Why Data Science is a Thing of the Past?
Updated: Feb 9, 2020
Facio's Artificial Intelligence is substantially different from any previous Big Data models and AIs, allowing it to give up its pieces for what a human would consider to be a transient initiative.
The global insurance industry has been paying close attention to Facio, which is the self-proclaimed “world’s most powerful Big Data insurance cloud.” There are several reasons for this. Most prominently, however, is the fact that Facio has built Artificial Intelligence that is substantially different from any previous Big Data models and AIs, allowing it to give up its pieces for what a human would consider to be a transient initiative.
This is a big deal. Facio’s technology broadcasts a move beyond traditional data science and toward a future of self-learning algorithms that do not require data cleaning, model generation and even data scientists. It signals that we are entering the futuristic world that we have always envisioned, where machines are taught simple rules and use those rules to automatically generate immense value for all of us.
Back To AlphaZero...
“By contrast, AlphaZero had not been taught any chess strategies by its human creators—not even standard openings. Rather, it used the latest machine-learning principles to teach itself chess by playing against itself. Nevertheless, out of 100 games that the novice AlphaZero played against Stockfish 8, AlphaZero won 28 and tied 72—it didn’t lose once. Since AlphaZero had learned nothing from any human, many of its winning moves and strategies seemed unconventional to the human eye.” Yuval Noah Harari
To understand why we are shifting away from data science, it is helpful to analyze one of the key breakthroughs in artificial intelligence. Back in 2016, researchers from DeepMind, which is Google’s artificial intelligence subsidiary, created an artificial intelligence system called AlphaGo that defeated the world’s best Go player. It was a groundbreaking achievement that was documented in a popular documentary on Netflix.
That AI was successful, however, by simply trying to mimic world-class players. DeepMind researchers wanted to build an AI that learned on its own (a so-called “tabula rasa”). It eventually did so, calling the result AlphaZero. The “Zero” indicates that the algorithm had no prior knowledge of Go, chess, and shogi (Japanese chess) except for the basic rules. At first, the algorithm made random moves. But two hours, AlphaZero started performing better than other chess players. Four hours later? It beat the best chess engine in the world.
Stupid Algorithms Learn Faster
“A good learner is forever walking the narrow path between blindness and hallucination.” Pedro Domingos
AlphaZero was such a big deal because it moved beyond the standard prediction model. Many prediction models are based on a mix of sophisticated search techniques, domain-specific adaptations, and handcrafted evaluation functions. Human experts have tested and refined this combination over several decades. These prediction models do everything from displaying ads that will most likely resonate with us to recommending our next movie selection on Netflix. As a self-learning algorithm, AlphaZero is completely different, delivering better results in a quicker amount of time. There is no need for traditional data science tasks like data cleaning. Instead, the algorithm, once it is presented with hard and fast initial rules, gets to work.
AlphaZero was created as an AI for playing games, but its effects have gone beyond that. This is where Facio comes in. Facio developed from scratch a self-learning algorithm for insurance. It uses a similar “trick” to AlphaZero, but on a quantum computer. Using only the underwriting rules and the carrier’s historical data, the company achieved tabula rasa in less than one hour. It obtained a superhuman performance level of insurance predictive capabilities, allowing it to underwrite insurance policies most efficiently. This is in contrast to AlphaZero and Google, which obtained tabula rasa in four hours.
The Underwriters Hunger Games
Within the hour, Facio’s algorithm was quicker and more accurate than 15,000 combined underwriters at the same carrier. Like AlphaZero, the algorithm does not need data scientists to clean data or generate models. Instead, the Facio team taught its algorithm the rules of insurance (including things like product coverages, claims, and underwriting). It then loaded the insurance carrier’s historical data into the algorithm and let it go to work. From there, the software underwrote and settled claims for millions of policies by itself.
As you can tell, firms like Facio show that there is an extreme paradigm shift occurring. Starting with some simple rules about a game or industry, these advanced algorithms can quickly interpret the offered data to offer truly novel insights for its users. All of the hard work in preparing and cleaning data can be bypassed. For as much value that data scientists provide, AlphaZero and Facio’s groundbreaking results show that the times are changing.
“The only limit to AI is human imagination." Chris Duffey
Ultimately, the world is changing. From AlphaZero’s stellar results in chess and Go to Facio’s successful work in the insurance industry, self-learning algorithms will become an increasingly important part of our society. Whether you are a data scientist yourself or have a management role in a tech company, it is worthwhile to monitor these trends. Self-learning algorithms are only getting stronger and will transform our relationship with learning machines.