Forget all the hype around artificial intelligence (AI), if you've really got the skills that Silicon Valley and Wall Street need, you could haul in millions of dollars a year.
“Right now, AI is an elitist sport – there are very few people who know how to practice it,” said Tom Eck, the CTO of industry platforms at IBM and a software developer who has been involved in AI as far back as the early ’90s. “The top-tier AI researchers are getting paid the salaries of NFL quarterbacks, which tells you the demand and the perceived value."
While Eck, speaking at the Markets Media’s Summer Trading event in New York, didn’t specify whether he meant starting quarterbacks or their backups, he did say that the best AI researchers are earning as much as the highest paid position in the National Football League. For reference, there are 31 quarterbacks who will make more than $5m this year, and the highest-paid QB in the league – the Oakland Raiders’ Derek Carr, fresh off signing a new contract last month – will get a cool $25m this year. He’ll earn a minimum of $70m over three years and could get up to $125m over five.
Financial services, healthcare and advertising are the three biggest adopters of AI at this stage, according to Eck.
AI is the act of imbuing a machine or a piece of software with the capabilities that we consider human cognition, basically making a machine act like a human brain. If we take neurons and model them mathematically as a simple formula with nonlinear components, we can create an interesting rules-based system using “if this, then that” algorithms capable of pattern recognition, Eck said.
But it's still at the early stage.
“AI is statistics-based – here’s a set of data; take the data, extrapolate and give me a prediction of the most likely thing to happen,” he says. “That’s about model-generating, given a set of data.
“The machine learning system has to come up with a generator that would spit out that data set, working backwards from the data to a mathematical model, and then it is no longer limited to the data you’ve observed – AI can predict what tomorrow’s equity prices will be.”
Deep learning uses the same neural-network type of approach as AI but at gargantuan sizes. All models are fueled by data,
“Big data really means more data than you have the capability to deal with: high volume, high velocity that you want to process quickly and variability,” Eck said. “The data science role not going away anytime soon – with data, it’s garbage in, garbage out.
"Deep learning is more challenging because of volume, velocity and variability,” he says. “The interesting thing about deep learning is its ability to operate on unstructured data."
The next step in the process of evolution is cognitive computing, a system that’s imbued with the ability to understand, reason, recognize the context and learn – and it has to have a human interface such as a natural language base.
Photo credit: Dmytro Aksonov/GettyImages