It is very early days in the brave new world of artificial intelligence. We have companies like Facebook, Microsoft, Google and more getting a lot of press for early victories with chatbots, driverless cars and other AI-related tech. But to think that AI is only a reality within the tech industry would be a mistake.
To start the panel discussion entitled "Artificial Intelligence: Friend or Foe?" Bloomberg Beta Partner Shivon Zilis showed a slide to demonstrate just how vast the reach of AI already is:
If the slide is overwhelming -- that's by design. What you are looking at is a list of startups that have some sort of machine learning capacity that are impacting industries as diverse as education to agriculture and enterprises as distinct as sales and security. As an investor interested in companies that are impacting the future of work, Zilis keeps an up to date map of all the startups impacting the AI space. The list grows every day and the problems the startups are tackling with machine learning only get bigger, she said.
"We are in the very first inning of a nine-inning game, and there is no industry that is untouched by machine intelligence today," she said.
The obvious dooms day conclusion to jump to after digesting Zilis' slide is that it's only a matter of time before a large portion of our labor market is unemployed. If artificial intelligence can automate jobs as diverse as farming fields to filtering spam out of websites, there is arguably no portion of the workforce that won't be touched in some way by these technological advances.
David Siegel, the co-chairman of tech-assisted hedge fund Two Sigma, says this reality should be top of mind for anyone who touches the labor market. Another panelist argued that this technological upheaval is not unlike the Industrial Revolution. Yet Siegel warned that in fact it is very different because while the Industrial Revolution gave way to millions of new jobs, it is unclear what jobs will be created when machines begin to takeover jobs in transportation, mining and agriculture.
"Vast numbers of jobs that us humans are doing will be done by algorithms," he said. "One of the big causes for the stagnation of middle class wages is because of clever computer programs."
To focus the machine takeover on blue collar work is missing the larger problem, said Merrick Ventures founder Michael Ferro. The trend he is tracking is how the AI-revolution will impact white collar jobs and the distribution of wealth in the country. Ferro says he knows hundreds of entrepreneurs who are working on everything from sensors on cars to algorithms for training that are going to make them a ton of money.
"Nobody is susceptible to how AI will shift a thousand billionaires to a group of another thousand billionaires," he said. "If your companies haven't adopted some [AI] strategy, there is some 18 year-old or 28 year-old or 88 year-old who is going to be adopting this technology. It will be a seismic shift."
So how do we guarantee that the machines don't take over? Stuart Russell, a professor at University of California, Berkeley and vice chair of the World Economic Forum Council on AI and Robotics, put things in perspective by bringing it back to our humanity and happiness. While a majority of the people in the audience at Milken likely enjoy their jobs, that is not a reality for millions of workers around the world, he said. If we can automatize the tedious, routine and unfulfilling tasks to machines to make way for workers to do more advanced work, that's ultimately a trend that leaders should get behind.
What remains to be seen is how exactly the labor market will shift to create new jobs to replace the ones we're giving to machines, warned Siegel of Two Sigma again.
"Will automation spur growth? We should be optimistic that we will come up with all these high-value things for our labor force to do, but I don't think we can assume it will happen," he said.
Resources: linkedin.com
Automated tools never do anything right. They just do it again. As we get better and better at recognizing patterns in the complexity, we will use automation to react to those patterns.
ReplyDeleteHowever, as automation spreads, it is applied too quickly. Things that look familiar are reacted to, but incorrectly. Hopefully, the few remaining aware human beings still in the loop will act to prevent disaster. The stock market, I think, is very close to this condition. Bots buy and sell based on known historical patterns. Unknown and unrecognized history, as well as changing present conditions, quickly make the algorithms more and more flawed. At the same time, fewer humans know and understand the underlying process and we race towards perfection and chaos at the same time.
This is also happening in engineering. New students are learning computer modeling and stress analysis without exercising the underlying math and mental visualization. We're removing the fuses while we revise the circuitry. Interesting times ahead.
The point about new students not practicing underlying principles as well is nothing new. Case in point, calculators, excel spreadsheets, etc. being used perform various mathematical tasks... when was the last time you performed long division, or computed the cosine of a some arbitrary angle, or the least squares regression line for some random dataset 'by hand' ... or for that matter in your head. These tasks used to require no small amount of human effort in the past, but computers now do them in the blink of an eye... In some cases such black-box endeavors are fairly innocuous, in others they can definitely go horribly awry... I think the roles, or perhaps more accurately, the skillsets needed by humans to fulfill those roles is going to shift. Computers will alleviate the burden of tedious algorithmic computations from the end user, but the end user must be able to determine when the computer is giving back something useful, and when the results are just so much garbage...
DeleteExactly. Call it what you want, but for now and the conceivable future, machines will only learn what we've told them to learn. They live in a constrained world. Their blind spots are impenetrable. The most average human has way more sensory and analytic power than the most complex model. Providing better tools to improve human performance is a much more robust path than full automation. I've been cleaning up after MIT's Media Lab and the rest of the "Lights Out" crowd for a long time. Humans are not yet redundant.
DeleteMachine learning occurs in a much smaller subset of reality than human learning. It can be faster, it can dig deeper, but it doesn't replace human creativity.
Automation has it's place, and it will replace much of what humans do. The transisiton will be far from seamless.
Chatbots are an updated protocol of the old conversational IVR phone system we are so used to.
ReplyDeleteBenjamin Kowarsch Benjamin Kowarsch
DeleteProject Management Consultant, Japan & Switzerland
@Armando, not really. At least not the chat bots we were discussing here. In an IVR all possible responses are predetermined. An Eliza like chat bot with machine learning capabilities does not have predetermined responses. It generates text on the fly following the rules of grammar and semantics and it's vocabulary is learned from interaction with humans.
That's how the Microsoft chat bot turned into a Donald Trump supporter. This wasn't predetermined, it was learned.