Conversations with Asia’s leading movers, shakers, thinkers, and provocateurs |
|
In conversation with
Walt Mayo
Machines That Listen
Click here to listen to the full Inside Asia episode
|
|
|
With so much hype around AI, it’s hard to know what to think of it. In one moment, AI applications can appear benevolent, even helpful. Like when Netflix offers up viewing suggestions, or Amazon recommends an item you might like. It can also be intrusive. Facial recognition, for instance, can and has been used to track and screen innocent people simply going about their business. AI, you might say, wears many faces. One moment, it’s Big Brother watching your every move. The next, it’s R2D2 trying to save your life.
I’m dramatizing, but you get the point. Now, apparently AI-enabled machines are learning to listen; so well, in fact, that advanced technology is able to decipher subtle tones, language variances and emotional nuances, virtually inaudible to the human ear. In so doing, it can determine the sentiment of the speaker. And that, says Walt Mayo, CEO of expert.ai, can make a world of difference.
|
|
AI Beats Pollsters in Predicting US Presidential Outcome
Take the just concluded US Presidential race, for instance, Walt and his colleagues put their technology to the test and three weeks before the election, calculated that Joe Biden would receive 50.2% of the popular vote compared to 47.3% for Trump. It struck me as a bit of a marketing ploy at the time, and a risky one at that. Why would a tech company stick its neck out with this kind of prediction when every other major poll prior to the election had Biden trouncing Trump by at least an eight percent margin?
The results are now official with Biden winning the popular vote by 51% to 47.2%. Yet, the real victor was AI. How did this relatively small tech company blow the doors off a phalanx of seasoned and highly certain professional pollsters? I asked Walt to join me in this episode of Inside Asia to share his story and talk about the future of so-called “sentiment analysis” AI.
|
|
The Things Machines Can Do
Our discussion fleshed out one big question: Are we entering a period of runaway technology so potent that it could replace many of the skills we as humans have come to take for granted? Walt points out that what’s easy for humans is not always easy for advanced machines.
|
|
Programming a robot to walk, for instance, requires thousands of data inputs, feedback loops, and adjustments to manage surge, sway, heave, yaw, roll, pitch….the list of commands goes on and on. I said it was difficult, but not impossible.
Check out this YouTube video of one particularly agile bi-pedal robot from Boston Dynamics. How quickly is the robot design field moving? So fast, that robots are increasingly able to teach themselves to walk, as evidenced by a recent Google project. It took only a few hours for this machine to teach itself to perambulate by tapping into AI and a data network that contained all the information required.
|
|
The Art of Listening
But what about some of the finer skills primarily associated with humans? Like listening. As a higher species, we’re blessed with the gift of language and while these days it feels like people are more interested in talking than listening, we have the ability. The question is: Are we making the most of it? Consumed, as we are, by social media, it leaves the impression that everyone has something to say. If only that were true. Sociologists and psychologists tell us that it’s shaping our listening habits and instigating new levels of so-called communication bias. Filters, in other words. We’ve started to hear what we want to hear and less of what’s actually being said.
To some degree, that might explain the repeated failings of pollsters. And it certainly speaks to the polarization of people who are selecting to hear whatever reinforces their own beliefs or frame-of-mind. If technologies like expert.ai have an imbedded ability to detect 80+ different “sentiments,” and segment them in order to best determine what someone really believes, then what chance do we and our biological brains have in competing with that?
|
|
If you like what you hear, please share this week’s and previous week’s episodes with friends and colleagues. Is there a topic you’d like to hear more about, drop us a note, or visit us at Linked-in, Twitter, Facebook, or Instagram. If you prefer to get a regular weekly synopsis of our discussions, subscribe to the Inside Asia Newsletter by signing up at www.insideasiaadvisors.com.
As always, we thank you for listening!
|
|
Steve Stine
Founder & Host
Inside Asia
|
|
Articles, Reports & Insights |
|
Article: "What Went Wrong with Polling? Some Early Theories," by Nate Cohn, The New York Times, Nov. 10, 2020 [here]
A bit of an exhaustive - and speculative - analysis of what went wrong, how pollsters failed to make adjustments following the 2016 election, and what a final data autopsy will likely reveal.
Article: "Why Polls were Mostly Wrong," by Gloria Dickie, Scientific American, Nov. 13, 2020 [here]
The article looks at the failure of US polling in an interview with Princeton University neuroscientist Sam Wang, who tends to excuse the polling process more than explain it.
