Tales of AI’s impenetrability are exaggerated. Big Tech should prepare for regulators to look deep into their platforms in the near future.
There is a perfectly good reason to open up the secrets of the social media giants. Over the past decade, governments have watched powerless as their democratic processes have been disrupted by misinformation and hate speech on sites like Meta Platforms Inc.’s Facebook, YouTube and Alphabet Inc.’s Twitter Inc. Now some governments are preparing for a comeback.
In the next two years, Europe and the UK are preparing laws that will curb the disturbing content that social media firms have allowed to go viral. There has been a lot of skepticism about their ability to get under the hood of companies like Facebook. After all, regulators lack the technical expertise, manpower and salaries that Big Tech boasts. And there’s another technical hurdle: the artificial intelligence systems that tech firms use are notoriously difficult to decipher.
But naysayers should keep an open mind. New techniques are being developed that will make searching for those systems easier. The so-called black box problem of AI is not as intractable as many think.
Artificial intelligence powers most of the actions we see on Facebook or YouTube and, in particular, the recommendation systems that sort out which posts go into your news feed, or which videos you should watch next—all to keep you moving. . Millions of pieces of data are used to train the AI software, allowing it to make very human-like predictions. The hard part, for engineers, is understanding how the AI makes a decision in the first place. Hence the concept of the black box.
Consider the following two pictures:
You can probably tell within milliseconds which animal is the fox and which is the dog. But can you explain how you know? Most people would have a hard time articulating what it is about the nose, ears or head shape that tells them which one it is. But they know for sure which picture shows the fox.
A similar paradox affects machine learning models. It will often give the right answer, but its designers often can’t explain how. This does not make them completely uncatchable. A small but growing industry is emerging that monitors how these systems work. Their most popular task: Improving the performance of an AI model. Companies that use them also want to make sure their AI doesn’t make biased decisions when, for example, analyzing job applications or granting loans.
Here’s an example of how one of these startups works. A financial firm recently used Israeli startup Aporia to test whether a campaign to attract students was working. Aporia, which employs both software and human auditors, found that the company’s AI system was actually making mistakes, giving loans to some young people it shouldn’t have, or withholding loans from others unnecessarily. When Aporia took a closer look, it discovered why: Students made up less than 1% of the data the firm’s AI was trained on.
In many ways, black box AI’s reputation for impenetrability has been exaggerated, according to Aporia CEO Liran Hosan. With the right technology, you can even – potentially – solve the ultra-complicated language patterns that underpin social media firms, in part because on computers, even language can be represented by numerical code. Figuring out how an algorithm might spread hate speech, or fail to address it, is certainly more difficult than spotting errors in the numerical data representing loans, but it is possible. And European regulators will try.
According to a spokesperson for the European Commission, the upcoming Digital Services Act will require online platforms to undergo annual audits to assess how “dangerous” their algorithms are to citizens. This can sometimes force firms to provide unprecedented access to information that many consider trade secrets: code, training data and process logs. (The commission said its auditors will be bound by confidentiality rules.)
But let’s assume Europe’s watchdogs couldn’t delve into Facebook or YouTube’s code. Suppose they couldn’t investigate the algorithms that decide which videos or posts to recommend. There would still be a lot they could do.
Manoel Ribeiro, a Ph.D. student at the Swiss Federal Institute of Technology in Lausanne, Switzerland, published a study in 2019 in which he and his co-authors tracked how some YouTube visitors were radicalized by far-right content. He didn’t need to enter any of the YouTube codes to do this. The researchers simply looked at comments on the site to see which channels users went to over time. It was like tracking digital footprints – painstaking work, but in the end it revealed how a portion of YouTubers were being lured into white supremacist channels through influencers acting as a gateway drug.
Ribeiro’s study is part of a wider body of research that has tracked the psychological side effects of Facebook or YouTube without having to understand their algorithms. While they provide relatively superficial perspectives on how social media platforms work, they can still help regulators place broader obligations on platforms. These can range from hiring compliance officers to make sure a company is following the rules, or providing auditors with accurate, random samples of the types of content people are directed to.
This is a radically different perspective than the secrecy with which Big Tech has been able to operate until now. And it will involve new technologies and new policies. For regulators, this could be a winning combination.
Parmy Olson is a Bloomberg Opinion columnist covering technology. A former reporter for the Wall Street Journal and Forbes, she is the author of We Are Anonymous.
