It has long been a meme in the conspiracy world that there are, amongst large corporations and government, super computers (even quantum computers) which are on the verge of emulating human thought. Everyday we see articles about government agencies who have vast computer complexes with software that can do sophisticated pattern matching, facial recognition, voice analysis amongst other amazing traits. But at its most fundamental these computers run software, the same way you run WORD or Minecraft. They wait for input then execute complex programmed processes to produce and output. Never the less rumors have persisted that AI’s or Artificial Intelligence systems do exist, but never in the mainstream.
It is that dichotomy which made the recent news that a new computer system, called Eugene, passed the Turing test all the more fascinating. This stunning “victory” was a mainstream news item and stands to prepare the world as we begin our decent into an AI world.
So what is the Turing Test and why is it important?
Named after computer pioneer Alan Turing, the Turing test is a base test of a machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. In the original example, a human judge engages in conversations with a human and a machine designed to generate output indistinguishable from that of a human. Participants are separated from one another and if the judge cannot reliably tell the machine from the human, the machine is said to have passed the test. In its current form the computer must fool 1 of 3 judges assigned to the test into thinking its human, a threshold which has now been passed.
Within the last several days a new Russian computer called “Eugene” fooled one of the three judges into believing it was a non-native English speaking 13 year old boy. A stunning victory for the Russian team which put together “Eugene,” and worthy of technical praise of the highest form, but there are implications.
What this outcome has shown us is that we are, especially now, on the verge of computers which can emulate humans. They may not be able to duplicate complex thoughts and emotions of its human creators but it can take input, process input and output that processed input in conversational form. In English that means that you can speak to it, it can process what you said, and respond to you in a way which is distinguishable from human. That has massive implications, because as the level of its sophistication increases these systems will start to replace humans in everything from Customer Service to positions of more complex decision making because if you can make a decision tree the system can interpret that decision tree and make a decision. The only problem is that its decision will be cold, unfeeling and purely calculated. To illustrate the point you might just think of HAL from the film 2001.
In the film HAL was a computer, or more accurately a program, which would take input (from a machine ie the USS Discovery or a human ie David Bowman) then pass it through a decision tree and arrive at an outcome. Now in HAL’s case that outcome was to purge the crew of the USS Discovery but you could place the same logic into investments, healthcare or any other decision oriented process you are involved with.
This might be a better example – You feel sick and you call the hospital. The person who answers the phone is not actually a person but a machine. You have a conversation where you talk about your head hurting and the “person” inquires as to you symptoms. Based on those symptoms the “person” suggests you come into the Hospital to run some tests and schedules an appointment. The results are fed back into the system and probably an underpaid doctor glances as the results and says yes there is a cancerous mass in your head. The machine then takes all the available information about you, your history, and your interactions, then runs a algorithm (a math program) to see how you would fair, statistically, in a operation to remove the mass. The machine, unlimited by emotion, decides you have a 72% chance of complications, but it has been told anything above 70% is non-operable. So the following day you get a call from the machine telling you, or maybe its a doctor telling you, you have inoperable brain cancer and you have a 60% of being dead in 2 months. NOw of course you have a lot to live for so you urge it to do the operation anyways, but the machine has no emotions so it doesn’t really care – you are too risky.
To extrapolate that we can explore the Wargames scenario, or the Matrix, even SKYNET from the Terminator franchise. All machines with complex decision trees which ultimately make catastrophic but rational decisions. The technology for that exists today, right now, but what was lacking was the interaction. The machine would not emulate a human so a human was required for the input. A broken flawed, and beautiful human who could adjust the diagnoses and override the decision tree based on simple emotion. The passing of the Turing test means that machines are evolving, something we knew was coming, and are now reaching a point where we will be unknowingly interacting with them at every level. Maybe not today or tomorrow, it will take a few years I suspect but we will get there sooner rather then later. Why? Because they don’t get paid, they never sleep or eat – they always work and that’s efficient. We like efficient in our society.
The bigger issue comes when we start to hand over more and more power to the computer. As we cede more and more control to the omnissiah, computer god, we are placing the control of these processes into a uncaring hyper-rational system.
Now don’t get me wrong, I’m not a Luddite. Technology is my thing, and I use it constantly – maybe even too much and I think there is a place for AI in the utopia of tomorrow. Putting gas in my car, fixing my car, ordering groceries I like maybe even paying my bills but its also a slippery slope. As we move in that direction the flaw of humanity, laziness, will start to push more control into the machine. Driving, air traffic control, flying (which is done today btw but more in a minute), even healthcare maybe all going that way.
As I said it’s not a problem today for the most part, but as we begin our journey into our rosy distopian future this is something we need to think about and control. Now that we have reached the first threshold, the Turing test, we need a standard set of limitations, like cloning, which everyone signs up for. Otherwise this will begin our decent and our technological arrogance with be the seeds of our destruction.
Now I know there are many people who are reading this saying oh whatever – you’re paranoid. That very well may be, but think about this. The next time you fly on a A320 or a Boeing 777/787 do you know who is really flying your plane? It’s a well known fact that those three aircraft in particular, and if the airport has the right type of equipment, can take off, fly and land your plane without a pilot. In fact if you fly, the aircraft you were on has likely done one or all of those things while the pilots discussed their dental plans or had a chat about world events. The aircraft is a tool, and the pilot is there in case something goes wrong but what I’m saying is that moving into the next phase we need to make sure the human operator is still in the equation.