Artificial Intelligence: Forget the Job-Killing Hype. Get Ready for a new era of Better, Faster, Cheaper Predictions (Part 2)
(This is the final episode of our two-part series on artificial intelligence, featuring AI expert Avi Goldfarb (avicgoldfarb), author of the highly-acclaimed “Prediction Machines: The Simple Economics of Artificial Intelligence. Read Part 1 here.
With the help of insight, observations and examples, Goldfarb shows us why we should focus on the astonishing evolution of artificial intelligence (AI) as a drop in the “cost of prediction.”
He also revisits his simple framework for understanding what the AI trend means for your organization.
The truth is, these are uncertain times.
Since the 1960s, the average lifespan of a company on the S&P 500 has shrunk from more than 60 years to less than 20.
Researchers expect that number to drop to just 12 years by 2027.
Meanwhile, 40% of the companies on the Fortune 500 today will disappear within the next 10 years.
Nobody really knows why the casualty rate for large organizations is higher than ever.
But we don’t have to know the answer to take advantage of AI’s amazing ability to uncover hidden patterns in truckloads of data to:
- Crush uncertainty
- Make better predictions
- Create new opportunities for revenue and growth
To paraphrase MIT Professor and Author Erik Brynjolfsson’s review of “Prediction Machines”: If you want to cut through the fog of AI hype, your first step should be to read this blog.
Hope you enjoy the conversation.
Appian: You’ve talked about the challenge of distributing the benefits of AI in a way that people think is fair and just? And that’s a hard question.Which touches on another hot topic—the ethical considerations around the use of artificial intelligence and intelligent automation.
Goldfarb: I think ethics are a very important part of the AI conversation, and in lots of different ways. One is an ethical question that we’ve been talking about for a long time, which is the ethics of running a large organization, and thinking through the trade-offs we get with the efficiencies that affect people in an organization.
Artificial intelligence is a technology of low-cost prediction and discovery, writes @SteveLohr via @nytimes w/ comment by @professor_ajay + Prediction Machines w/ co-authors @avicgoldfarb @joshgans https://t.co/nwdVlMu323
— Rotman School (@rotmanschool) October 21, 2018
Dealing with the Ethical Issue of AI Bias
Appian: But that’s not unique to AI.
Goldfarb: No, but it’s certainly a key part of the conversation. There are also other ethical questions around jobs and replacement of people. So, number two is: How good does the technology need to be to replace humans?
A third one is how do we codify what we value. The final piece has to do with the fact that the data machines are trained on is provided by humans. And this means the machines are going to be biased.
And the ethical question is: How do we feel about running machines that are less biased than humans—but still biased. The challenge is, how do we deal with that ethical issue?
Appian: That’s a good segue into the next question. In business, we talk a lot about leveraging AI to streamline business processes and reduce labor costs. But your expectations for AI are much more transformative than just taking costs out of an organization.
What’s the risk of getting hung up on efficiency? And if we do that, are we missing the real business value of AI as a transformative technology?
The Evolution of Machine Translation and Trust
The most exciting opportunities around AI are not around costs savings. They’re around transforming business and creating new opportunities—new business models that were not possible before, because our prediction capability was so bad.
Take machine language translation. It’s a long way away from being as good as human translation. But it’s still good enough for lots of contexts, like allowing people to trade across borders in a way that wasn’t possible before.
There’s some recent research which says that this can have a significant impact on a company’s ability to sell abroad digitally.
The evolution of translation capability allows for better communication and trust, which wasn’t possible before.
An animated introduction to Prediction Machines https://t.co/wDO1LjTO0k
— Joshua Gans (@joshgans) September 19, 2018
Appian: What about the impacts on customer experience. How will better prediction capabilities impact customer experience?
Better predictions should lead to much better customer experiences in all sorts of ways: From machine translation that we can better understand people, to better predictions about what I’m going to purchase, so that I get what I want as a consumer.
