Here, we talk to the podcast’s executive producer and co-host, Katherine Gorman
Nothing in technology scares people more than machines replacing their jobs, and further down the road, taking over the universe.
Most major media outlets capitalize on this fear with pieces such as the BBC’s “Will a robot take your job?” quiz, and scores of movies including the Matrix and Her imagine automation and artificial intelligence worst scenarios.
But when many people think of “machine learning,” a subset of artificial intelligence (AI) where computer algorithms can receive new information and learn without supervision, they concern themselves with humanity’s ability to survive a cybernetic army that has become too intelligent for our own good: also known as Terminator‘s Skynet.
And yet, this fear is completely off topic.
That is one of the points Executive Producer and Co-Host Katherine Gorman is trying to get across with her biweekly podcast, Talking Machines, which you can download or stream in a variety of ways. The show’s creators are pushing a Kickstarter campaign to fund the second season as they wrap up their first 20 broadcasts. For anyone interested in learning about machine learning in real depth, they serve as a wonderful crash course. With 11 days to go, they have raised almost one third of their $45,000 goal.
She laments how science fiction and media hype about the incoming AI apocalypse obscure where the industry is at and also masks some of the other kinetic – not potential – ways the technology is having an impact.
“The research is nowhere close to creating [Skynet] or the things Elon Musk is worried about. We want to move the conversation back to reality,” she tells Geektime.
“Machine learning is about creating algorithms that feed on data and can learn on its own,” she says, creating an “elegant” development in computer science where data is analyzed and synchronized faster.
For example, “Research into machine learning and AI has created some very powerful tools that can recognize cats and understand what movies we like from our previous preferences, but that’s the same grasp on the world that maybe a two- or three-year-old human has: It’s just nowhere near becoming self-aware,” what would be necessary for artificially intelligent machines to act like full-on humans.
From NPR to independent podcasting
After a decade at NPR (at the beginning of which she was studying with this reporter at the University of Michigan), machine learning grabbed her attention because of its enormous presence in conversations taking place in the research and development world. At the beginning of this year Talking Machines, a biweekly podcast aimed at re-informing listeners where machine and robotics research really is, first aired. The show is co-hosted by Ryan Adams, an assistant professor of computer science at Harvard University whose professional focus is machine learning (ML), and is the flagship broadcast of Gorman’s new production company, Tote Bag Productions.
To get listeners interested, guests frequently are the biggest names in the field today: academics, entrepreneurs, and the like. The response has been overwhelming, bringing in 30,000 people per broadcast after launching only a year ago. A central theme of the show has been the technology’s applicability across fields, such as business, biology, sports, and advertising.
Where is machine learning advancing?
Machine learning excites Gorman most in its big data applications.
“Using data science and being able to apply these quantitative views to formerly qualitative fields changes the way we explore things,” she says. “You can explore your field in a way you couldn’t before at all.”
For example, she mentions “David Blei of Columbia,” who she cites “is really at the forefront of the field. A while ago he was working on a project to use machine learning, specifically an approach called topic modeling, to sort through all the cables that had been sent between foreign offices during a certain time period. Using ML to automate that work allows the researchers on his team to greatly expand the amount of ground they were able to cover.”
She also explains how machine learning has completely revolutionized the field of advertising, and says Claudia Perlich, who they interviewed on the show, “is really on the edge of that work.”
Gorman believes that it is clear that predictive algorithms are the hottest trend in the field right now, pointing out that even preference plugins on sites like Netflix or Amazon are an everyday expectation that are building blocks for machine learning and deep learning projects. Going forward, she is excited about where the fast-paced industry will likely be in the next few years.
“We’re going forward being able to interact verbally with our devices, to respond in a natural way so you feel like you’re having a conversation. It will change the way we interact with handheld devices and give it more of a human-like flow.”
However, she clarifies that by a “human-like flow,” she does not mean Scarlett Johansson’s Siri-esque character from Her. Rather, she thinks that “a device that has a more fluidly human interaction feel would be great: a device that understands my accent, or is able to fully understand my question or do very accurate transcription of text. That’s the level that I’m talking about.”
She also believes this level of “humanness” will increase technological adoption. “I think it would really lower the barrier to entry to using technology, which as it becomes cheaper and more ubiquitous would be even more impactful.”
How will smarter machines impact our lives, both for the better and the worse?
Critically, the show explores major social questions brought on by more efficient data machines.
“When these tools become better and easier to use, how will that change our daily lives?” she asks. “What sort of questions do we ask about training, economy, education?”
She reflects that, “Growing up outside of Detroit, [I] watched that very sea change, of humans to automation, first hand. I think what scares me the most about it is that our educational system is not ready for a world in which those jobs that might become automated, are automated. What an educational system that is ready for it looks like, I don’t know but it will be very different and I think getting there will take a lot of cooperation from a lot of slow moving institutions.”
On the bright side, she thinks more efficient tools will allow humans the ability to flourish in a variety of ways. “Think of the time and space people will have to be creative, explore, and advance human understanding if smaller tasks are automated.”
Ultimately, she compares the changes in the years and decades ahead to the Industrial Revolution, when the kinds of employment people were getting were greatly different from the jobs their grandparents had.
How the Kickstarter campaign will support the show
The bulk of the Kickstarter campaign will finance expanding online offerings, production quality and reaching a larger audience. Some of it will go to attending industry conferences, where Gorman and Adams will able to get in touch with and interview a number of people in one place.
“We’ll see how Kickstarter goes. We are inviting companies and projects to sponsor the show right now.”
Gorman is hoping this is just the beginning, optimistic about the interest the show has drummed up thus far. Furthermore, she sees this as a big chance for the public at large to demand a newer, stronger level of attention from the media on scientific developments.
“Any time spent on science reporting, for conversation to unfold, talking to the real researchers at the forefront of the field is a huge testament to sustainable science reporting.”
Gedalyah Reback contributed reporting.