Machines That Learn

The Second Era of Computing

Alive

Conversations About Life When Technology Becomes Sentient

Chapter (1) - So Much Has Happened
Post #6 - Machines That Learn

---

It’s easy to see that this form of intelligence augmentation will continue well into our future, but soon it will be joined with another form of relationship between us and the machine. One that does not include humanities intelligence to take part in solving any problems or making any decisions. I call that the second dilemma and I will be explaining it in detail later in this chapter.

End Of Detour

---

Machines That Learn

Our first taste of computer systems that enjoyed a level of genuine intelligence was at the start of the 21st century through the early versions of machine learning and deep learning. With machine learning, computers used algorithms to discover patterns in large sets of data and autonomously used those patterns to make predictions and decisions. Deep Learning used raw data to create higher levels of abstraction—which did not exist in the original data and and from those levels, built further layers to extract the essence of a concept and grasp a better understanding of the world presented to it.

My favorite example of the way we created early progress with deep learning was documented in a white paper that Google published back in 2008 which later became commonly known as the "cat" paper. Using large-scale brain simulations, Google researchers set up a system to, well, go and watch YouTube. Yep, that was the task. They didn’t really tell the machine what to do or what to look for but used deep learning algorithms to teach it to look for patterns in what it “sees”. Frame by frame it started to “watch”, turning an image into thousands of meaningless pixels and seeking out commonalities and logical trends.

Allow me to digress here for a minute and please note that those machines could—can—fully see, though they have no eyes, through a digital record captured by another machine—a camera. They can recall a past stored in megabytes instead of memories. My point is, they can do what we do and more. They do it differently, but faster and better. More intriguing, when you really think about it, we—despite our ego—are similar but not necessarily the superior race. We too run on hardware, though more vintage. Our cameras are integrated with the processing hardware in our skull and our memories are stored and retrieved through electrical signals too.

To find our similarities, don’t dwell on how they are not human—look at us as if we were machines…

… which, I assure you, we are. Change the vantage point and then you’ll see we’re not that different after all. Please keep that thought for later chapters.

Anyway, back to the experiment. It didn’t take long for the machine to recognize its first “thing” favored by YouTubers … a cat.