When Elon Musk win-tweeted it Acquisition of Twitter for $44 billion On Monday night, he pledged to improve the social network by, among other things, “making the algorithms open to increase trust.”
AT TED talk earlier this month, the entrepreneur suggested that the algorithm that determines the promotion and downvote of tweets could be uploaded to the GitHub software hosting platform, making it available to people outside the company. “People can look at it and say, ‘Oh, I see a problem here, I don’t agree with that,'” Musk said. “They can identify issues and suggest changes, just like you update Linux or Signal.”
In fact, hacking Twitter to see how it actually works would take a lot more than just uploading the code to GitHub. And proving the existence—or the absence—of biases, which can be subtle in nature and depend on a variety of ever-changing factors, can be much more difficult than Musk suggests.
At first glance, more transparency makes a lot of sense. Social platforms such as Twitter, facebookand tik tak wield enormous influence and power, but are largely opaque to their users and regulators. And just as the source code of a computer program allows it to be checked for bugs or backdoors, revealing the code that makes Twitter work could theoretically show that the platform promotes certain types of content over others.
“I’m very happy to see what’s going on,” says Derek Roots, an associate professor at McGill University in Canada who studies major social platforms. Roots says he has so far refrained from teaching social referral systems to his students because they are too opaque.
While Roots admits he doubts what less moderation — another of Musk’s promised “improvements” — could mean for the platform, he believes more transparency would be helpful and hopes other social networks feel pressured to reveal more about how they work. “This could be a really interesting experiment that is long overdue,” says Roots.
This idea sparked controversy about the political bias inherent in the platform. Some to the right of political division rubbing hands in front of the prospect conclusively prove that conservative views are usually “shadow forbidden— or they have not been given the publicity they truly deserve. But they may be frustrated by the difficulty of understanding how the platform actually works.
The first problem is that there is no single algorithm that determines how Twitter decides to upvote or hide content, as opposed to Musk hinted in past. Rather, according to sources on the Twitter tech team, who spoke on condition of anonymity, the decisions are the result of many different algorithms performing an elaborate dance on mountains of data and a multitude of human actions. The results are also tailored to each user based on their personal information and behavior. “There is no ‘core algorithm’ for Twitter,” says one source at the company.
Another problem is that Twitter uses machine learning to make many decisions. For example, Twitter trains numerous machine learning models to help decide which posts to prioritize in users’ feeds based on a dizzying number of factors. These models cannot be validated like normal code; they need to be tested in an environment that reproduces the real world as closely as possible. Models also change rapidly in a real system in response to a constant stream of new data, user behavior, and input from moderators. This would quickly make them an unreliable source of information.
“In this age of machine learning, it’s not algorithms that matter, it’s data,” says David Karger, professor and computer scientist at the Massachusetts Institute of Technology. Karger says Musk could improve Twitter by making the platform more open so others can use it in new ways. “What makes Twitter important is not the algorithms,” he says. “These are the people who tweet.”
A deeper picture of how Twitter works also means discovering more than just handwritten algorithms. “The code is fine; data is better; code and data combined into a model can be the best,” says Alex Engler, a fellow at the Brookings Institution who studies the impact of AI on society. Engler adds that understanding the decision-making processes that Twitter’s algorithms are trained on will also be critical.
The machine learning models that Twitter uses are still only part of the picture, because the whole system also reacts in complex ways to real-time user behavior. If users are particularly interested in a particular piece of news, then relevant tweets will naturally be boosted. “Twitter is a socio-technical system,” says a second Twitter source. “It reacts to human behavior.”
This fact has been illustrated research posted on Twitter in December 2021, showing that right-wing posts received more reinforcement than left-wing posts, although the dynamics of this phenomenon was unclear.
“That’s why we audit,” says Ethan Zuckerman, a professor at the University of Massachusetts Amherst who teaches public policy, communications, and information. “Even the people who create these tools end up finding unexpected flaws and flaws.”
Zuckerman says one irony of Musk’s stated motives for acquiring Twitter is that the company has recently been surprisingly transparent about how its algorithm works. In August 2021 Twitter launched a contest this gave outside researchers access to an image cropping algorithm that exhibited biased behavior. The company is also working to give users more control over the algorithms that render content, according to those familiar with the work.
Releasing some Twitter code will provide more transparency, says Damon McCoyAdjunct professor at New York University who studies the security and privacy of large and complex systems, including social media, but even those who created Twitter may not fully understand how it works.
The Twitter engineering team is concerned that amid all this complexity, some code may be taken out of context and highlighted as a sign of bias. Revealing too much detail about how Twitter’s recommendation system works can also lead to security issues. Access to a recommender system will make it easier to play the system and achieve fame. It may also be possible to use machine learning algorithms in ways that may be subtle and difficult to detect. “Attackers are investigating and testing the system right now,” McCoy says. Access to Twitter models “may well help outsiders understand some of the principles used to elevate one content over another.”
On April 18, as Musk stepped up his efforts to acquire Twitter, someone with access to Twitter’s Github, where the company is already releasing some of its code, created a new repository called “algorithm” – perhaps the developer is digging into the idea that the company might it’s easy to post details on how it works. Shortly after Musk’s acquisition was announced, it disappeared.
Additional report by Tom Simonite.
- 📩 Latest news about technology, science and more: Receive our newsletters!
- Farmer’s quest to win waves of drought and flood
- Hospital robots Help fight nurse burnout
- How to download video watch offline
- Humanity vibration check yourself to death
- Climate costs the most modest battery material
- 👁️ Explore AI like never before with our new database
- 💻 Update your working game with our Gear team favorite laptops, keyboards, input alternativesand noise canceling headphones
Credit: www.wired.com /