In June, Stephen Wolfram was asked to advise Congress on how it might regulate the social media platforms whose algorithms were force-feeding the most extreme viewpoints to people, creating cocoons of information that lead to polarization.
Wolfram, CEO-founder of Wolfram Research and a pioneer in the development and application of computational thinking, came up with a surprising solution: More bubbles. “I’m proposing bubbles,” Wolfram testified. “I think it’s the way of our species.”
Wolfram suggests taking the decisions out of the hands of Facebook, Twitter, YouTube and other platforms. Let outside brands develop algorithms that can sort digital media feeds, and let people choose the ones they want.
This summer, Wolfram was in Washington to explain the mysteries of Silicon Valley’s impenetrable algorithms. His hearing focused on the addictive qualities of digital platforms, and how they compel users to obsessively check their services. There are subtle mechanisms that hook people, and one of those is a steady diet of extreme political viewpoints served up by mysterious algorithms catering to people’s stated, and unstated, preferences.
“Let’s just open the AIs and audit them, and check that they’re not doing anything wrong,” Wolfram recalls lawmakers suggesting. “And I was like, ‘That just can’t possibly work. That’s just a completely doomed idea.’”
We recently talked with Wolfram by phone to find out how the ideas bubbling out of Congress could shape the future of social media.
Why did you propose to Congress to hand over the algorithms, giving outside companies the ability to sort posts for users in News Feed?
The reason I agreed to testify when I did was because things were being proposed that didn’t make any scientific and technological sense, and so I thought I should point that out. And then I realized, “Gosh I’m just going to come and say a bunch of negative things about things that won’t work. I better come with suggestions of things that might work.”
To my considerable surprise the suggestions that I came up with [have] had a lot of interest and positive feedback. And basically nobody—and I’ve talked to a lot of people about them now, including people very involved in the actual implementation of these kind of things—nobody has come and said there’s a fatal flaw. So I think that’s quite interesting.
Are you saying there’s been receptivity from within the platforms?
I would say yes ... The one thing that is easy to see is [Twitter CEO] Jack Dorsey tweeted about my testimony.
What entities would create the social algorithms if not Facebook and Twitter?
The thing I was most dealing with was automated ranking of content. That is one of the big issues, obviously. There is lots of content out there in the world. A lot of what is important about these platforms these days is their ability to rank what content a particular person should see and when. That seems to be the place where there’s sort of a lot of angst. How is that ranking working? What’s blocked? What’s ranked highly? How does that all work?
These platforms have gotten themselves into what seems to me to be a very difficult business position. They are effectively trying to become kind of moral arbiters for the world. By saying this is content we think should be here, this is content that shouldn’t be here. That’s a very difficult position for them, and people will inevitably be upset with them.
Will there ever be a perfect AI moral arbiter?
The answer is no. We humans would have to agree this is the perfect ethical code we want to follow, and the reality and the lessons of history say that’s not something everybody is going to agree on.
So, what do you actually do? You can make the final ranking of content be something that users are not trusting these platforms to do but something that they’re trusting brands to do. Trust entities that stand for something people know about. These entities might be media, they might be Fox News, they might be Disney. They might be some influencer from YouTube. When you say it’s a Disney-ranked thing or it’s a Fox News-ranked thing, people would have some concept of what that should mean.
So how would this work?
Build an AI that can capture values, one might call it preferences, associated with some particular brand. Then insert that AI as part of the pipeline of content delivery into one of these large-scale platforms.
Why would people want a brand to run their News Feed?
I trust my final ranking provider, which is some brand that I’ve known about for decades, to be the thing that is defining my way of thinking about the world that represents the values that I want in my News Feed.
It’s nothing really new. People have always had a choice to read this newspaper or that newspaper. We’re in the situation where because of the technology of these platforms and the network effects that exist, it’s kind of like only one newspaper that is intimately tied into the infrastructure of these platforms.
There has been a lot of concern that CEOs like Mark Zuckerberg decide what can and can’t be seen on platforms. How do you address that?
One of the mistakes that shouldn’t be made is [assuming] that somehow technology will solve the problem. There’s a technical component to it that is quite sophisticated and complicated, but it’s not something that at end of day Mark Zuckerberg or anyone is going to be able to say “We’re going to have AI figure out what the right thing to do is.” It’s philosophically impossible. Ethics is not like mathematics. There’s no right answer. It’s something that people have to decide what they want.