
Apr 21, 2014 6:01 AM PT
I saw the movie Transcendence last week, and it was an impressive telling of what might happen were we to achieve singularity. Given this is likely where AI research is taking us and that analytics is a major component to getting there, I not only recommend the movie as entertainment but also as a means to wrapping your head around what we must correct before we take that next big step. The problem resides in us.
Much of the world operates on lies. Governments, religions, employers, coworkers and relatives all lie at times, with impacts that range from trivial to catastrophic. In effect, there are two kinds of people: those that know they are lied to and those who haven't really figured it out yet.
Think of all the things your parents told you that weren't true. My stepmother told me pregnancy resulted from kissing, which kind of confused me about sex for a couple, fortunately early, years. When life begins, the cause of homosexuality, global warming, infidelity, politics, and whether someone will respect you in the morning all form the basis of massive arguments in which both the side that is wrong and the observer often have no idea who is right.
Heck, even the side that is right likely has doubts. I often wonder why we believe anyone. Since valid analytics could be a massive weapon for either the right or the wrong side, I wonder whether this fact- based tool actually will change hard-set opinions.
I'll close with my product of the week: something that could save you from being embarrassed by spreading a lie you thought was true.
Analytics is a class of products that analyzes large data sets -- structured, unstructured or both. The last is still a bit rare. Analytics is a fact-based tool, designed to assist in decision making, that is both underutilized and misused today. I say "underutilized and misused," because few companies, including those that sell these tools, use them broadly, properly or well.
A perfect example of this is how the Republican Party lost the last election. It used analytics heavily and concluded Republican candidates would win by a landslide. So convinced of the win were Republicans that they entered the final days of the race overconfident and had their collective butts kicked.
Granted, they made a significant number of errors, but they clearly believed they would win -- and by a landslide -- until they didn't. This showcases the danger of bad analytics. It can result in more effectively promoting a lie that could, and did in this instance, ensure failure.
To conduct any analysis correctly, automated or not, you have to ensure three things: that the data set is unbiased; that the analysis tool works; and that the person interpreting the results can do so with accuracy. That means the analyst has to both have the skills and the willingness to accept results that contradict any pre-existing conclusion or wishful thinking.
If the sample is biased or the recipient of the data is unable to do the job properly, it renders even the most expensive analytics process less valuable than a coin flip. The coin flip has a 50 percent probability of being right, while a corrupted analysis process will virtually always result in an inaccurate answer. An inaccurate answer you believe to be accurate, as the Republicans painfully found out, is worse than no answer at all.
Accepting the Right Answer
Now this is the tricky part, and it came to me while reading a Salon article about a guy from the religious right arguing that being gay was a choice and the audience getting up and walking out during his presentation. The speaker was pissed, and audience members argued they weren't against free speech but they didn't have to listen to a guy with whom they didn't agree.
This is a highly polarized issue. Let's say we had a report done using valid research (unbiased) that proved either side wrong based on a significant weight of valid medical facts backed up by an analysis of the Bible, also fact-based. Now stop for a moment. I'm saying whichever side you were on, this evidence -- which was unimpeachable, using valid facts -- would prove you were wrong. Would you accept it? Based on past experience, I doubt most would. We would assume there was some bias or a trick and reject the conclusion.
Let's take life vs. choice on the abortion debate. If analytics could prove that the Bible didn't allow that a fetus at any age was human and instead actually supported abortion or if analytics could prove that women incurred massive psychological damage when having an abortion or that the fetus could experience and understand an unacceptable level of pain during the process. If analytics could then weigh those facts to conclude that one position or the other was invalid, would the losing side accept the results? I doubt it.
It is even more deadly in politics. For instance, analytics would have shown there were no WMDs in Iraq, but you may recall the administration in place at the time aggressively suppressed that information and manufactured false results because it wanted war. How many people died because of a lie?
Wired to Be Fools
There are two theories that really mess up our decision making. One is Confirmation Bias, which makes us see only things that agree with a position we've already taken, and the other is Argumentative Theory, which maintains that status goes with being seen as right.
Note I said "being seen as right," not actually being right, which likely explains the unfortunate number of politicians in every party who seem to be idiots.
Analytics Could Save the World
This is one area in which analytics absolutely could help us. If we were to rank politicians based on their capabilities, we could fix government -- and given the satisfaction rating for the U.S. government, I'm thinking it wouldn't be that hard to convince most people that their elected representative was an idiot.
The difficulty would be setting up a service that wasn't biased by some special interest group or corrupted by one or both parties -- but wouldn't it be wonderful if we could?
We often pick based on party affiliation and how appealing a politician is. Wouldn't it be far better to pick candidates based on competence and whether they actually agreed with positions we felt strongly about? Analytics could give us those answers, but would we take them?
The movie Transcendence makes a compelling argument that we wouldn't, and it makes some rather subtle and painful points about whether we are our own worst enemies.
This is as much about how not to do stupid stuff ourselves as it is about analytics. We all, and I definitely include myself, get into arguments with people that are deep and heated without first ensuring our position is supported by actual facts rather than things we want to be true. Even companies do this.
For instance, Google is in a massive push to prove mobile devices are better than PCs, and it is winning -- but PCs make more revenue for Google than mobile devices, so it is on the wrong side of the argument.
Microsoft is fighting Google with search, a free product, and Bing has become a huge money hole. You'd think the company would come up with a service that was both better and more secure, but charged a nominal fee instead. Pulling a line from the movie A Few Good Men, the problem with analytics -- and eventually AI -- is that for critical things, we really can't handle the truth.
I'll leave you with the one big thought from the movie Transcendence: What if "the end of mankind as we know it" were a good thing?
Product of the Week: Snopes
Of all the websites I use, I find Snopes to be the most valuable. That is because, when I remember to use it, it showcases accurately whether something I don't want to believe is a lie -- and, more importantly, it showcases whether something I may want to believe is a lie.
We naturally want to believe that scandalous stories about the government -- or the people in it with whom we disagree -- are true, and we are far more likely to test arguments we don't agree with than stories we wish were true, because we want to be right. Once our mistake is uncovered, though, we lose credibility -- and if we make this mistake, we can become a bad joke.
If we are willing to consider that what we were told might be false -- and often we aren't -- the Web is an excellent place to go to find out, and because Snopes is an invaluable resource on the Web, it is my product of the week.
By the way, as it is also a good place to read about the top scams going around, it could keep you from accidentally sending your life savings to the new version of that poor Nigerian prince. Apparently that was still the top scam last week. I know several people who have fallen for one of those scams...