If you’ve seen Maximum Overdrive, then you know exactly where I’m coming from. If you haven’t, then I highly suggest you watch it – it could very well be foreshadowing of what’s to come. Seriously, though, artificial intelligence (AI) is starting to teach itself algorithms and it’s actually quite amazing.
Over the past few decades, companies and groups have used machine learning to get a number of tasks completed. As technology as a whole grows bigger and becomes more developed, many of these groups are essentially holding a mirror up to technology’s face and showing it how to teach itself how to get work done.
Using AI in the real world
This isn’t something being spearheaded by NASA or a government-funded project that will take years to produce data which we may or may not put to use in our everyday lives. This is happening now, and it’s right in our back yard.
Scot Barton, who works with Farmers Insurance to help keep the company at the cusp of the technological future, is about to go down in history books for his latest project. Barton leads a group of engineers and researchers that study to find the answers to questions that revolve around customer behavior and even construct policies. Now the team is using data from machine learning to create a system called DataRobot, which automates a lot of difficult work involved in applying such techniques.
Other experts are also building the most basic AI-powered operating systems that are designed to make applications in the very near future as accessible as Microsoft Excel is today. Farmers’ use of DataRobot is just one example of this. Barton says he puts in raw data and AI reformats it. Then it runs dozens of different algorithms at once against it, ranking their performance. Once done, it gives Barton’s team the best options for both the company and the customer, resulting in better deals overall. It’s small scale compared to the other possibilities out there, but Barton and his team have the right idea and they’re running with it.
How feasible is self-learning AI?
In June, consulting company McKinsey published a report that put artificial intelligence in the spotlight. The company pointed out that artificial intelligence, specifically machine learning, could likely overhaul several different industries. These include manufacturing, finance, and health care, and could potentially add up to $126 billion to the U.S. economy by 2025. It sounds pretty awesome, but there’s a catch. According to the report, there is a huge shortage of AI talent and the economy just isn’t ready for it yet.
This seems to be a case of having too many jobs and not enough workers. While in the physical world, this could be a huge issue. You can’t just create humans to do jobs whenever and however you want, but machines are different. If you need a task completed, the only obstacle is programming a machine, whether virtual or otherwise, to get it done.
Many are already onboard the idea of self-taught AI, even if the proof isn’t quite all there yet. DataRobot, which started in 2014, has raised more than a hundred million dollars, including $54 million this March, to keep the train moving. The company’s CEO Jeremy Achin says data scientists are really the ones who are pushing back against the idea right now. He says that half of them feel that their skills cannot be automated and the other half worry that they will be. ““I don’t care how many people change their title to ‘data scientist’ on LinkedIn,” he said in a statement. “You’re not going to move the needle.”
So while the thought of AI teaching itself is still new, it’s already causing quite a ripple in the tech world and has become quite the radical idea. As we all know, radicals are the only ones who ever get anything accomplished, so maybe AI is on the right track.