Machine Learning: The Curiosity of Data
There are a lot of things we ask of those interested in working at Dickson, but there is one quality we’re always on the lookout for. It’s curiosity. The desire to learn is fundamental to one’s success. That’s true in life as much as it is here at our company.
In today’s society, that trait is becoming increasingly important for our machines. You may not realize it, but our machines are learning. Not only are they smarter than ever before, but they’re adapting to the data they’re given to optimize their procedures in order to better their efficiency. It’s a type of artificial intelligence that limits the need for programming to help learn.
A simple example of this is often on display by IBM’s Watson. By asking people a few simple questions, Watson recommends a beer to the answerer. Each answer that the system receives is an additional data point that helps the machine provide better recommendations. Below is a video that details the process at work.
This is a very simplistic example of a very bold idea. The ability for a machine to almost immediately calculate data points and come to a conclusion can mean a tremendous amount for those scattered across the globe. Its potential is incredible. IBM believes that the investment in the practice will lead to a better future, and not just one that helps recommend you your next craft beer. It could cure disease and even positively impact world hunger.
The practice isn’t perfect though and is directly impacted by the quality of the data delivered. Microsoft learned this the hard way when it launched Tay, an AI twitter bot, that used machine learning to communicate on social media. The bots data points were driven through communications with others on the networks. Thanks to the types of language and comments directed to the bot from the community, Tay had to be taken down within 24 hours of launch for tweeting out offensive comments it had learned to the world.
The curiousness of the entire project means that attentive humans and machines can, and will, both learn from each other. While the potential for downfall is concerning, the potential benefits are mind blowing. Eventually, we at Dickson even hope to collect enough data to practice machine learning ourselves. Imagine being able to alert you before you experience an excursion because of trends our system has recognized in the past. That’s the potential we see in the future. Now we just have to maintain our curiosity until we learn how.