Technology has come a long way in the last decade and machine learning and artificial intelligence are perfect examples of how a machines can help us get the job done. But before you buy a piece of technology for your business, you need to take a step back, pinpoint your core problem, and decide if you actually need AI or ML to solve it. In this episode of the Dr. Dark Web podcast, host Chris Roberts answers vital questions regarding AI and ML, such as, “Do you actually need AI to solve your problems?“; “How can you find a suitable vendor for your solution?“;”Do you have enough data to train an AI model?” and more.
⚡Do you actually need artificial intelligence to solve your problem? This is the first question you need to ask yourself when you are considering implementing AI in your processes. Chris explains, “If I’ve got a vendor in front of me and they’re telling me that they’ve got this amazing technology and solution, I’m going to first turn around and say, ‘Do I actually need it?’ And that’s an introspective look at this and go, ‘Hey, have I identified the people? Have I identified the process? And this is the technology, but which process is broken?'”
⚡ Find a vendor with proven domain expertise. Another critical question to consider when implementing technology in your processes is, “Where do I apply ML and AI?” Chris explains, “If you’re looking to solve, let’s say, a financial issue. So, you run a financial institute, you want to model financial data, you want predictive analytics run on like the stock market, and you want somebody who understands that type of data. You want somebody who understands those models just the same as us in our industry. If you’re going after an AI/ML-type solution in security orchestration or in heuristics or in user behavior analytics, you don’t want somebody who’s taken IBM Watson. […] ‘We’ll make it work’ is not a good way of dealing with it. You want somebody who has, to some degree, got a good track record. They have domain expertise in the very field that you’re working on.”
⚡ Not all AI/ML models fit all solutions. AI and ML models are not one-size-fits-all. That’s why you need to find a suitable model for your particular problem. Chris says, “It sounds easy, but all too often, we get distracted. All too often, we don’t necessarily ask what and how have you trained your solution? How have you fit what you’re doing into my model? Training is a huge thing — training data models, training data science. If you have a security orchestration model, how much are you training it? And then, quite honestly, how often are you retraining it? […] If you think about it, an intelligence system feeds off of data, intelligence, and information. It doesn’t just feed off of the raw underlying stuff you’re giving it. It’s got to learn. It has a thirst to learn. So how are you continuing to train it?”