As Watson utterly destroyed Jeopardy’s two reigning human champions at their own game, the tech world applauded IBM’s creation, but the applause was tempered by an almost palpable sense of anxiety. It’s that looking-over-the-precipice feeling you get when you see something that has the possibility of changing the world, but you’re not sure if that change will be beneficial. As Ken Jennings and Brad Rutter tried helplessly to at least make their combined winnings match that of Watson’s (they failed to do even this), one word reverberated throughout the Webosphere like a hushed premonition of the ides of March: Skynet.
Skynet is the digital villain in the popular storyline of Terminator that was built by the government to protect its citizens from foreign attack, and instead becomes self-aware and turns on humankind as it attempts to deactivate the rogue software. The AI decided that all humankind was considered a threat, and used the US’s nuclear arsenal to unleash a nuclear war on the world, annihilating the vast majority of the population. When faced with such a creepily powerful artificial intelligence in a game of Jeopardy, where it seems to be solving natural language problems faster than a human can, the first thought of the tech world was understandably, “Uh-Oh.”
However, as twistedly romantic as it is to imagine a scenario like that actually occurring as a result of IBM’s research, the truth is that all the pessimists are faced with the slippery slope logical fallacy. The idea of a “Skynet” scenario is extremely far removed from the reality of what IBM has actually accomplished, and it is unfair to overshadow the amazing possibilities that Watson brings to the table with a slippery slope argument that has numerous flaws. Here are 5 reasons why Watson is not Skynet:
Watson is a search tool – Granted, it’s a search tool on steroids, but all Watson is at its core is a group of algorithms that gets applied to a massive database of searchable data sources, albeit in creative and innovative ways. It uses massively parallel computing to apply proprietary IBM natural language parsing and searching algorithms to a vast body of what IBM calls “as-is” text. “As-is” includes current dictionaries, encyclopedias, and other publications to build the source of data that Watson can search. It is entirely self-contained in that Watson cannot glean information from anything other than the specifically structured data that the IBM has fed it. Being afraid of this aspect of the Watson project is akin to being afraid of Google’s search algorithms. In the end, it is simply searching a discrete set of data for an answer, and determining a confidence level that it’s the correct one.
Self-Unaware – The key component to a doomsday scenario spelled out in the Terminator world is the idea of self-awareness. Self-awareness is the idea that something knows that it exists. In science fiction, the consequences of an inanimate object becoming self-aware is that sees itself as an individual as much as a regular human would, going so far as to defend itself when attacked. Another key component is that it is able to formulate knowledge regardless of what its written code tells it to formulate. In other words, it thinks for itself. Any scenario that involves a situation of self-awareness would not include basic search and parse algorithms and a static data source of just text. It’s not even close to the same thing.
Only plays Jeopardy – At least in its current incarnation, Watson was designed to do one thing: play Jeopardy. Many of its algorithms and risk analysis processes are based on the rules and flow of the game of Jeopardy. It uses categories to build confidences, it wagers based its own standing and the standings of the two opponents, and it even knows to answer questions in question form. It is a single-purpose machine, and only the underlying technology can be used to adapt it to other environments. The implementation is only useful for game shows; it can’t annihilate civilization in a global nuclear winter just yet.
Data, Information, Knowledge, Understanding, Wisdom – Systems Thinking posits that there are different kinds of data, and this applies to both humans and computers. Data is raw symbols. It means nothing in and of itself, and is just there to be read. Information is what happens when relational connections are made between data points to create something that may or may not give meaning to data. Knowledge is the application of information to suit some purpose. Memorization of a multiplication table is knowledge. It is a useful implementation of information, but it is missing a key function. It lacks understanding. Understanding is when you can use knowledge to create more knowledge. In the multiplication table example, of you understand that 2 *2 = 4, and not just know it, you can also deduce that 20 * 20 = 400, even though it isn’t in the actual multiplication table. Understanding is where Artificial Intelligence has a home. It tried to find ways to make computers understand knowledge so they can make further knowledge without human assistance. Wisdom is a uniquely human trait, and it relies on non-linear, non-deterministic trains of thought to arrive at conclusions. This is where moral right and wrong as well as any emotional decisions are determined, and this is where the ultimate difference between a human and the most advanced AI will always be. Until science can begin to understand the mechanisms of wisdom and consciousness, computers will not replace humans in any meaningful way.
It wasn’t a blowout – In the excitement surrounding Watson’s trouncing victory, the accomplishments of Jennings and Rutter are unfairly overshadowed. In the face of the one of the most powerful AIs known to mankind, two humans actually accumulated some points. They may not have won (not by a long shot), but the Jeopardy demonstrated two important things: 1) Supercomputers are wrong sometimes, 2) Humans were able to beat a massively parallel supercomputer in a game of analytical speed at some points. I find that pretty awesome.
Ultimately, it’s vastly more important to focus on the advantages that Watson can potentially bring to civilization than to dwell on science fiction doomsday scenarios. The reality is that the natural language data mining technology in Watson could be a life-saver in the world of medicine. With every medical journal, every textbook, and every clinical trial in its array of data, Watson could become the ultimate diagnostic tool. Used as an Internet search tool, Watson could beat Wolfram Alpha at its own data mining game. The key difference between Watson and other supercomputers is its ability to parse natural language, and this is where any future application of IBM’s DeepQA technology is going to go.
Watson isn’t going to become Skynet anytime soon. We should be focusing on the huge potential and applications this technology promises to deliver and not worrying about science fiction doomsday scenarios that are as unrealistic as the works of fiction they spawned from.
IBM researchers will be the first to tell you that they have big plans for the tech, and those plans are certainly not world domination (at least not yet...).