Why is Artificial Intelligence (AI) and Machine Learning (ML) so hot right now? What changed?
Advances in algorithms, technology, as well as the availability of vast amounts of data have transformed predictive analytics. It turns out that many of the most challenging problems require a lot of data to solve effectively. Humans collect vast amounts of data from our senses every day and learn from this data. Consider how you learned to ride a bicycle. The only way to learn is to try, fail, and try again until you internalize the physics of riding a bicycle. Machines do the same thing today, and in some cases, better than individual humans.
Machines, like humans, need training data in order to learn. More data is better, just like more practice can make you a better golfer or musician. The software that machines use to learn are called models. In the past, there really wasn’t a lot of training data available to build high quality machine learning models. The Internet as well as low cost storage changed all of that. Now, there are vast data sets that can be used to train machine learning models. In some cases, organizations will spend time and money to prepare training data suitable for machine learning models. High quality data is now extremely valuable. Some say data is the new oil.
Lots of data alone is not enough. The right algorithms are needed to make sense of data and learn from it. Many algorithms have been tested over the years with mixed results. Today, deep learning models based on neural networks outperform all other methods given enough data. Deep learning models excel at computer vision, speech to text/text to speech, language translation, and many other tasks. Deep learning models can also learn to play games at levels exceeding even the most experienced grandmasters.
Finally, training models with vast amounts of data requires a lot of computational power. Modern Graphics Processing Units (GPUs) are widely used to floating point calculations in parallel at very high speed. GPU instances are available at an affordable cost in the cloud.