Don’t Build Big Data With Bad Data

I was at Pure Accelerate 2017 this week and I saw some very interesting things around big data and the impact that high speed flash storage is going to have. Storage vendors serving that market are starting to include analytics capabilities on the box in an effort to provide extra value. But what happens when these advances cause issues in the training of algorithms?

Garbage In, Garbage Out

One story that came out of a conversation was about training a system to recognize people. In the process of training the system, the users imported a large number of faces in order to help the system start the process of differentiating individuals. The data set they started with? A collection of male headshots from the Screen Actors Guild. By the time the users caught the mistake, the algorithm had already proven that it had issues telling the difference between test subjects of particular ethnicities. After scrapping the data set and using some different diverse data sources, the system started performing much better.

This started me thinking about the quality of the data that we are importing into machine learning and artificial intelligence systems. The old computer adage of “garbage in, garbage out” is never more apt today than it has been in history. Before, bad inputs caused data to be suspect when extracted. Now, inputting bad data into a system designed to make decisions can have even more far-reaching consequences.

Look at all the systems that we’re programming today to be more AI-like. We’ve got self-driving cars that need massive data inputs to help navigate roads at speed. We have network monitoring systems that take analytics data and use it to predict things like component failures. We even have these systems running the background of popular apps that provide us news and other crucial information.

What if the inputs into the system cause it to become corrupted or somehow compromised? You’ve probably heard the story about how importing UrbanDictionary into Watson caused it to start cursing constantly. These kinds of stories highlight how important the quality of data being used for the basis of AI/ML systems can be.

Think of a future when self-driving cars are being programmed with failsafes to avoid living things in the roadway. Suppose that the car has been programmed to avoid humans and other large animals like horses and cows. But, during the import of the small animal data set, the table for dogs isn’t imported for some reason. Now, what would happen if the car encountered a dog in the road? Would it make the right decision to avoid the animal? Would the outline of the dog trigger a subroutine that helped it make the right decision? Or would the car not be able to tell what a dog was and do something horrible?

Do You See What I See?

After some chatting with my friend Ryan Adzima, he taught me a bit about how facial recognition systems work. I had always assumed that these systems could differentiate on things like colors. So it could tell a blond woman from a brunette, for instance. But Ryan told me that it’s actually very difficult for a system to tell fine colors apart.

Instead, systems try to create contrast in the colors of the picture so that certain features stand out. Those features have a grid overlaid on them and then those grids are compared and contrasted. That’s the fastest way for a system to discern between individuals. It makes sense considering how CPU-bound things are today and the lack of high definition cameras to capture information for the system.

But, we also must realize that we have to improve data collection for our AI/ML systems in order to ensure that the systems are receiving good data to make decisions. We need to build validation models into our systems and checks to make sure the data looks and sounds sane at the point of input. These are the kinds of things that take time and careful consideration when planning to ensure they don’t become a hinderance to the system. If the very safeguards we put in place to keep data correct end up causing problems, we’re going to create a system that falls apart before it can do what it was designed to do.


Tom’s Take

I thought the story about the AI training was a bit humorous, but it does belie a huge issue with computer systems going forward. We need to be absolutely sure of the veracity of our data as we begin using it to train systems to think for themselves. Sure, teaching a Jeopardy-winning system to curse is one thing. But if we teach a system to be racist or murderous because of what information we give it to make decisions, we will have programmed a new life form to exhibit the worst of us instead of the best.

AI, Machine Learning, and The Hitchhiker’s Guide

Deep_Thought

I had a great conversation with Ed Horley (@EHorley) and Patrick Hubbard (@FerventGeek) last night around new technologies. We were waxing intellectual about all things related to advances in analytics and intelligence. There’s been more than a few questions here at VMworld 2016 about the roles that machine learning and artificial intelligence will play in the future of IT. But during the conversation with Ed and Patrick, I finally hit on the perfect analogy for machine learning and artificial intelligence (AI). It’s pretty easy to follow along, so don’t panic.

The Answer

Machine learning is an amazing technology. It can extrapolate patterns in large data sets and provide insight from seemingly random things. It can also teach machines to think about problems and find solutions. Rather than go back to the tired Target big data example, I much prefer this example of a computer learning to play Super Mario World:

You can see how the algorithms learn how to play the game and find newer, better paths throughout the level. One of the things that’s always struck me about the computer’s decision skills is how early it learned that spin jumps provide more benefit than regular jumps for a given input. You can see the point in the video when this is figured out by the system, whereafter all jumps become spinning for maximum effect.

Machine learning appears to be insightful and amazing. But the weakness of machine learning is be exemplified by Deep Thought from The Hitchhiker’s Guide to the Galaxy. Deep Thought was created to find the answer to the ultimate question of life, the universe, and everything. It was programmed with an enormous dataset – literally every piece of knowledge in the known universe. After seven million years, it finally produces The Answer (42, if you’re curious). Which leads to the plot of the book and other hijinks.

Machine learning is capable of great leaps of logic, but it operates on a fundamental truth: all inputs are a bounded data set. Whether you are performing a simple test on a small data set or combing all the information in the universe for answers you are still operating on finite information. Machine learning can do things very fast and find pieces of data that are not immediately obvious. But it can’t operate outside the bounds of the data set without additional input. Even the largest analytics clusters won’t produce additional output without more data being ingested. Machine learning is capable of doing amazing things. But it won’t ever create new information outside of what it is operating on.

The Question

Artificial Intelligence (AI), on the other hand, is more like the question in The Hitchhiker’s Guide. Deep Thought admonishes the users of the system that rather than looking for the answer to Life, The Universe, and Everything, they should have been looking for The Question instead. That involves creating a completely different computer to find the Question that matches the Answer that Deep Thought has already provided.

AI can provide insight above and beyond a given data set input. It can provide context where none exists. It can make leaps of logic similarly to those that humans are capable of doing. AI doesn’t simply stop when it faces an incomplete data set. Even though we are seeing AI in infancy today, the most advanced systems are capable of “filling in the blanks” to cover missing information. As the algorithms learn more and more how to extrapolate they’ll become better at making incomplete decisions.

The reason why computers are so good at making quick decisions is because they don’t operate outside the bounds of the possible. If the entire universe for a decision is a data set, they won’t try to look around that. That ability to look beyond and try to create new data where none exists is the hallmark of intelligence. Using tools to create is a uniquely biologic function. Computers can create subsets of data with tools but they can’t do a complete transformation.

AI is pushing those boundaries. Given enough time and the proper input, AI can make the leaps outside of bounds to come up with new ideas. Today it’s all about context. Tomorrow may find AI providing true creativity. AI will eventually pass a Turing Test because it can look outside the script and provide the pseudorandom type of conversation that people are capable of.


Tom’s Take

Computers are smart. They think faster than we do. They can do math better than we can. They can produce results in a fraction of the time at a scale that boggles the mind. But machine learning is still working from a known set. No matter how much work we pour into that aspect of things, we are still going to hit a limit of the ability of the system to break out of it’s bounds.

True AI will behave like we do. It will look for answers when their are none. It will imagine when confronted with the impossible. It will learn from mistakes and create solutions to them. That’s the power of intelligence versus learning. Even the most power computer in the universe couldn’t break out of its programming. It needed something more to question the answers it created. Just like us, it needed to sit back and let its mind wander.