Machine learning is a fixture of modern life, underpinning many things we do every day, but most people have no clear idea of what it is. The basic principle is this: machine learning is programming that allows a machine to learn new things based on available data, roughly the way a human would. It may sound simple, but the reality is complex. Machines, after all, can only do what humans program them to do. Machine logic is nothing like human logic. How does it work? How is it used?
The beauty of machine learning is that it is, by its very nature, flexible and adaptable. There are two kinds of machine learning: supervised and unsupervised. Supervised learning is used to quickly classify data or make regressions. Unsupervised learning is used for clustering and other tasks that explore and investigate a lot of loose data. Machine learning operations, or MLOps, integrate these machine learning tasks into a business and creates better workflows. This may all sound pretty abstracted, but these functions have a wide array of applications, from the algorithms that analyze customer data to facial recognition software to the calculation of credit scores. The main advantage that machine learning and AI has over a human task force is that a machine can analyze data in bulk; the kind of bulk data that would quickly overwhelm and exhaust a human workforce. It does not eliminate errors, and it's always best to double-check that work as much as possible with human eyes, but the potential is boundless.
Soon, we may see less talk about machine learning and more talk about AI. Many people consider them to be essentially the same. While there is still debate over whether the algorithms that use machine learning represent true artificial intelligence, there is no doubt that artificial intelligence is what the future holds for machine learning as it continues to grow and evolve. For humanity, this means that machines could continue to take on more and more complex jobs. This will likely stay in the realm of managing the digital world, from stock exchanges to cybersecurity threats. It may creep into the world of healthcare; there is a great deal to be said for artificial intelligence's ability to predict the spread of contagious diseases, for example.
As business pushes the boundaries of what is possible, a clearer image of what is likely will emerge. Potentially, an AI in a sufficiently advanced robotic interface could take on certain jobs that are currently very dangerous for humans. For example, imagine robot first responders smothering forest fires or clearing the debris after an explosion or earthquake. Robots at that level of sophistication could build human habitats on Mars or the Moon. These possibilities are endless and fascinating, but they do raise concerns as well.
When you talk about machine learning and AI, people's minds automatically turn to science fiction and apocalyptic scenarios. The real dangers are far more subtle and creeping. At this very moment, a machine learning AI is probably watching your every online purchase and subtly tracking your online movements. Most of that is benign, but there's no reason to think that it will stay benign. Similarly, imagine a drone that no longer needs a human pilot to pull the trigger. The danger is not that the drone will go rogue and turn on all humanity, but that it will advanced warfare the same way mass production advanced warfare in WWI.
There are certain dangers, but they are probably not the dangers that you imagine. The danger lies less with the AI itself, which is only a tool, and more with the humans behind these tools. The power is in humanity's hands; what you choose to use this power for is entirely up to you. There is potential for great good and benefit to all humankind, but there is equal potential for evil. The trick is to choose one more often than the other.