Article: "Election Polls were a Disaster this Year. Here's how AI could Help," by Arijit Sengupta, Fast Company, Nov. 16, 2020 [here]
The writer describes how AI-enabled sentiment analysis revealing not what consumers say they want, but by their actual behaviors. Amazon, Netflix and other mega e-commerce companies have long employed the technology to make proactive recommendations on what they believe a customer is most inclined to buy.
Report: "Artificial Intelligence and Southeast Asia's Future," McKinsey Global Institute, September 2017. [here]
A slightly dated report showing ASEAN's rate of AI adoption and the need for governments to take a stronger role in promoting its development and application.
Report: "Freedom to Conduct Opinion Polls: A 2017 Worldwide Update," ESOMAR and the World Association for Public Opinion Research, 2018. [here]
The two organizations sponsoring the report seek to track the freedom to gather public opinion insights through polling surveys and show that 79 out of the 133 countries surveyed for this report place some sort of embargo or restriction on public polling - an increase from three years earlier.
Blog: "Making Sense of the 2020 Presidential Polls with NLU," by Jay Selig, expert.ai, Nov. 16, 2020 [here]
This is expert.ai's attempt to explain how their technology proved superior in predicting the US Presidential election outcome.
|
|
 |
Click photo to link and listen |
James Barrat:
Our Final Invention
| August 10, 2018 |
|
Dystopian - This is not the take on artificial intelligence (AI) that you’re going to find in the popular media or on corporate websites. Most everyone is thrilled about what AI can and might do someday soon. Self-driving cars, remotely controlled delivery vehicles, homes that clean and restock themselves automatically, medical technologies that can predict (and prevent) our every illness—these are the tip of the iceberg, and for the companies that will manufacture and control them, they can’t get here soon enough. But there is a dark side. My guest this episode is James Barrat, futurist, filmmaker and author of AI: Our Final Invention. With a title like that, you know the news he brings is not good. Among other things, James and I talked about the “intelligence explosion”—the point at which a computer is able to replicate itself better than human beings can. That’s not so far in the future, James says, but for some reason its seen by very few as a clear and present danger. How far away are we from the intelligence explosion? Ten years? Fifty years? Three quarters of a century? As with all things tech, there’s the prospect of acceleration. And in the case of AI, that means we humans may not be ready. Are we looking at a handover or a takeover?: In these times of rampant corporate competition and geopolitical intrigue, what hope can we have for agreeing on the ground rules for AI development and deployment?
|
|
|
|
Click photo to link and listen |
Anurag Banerjee:
The Social Impact of AI
| Dec 3, 2018 |
|
Utopian - In this episode we focus on the utopian side of artificial intelligence. It’s not a subject you hear much about these days, and this program is no exception. Our guests have had a lot to say about AI’s possible horrifying applications. But like most things, there’s another side. In this episode we turn to AI’s better angels—and see if we can believe in them. My guest is Anurag Banerjee, Founder and CEO of QUILT.AI, an early-stage tech and consulting firm that merges machine learning with human and cultural intelligence to create insight for good and for profit. It’s a fascinating conversation.
I find myself wanting desperately to believe that AI can cure our most complex social and cultural challenge, and to worship at the altar of AI Altruism. And certainly there’s a great deal that might be possible. But it won’t come for free. We must sacrifice privacy in return for insights that could change our world. And it might be worth it. Anurag Banerjee believes that technology – powered by AI – can connect us and solve society’s biggest problems. Our machines know us. They know what we want before we do. That’s a boon for any organization with something to sell. But what about the ability to save someone’s life? Intervene in a crisis? Prevent a terrorist attack? Or slow the rate of suicides? Will large corporations with profit-hungry shareholders go toe-to-toe with tiny start-ups wielding AI sling-shots?
|
|
|
|
Prefer reading vs. listening? |
|
|
If podcasting isn't your thing, or you're just too busy to take 30 minutes to listen, check out my articles here on Linked-in where I set out to summarize each week’s conversation. We touch on a wide range of topics from healthcare to finance, politics to cleantech. I hope you’ll join us. |
|
Steve Stine | LinkedIn
Check out professional insights posted by Steve Stine, Senior Regional Leadership Advisor and INSIDE ASIA radio talk show host
Read more
www.linkedin.com
|
|
|
|
|
|
Inside Asia | WhatsApp +65 98802184 | Asia Pacific |
|
|
|
|
|
|
