With each passing year, artificial intelligence becomes a more common topic of discussion as well as a prevalent feature in sophisticated systems that are already available.
Siri and Alexa, for example, are two of the most well-known instances of AI in the modern world today. However, there are a variety of other systems that are exponentially more capable of both intuitive and analytical “thought” than these commercially popular counterparts.
IBM’s Watson and Google’s DeepMind are the frontrunners of such systems, and while they both boast superior capabilities, they are quite different from each other in terms of their histories and applications.
A Little Bit of Background: IBM’s Watson
With the original goal of becoming the grand champion of Jeopardy! in 2011, Watson was created to be the poster boy of technology aimed at producing a better automated system of understanding standard search systems.
It boasts processing speeds of over 500 gigabytes a second and, in general, is the standard by which most other AI systems are judged.
Watson was created by tech giant IBM and gained its initial fame by facing off against two Jeopardy! superstars, Ken Jennings (with 74 winning appearances) and Brad Rutter (who earned the show’s biggest prize pot at $3.25 million).
Given the impressive stats and record-setting numbers of both of these contestants, defeating them was clearly a natural benchmark for which to test IBM’s new AI/analytical software machine.
The competition was admittedly close. Jennings and Rutter were able to catch up to Watson’s commanding lead at some points (especially when it came to questions focused specifically on decades).
But no matter how close the knowledgeable pair came to matching its impressive score, Watson still pulled away even further, leaving the two Jeopardy! titans looking like nothing more than trivial competitors.
In the end, Watson beat out its human competitors by almost $25,000 in one game and an astonishing $53,147 total over two games. The result left runner-up Ken Jennings quoted as saying, “I, for one, welcome our new computer overlords.”
What Makes Watson Different from the Rest
Watson was created specifically to show that AI systems could beat human counterparts in Jeopardy! and as such, is ingrained with a high level of natural language processing.
This foundation gives it considerable power over other AI systems in computing and analyzing both written and spoken text. The intricacies of Jeopardy! questions, then, are easily spotted and analyzed by Watson’s systems.
It uses hundreds of algorithms (all calculated simultaneously and in the span of about 3 seconds) to determine the correct answer to a wide variety of questions that may be posed in common parlance – an aspect of modern language that might bring other systems to a computational standstill.
The applications for the Watson system, though, are primarily confined to question-answering, making this quality both its strength and its weakness. While it excels at knowledge-based applications, it has a hard time understanding unfamiliar systems or finding creative solutions.
It’s currently in the process of being applied to a wide range of industries like healthcare, retail, and travel to provide split-second decision-making power that can boost productivity, accuracy, and more.
A Little Bit of Background: Google’s DeepMind
Similar to Watson’s triumph over its human rivals on Jeopardy!, Google’s DeepMind first garnered the majority of its fame after it bested the Go 18-time world champion Lee Sedol in 2016.
The competition between AlphaGo, a DeepMind-created AI system built specifically for competing in the highly variable strategy board game Go, and Lee Sedol was, in fact, little competition at all. In the end, DeepMind’s AlphaGo bested the world champion in a set of 5 games at 4 to 1 – a significant triumph for man and machine alike.
This game of strategy in particular has long stood as a hurdle for AI due to the level of intuition required in the game. Beyond that, with more possible positions than there are atoms in the known universe, mastering this strategic game can be an especially daunting task.
AlphaGo’s triumph over Lee Sedol signified a first for artificial intelligence as many believed such systems were simply incapable of besting any human champion.
What Makes DeepMind Different from the Rest
Google’s DeepMind is far different from IBM’s Watson. This system, rather than being produced for a particular purpose, was created as a more open-ended approach to how “deep learning” capabilities can make use of incoming information.
Rather than being tasked with trying to figure out the correct answer to a certain question then, DeepMind is instead fed a variety of information from different sources without any particular goal in mind.
How this AI system makes use of and changes based on this information is called “deep learning.”
For instance, DeepMind was given a number of vintage video game without any previous input. Over time, DeepMind was able to not only learn from its mistakes and failures while playing through the games, it even surpassed the abilities of a human expert in a number of them.
The most impressive aspect of all of this is that DeepMind learned to do so without any guidance from developers.
As such, it can be applied to numerous fields to help identify patterns that the human eye may not be able to detect. In healthcare, for example, DeepMind has been used to automatically analyze test results to spot the early signs of disease.
IBM Watson vs. Google DeepMind
While modern AI systems are becoming increasingly powerful with each passing year, the cases of IBM’s Watson and Google’s DeepMind show that artificial intelligence can be both created and applied to the world in different ways.
With Watson’s highly structured creation and application, knowledge is the ultimate goal. DeepMind, on the other hand, seems to have the idea of independent learning at the core of its being.
But while these two conceptions of AI may be quite different, they both prove that artificial intelligence is undoubtedly on the rise.
Image Source: Adobe Stock