Taking Prediction Retailing to a Whole ‘Nother Level
Think about Amazon’s recommendation engine, which can currently recommend what you might want to buy. And it’s pretty good. Maybe for every 20 things they recommend, I might buy one. And if you realize that they have millions and millions of items in their catalog, the benefits of making better predictions can really add up.
But at the same time, Amazon’s business model looks like an old-fashioned catalog company. At a very high level, it’s the same as the Sears Catalog model was. It’s online. And they do the processes much, much, better. But there are parallels to shopping on Amazon and looking through a catalog, picking what you want, and then having it ship to your door.
Appian: So, with better prediction capability, they wouldn’t have to wait for you to buy?
They potentially could just ship stuff to your door because they know you want it. You could open the box at your home, and you’d be pleasantly surprised, because you use detergent and they shipped you the right brand of detergent at the right time.
It might not be perfect. And they may have to figure out how to deal with returns. But my point is that better predictions will improve the customer experience in all sorts of ways and create new opportunities for business.
AI Is Getting Better at Filling in the Blanks
Appian: So, with the fast evolution of AI, there’s a lot of talk about us being at a tipping point in terms of AI adoption? Or is that just hype?
Goldfarb: We’re seeing plenty of great companies adopt AI and identify important use cases. At the same time, we’re also seeing companies that thought they were going to be able to use AI but couldn’t figure out how to take advantage of it.
So, yes, there’s absolutely hype. And in some contexts, that hype is going to be overdone. But at the same time, this is a technology that’s changing the way businesses operate.
So, as I mentioned earlier, what machines have gotten very good at is filling in missing information, making predictions.
And so, as an economist, I think about, well, making predictions is getting easier. We can think about that as a drop in costs. And we all know what happens when something gets cheaper: We want to buy more of it.
So, as prediction gets cheaper, as it gets better, we’re going to do more and more prediction by machines.
And 10 to 20 years from now, the most successful companies will be using AI in some form to stand out from the crowd.
Remember how in the late 1990s, there was so much hype around the internet? Well, a lot of people made bad bets on it.
But it’s hard to say that after 20 years, internet technology hasn’t been transformative.
Moneyball: The Art of Winning with AI and Big Data
Appian: Speaking of transformative, you also talk about machine learning in your book.
Goldfarb: What’s happened in the last 10 years and especially in the last six years, is that a particular branch of AI called machine learning has improved to the point where a lot of things that just 10 years ago we thought of as inherently human problems can now be done by machines.
The reason we’re talking about AI in 2018, and we didn’t talk about it in 2008 or 1998, is because machine learning—this particular type of AI—has gotten much, much better over the last decade.
Machine learning is a type of AI. But in the context of what we’re talking about today it is “the AI.”
Appian: You’ve also talked about how the rise of AI will drive up the value of data. What does that mean for organizations that deal with massive amounts of data, like banks, financial services and healthcare companies?
Goldfarb: Prediction machines require data to work. The more data you have, the better predictions you will be able to make.
And so, companies and industries that deal in large amounts of data will benefit most from the AI trend.
But the key opportunity will lie in the ability to identify new types of data and create the infrastructure to collect these types of data.
So, it’s not just about the data you have right now. It’s also about the kinds of data you have the ability to collect going forward.
AI Meets the Internet of Things
Appian: Which brings to mind the massive amounts of new data that will be generated by the Internet of Things.
Goldfarb: Exactly. Particularly for organizations that have a relationship with an end customer. That’s an opportunity to collect and own data.
And so, lots of companies that have these direct (IoT) relationships with customers will have enormous opportunities to leverage AI.
Appian: Finally, what are the biggest trends on your radar for 2019 and beyond?
Goldfarb: I expect to see more and more companies bringing AI into their organization.
So, in some sense 2017 and 2018 were about all of us understanding that the leading tech companies in the world—Amazon, Google, Microsoft, Facebook and others—were emphasizing and investing in AI.
I think what’s going to happen in 2019 and over the next couple of years is that more and more companies in insurance, finance, healthcare, retail, etc., will be looking into and taking advantage of these technologies.