When I attend customer engagement and industry events, I inevitably field lots of questions that are close to the heart of a data scientist. Many executives are confused by the concepts of machine learning, deep learning, memory-based learning, and artificial intelligence. They wonder about the differences in these technologies, how everything fits together, and what they need to pay attention to. They wonder whether they need all of it or just some of it, and what they need to do to get started. And, yes, I hear people ask whether the ultimate goal is to replace humans with computers. In this post I will try to answer these frequently asked questions.
Let’s begin with a quick primer in this conglomeration of technologies.
- Memory-based learning enables humans to infer important connections within large, complex data sets by drawing analogies between new data and what has been seen in the past.
- Machine learning uses algorithms to learn iteratively from data, identify patterns, and predict future results—all with minimal human intervention.
- Deep learning is a branch of machine learning that uses artificial neural networks that learn from experience to automatically form models that can glean insights from data to carry out tasks like image recognition, speech recognition, and natural language processing.
- Artificial intelligence (AI) draws on all of these technologies and more to do things that previously could be done only by humans. AI, for example, can use deep learning algorithms to identify a kitten in a photo of children at play or to determine the type of object on a roadside and whether that object is likely to move into the roadway.
So do you need all of these technologies, or just certain technologies? The answer here has two parts. The first relates to the problem you are trying to solve. If you have a targeted problem, like trying to identify patterns in the records of thousands of hospital patients, you probably need memory-based learning or machine learning.
Now here’s the second part of the answer: In the long run, you’re going to need all of these technologies, because they all play an important role in the ultimate destination of artificial intelligence. Just consider the current reality of autonomous vehicles on our roadways. These vehicles are made possible by the mix of technologies that play into AI, and the future is going to bring countless other use cases along these lines.
As for that future, we are now segueing from the use of advanced analytics technologies to answer questions into a new era in which AI will ask the questions for us. The AI system will sense what’s most important in massive amounts of data, apply reasoning to understand the meaning of the data and which information is most important under the current circumstances—for example, whether the object on the side of the road is likely to move and therefore poses a risk or whether it is a sign and can provide useful navigational information—and to take actions in response to its conclusions.
Now for the question that always lurks in the background when you’re talking about AI: Are we trying to replace humans with computers? The answer is no, not at all. Instead, we are using AI to augment human capabilities, and to do things that humans could never do on their own, given the scale of today’s data.
For example, humans could never search through the data on billions of websites to identify the information that a search-engine user is most interested in. But Google does that all day long, and learns from users’ queries as it goes, so it can deliver even-better results in future searches. And it is really good at what it does. Just think of the many times that the top-ranked results on Google were the ones you were most interested in. That’s an example of AI augmenting human capabilities.
While AI is here today and will be a huge part of our future, the reality on the ground for data scientists is something else. I recently came across the results of a KDnuggets poll that looked at the Top 10 algorithms used by data scientists. The algorithms all share one thing in common: They are all focused on traditional machine learning tasks, such as regression, clustering, and decision trees/rules.
This brings us to the question of how you get started down the path to AI. In the Intel perspective, you start with your current systems and skillsets, and your experience with traditional machine learning algorithms, and then add capabilities that allow you to tackle new and harder problems, such as memory-based learning and deep learning. These new capabilities also allow you to include new, unstructured data such as images, text and video, that were previously impossible to use. New data means new insights. The idea is to begin with what you are best at, and work your way up from there.
At Intel, we are working to help organizations move along this evolutionary path by bringing together leading-edge tools for memory-based learning, machine learning, and deep learning—which, of course, are key building blocks for AI. You can learn all about these efforts by visiting our AI site, at intel.com/ai.