Why open sourcing social media algorithms is a useless sound bite
From a strictly theoretical standpoint, Elon Musk isn’t necessarily wrong about some of Twitter’s shortcomings. Content moderators do make mistakes. Spambots are annoying. An edit button would be nice if done right.
The same goes for his general theory about Twitter’s private algorithms, which can dictate the content we see on the platform.
In Musk’s mind, these internal algorithms are subject to wholly insufficient scrutiny, giving the company immense power to manipulate and shape our public square. In response, he proposes open sourcing Twitter’s algorithms, which would allow developers to observe how Twitter engineers control the platform and create their own competing algorithms. To make this happen, he’s offered $43 billion to buy the company and take it private.
“One of the things that I believe Twitter should do is open source the algorithm, and [if they] make any changes to people’s tweets, if they’re emphasized or de-emphasized, that action should be made apparent so anyone can see that action has been taken, so there’s no behind-the-scenes manipulation, either algorithmically or manually,” Musk said Thursday at the TED conference in Vancouver.
It’s certainly a virtuous goal. But as with so much about Musk’s views on Twitter, the reality is far more complex than a simple sound bite. In no space has this disconnect become more apparent than his focus on algorithms.
For understandable reasons, algorithms have become a catchall boogeyman for social media critics. The secretive algorithms powering TikTok, Instagram, Facebook, Twitter, and their peers often dictate what you see—or don’t see—on the app. That selectivity leads to suspicions about internal manipulation designed to suppress certain content, prompting calls for greater algorithmic transparency.
Musk and other advocates for open sourcing envision an idyllic platform where users could choose from many algorithmic options, rather than relying on a profit-driven company’s code. Tech writer Nathan Baschez made a compelling case for this in a post late last week, arguing the feature could “rebuild some marginal trust in Twitter as an institution, by demanding less of it.”
But two major hurdles make open sourcing of algorithms highly unlikely: scale and money.
Musk’s commentary fails to convey the sheer magnitude of today’s social media algorithms, which filter immense amounts of consumers’ personal data and usage habits through overlapping webs of code and machine learning systems. As Fortune reporter Jonathan Vanian wrote Saturday: “Like other apps built on sophisticated machine-learning tools, it’s incredibly difficult for the layman and even A.I. researchers themselves to fully comprehend how Twitter’s core service functions.”
A Twitter technologist, speaking on the condition of anonymity, told Vanian that open sourcing the company’s algorithm is a “slippery slope to open sourcing Twitter’s entire data pipeline, which isn’t worth the effort unless you can fully understand the scale of potential upsides and downsides.”
Even if a third-party developer could make sense of Twitter’s algorithms, the amount of labor and financing needed to manipulate the company’s source code and maintain the new product appears prohibitive. Consider that Twitter has spent $5.4 billion on research and development alone in the past seven years—five of which resulted in the company posting a net operating loss.
A Musk-controlled Twitter still could cut back on content moderation and offer clearer explanations for censorship decisions. Simple tweaks to Twitter, akin to the “latest tweets” scrolling option, also offer hope for reducing the impact of algorithms.
But open sourcing the company’s algorithms reeks of empty verbiage, a sentiment summarized well by Twitter vice president of product Steve Teixeira.
“The ‘open source the algorithm’ thing drops more of a potential tactic than a strategy itself,” Teixeira tweeted early Monday morning. “I suspect it would be a pretty inefficient tactic as well, given the scope of moving parts and pace of change.”
Want to send thoughts or suggestions for Data Sheet? Drop me a line here.
Buying some time. Twitter formally notified the Securities and Exchange Commission on Monday of its “poison pill” strategy, which gives the company’s board more time to maneuver amid Tesla CEO Elon Musk’s hostile takeover attempt, Bloomberg reported. A new securities filing allows Twitter to offer new stock to all shareholders, except Musk, at a discounted price. Twitter shares rose 5% in midday trading Monday, the first time the markets reopened since news of the “poison pill” tactic broke. In a separate development, former Twitter CEO Jack Dorsey criticized the company’s board Saturday in a rare show of public dissension.
Under video review. Two federal agencies are investigating TikTok’s handling of videos depicting child sexual abuse, adding to the app’s content moderation and legal issues, the Financial Times reported. The departments of Homeland Security and Justice are reviewing the growing use of TikTok to share child pornography, which the app’s Chinese owner, ByteDance has struggled to police. A Homeland Security official accused ByteDance of failing to work as proactively on child sexual abuse issues as domestic social media companies.
A dead end? The CEO of electric vehicle startup Rivian warned of a fast-approaching shortage of battery supplies, a challenge that could threaten the growth of climate-friendly automobiles, the Wall Street Journal reported Monday. Rivian chief RJ Scaringe estimated that the industry’s battery needs will increase roughly 10-fold in the next decade, requiring a major upgrade to supply chains already straining under shortages. Scaringe predicted that the scarcity would impact virtually all links in the supply chain, including raw material sourcing and battery cell assembly.
Harvesting some coin. In the latest cyberattack targeting cryptocurrency-based platforms, a cyberattacker drained $182 million from the Ethereum-based stablecoin protocol Beanstalk Farms, CoinDesk reported Sunday. Beanstalk officials said the hacker took out a flash loan, used the funds to acquire a large stake of Beanstalk’s native governance tokens, then unilaterally passed a governance proposal to place the cryptocurrency in a private wallet. Beanstalk hasn’t said whether it will reimburse those who lost assets, a step taken by Axie Infinity following a $625 million cyberattack last month.
FOOD FOR THOUGHT
One man’s Treasure Data. It’s well-documented that Meta, Twitter, and other social media companies felt the sting of Apple’s iOS privacy changes last year, which made it harder for developers to track users across their smartphone apps. What’s gotten less attention, however, is the opportunity for other companies to benefit from the switch. As Protocol reported Monday, several customer-data platforms are regaining steam after Apple’s new privacy protocols reduced corporations’ reliance on social media platforms to identify and target consumers with ads. The winners include household tech names, such as Adobe and Salesforce, along with lesser-known startups.
From the article:
In the wake of data privacy changes by mobile platforms last year, the enterprise tech world is suddenly very interested in customer-data platforms (CDPs). With Twilio’s acquisition of Segment, Treasure Data’s $234 million fundraise late last year, and Salesforce’s push into CDPs, the hot new buzzword is potentially a hot new market.
“The way I think about [CDPs] is, it’s trying to create a 360-degree view of each of your customers to help you more accurately identify what would most resonate with this customer,” said Derek Zanutto, a general partner at CapitalG.
IN CASE YOU MISSED IT
Congress seeks to pass legislation that addresses the global chip shortage, by Kevin Freking and the Associated Press
Elon Musk’s cryptic tweet channeling Elvis baffles investors, by Michael Tobin and Bloomberg
BEFORE YOU GO
The silly season. College football remains the premier home for lunacy. In the latest example, Sports Illustrated reported that the University of Central Florida replaced players’ jersey numbers with a giant QR code for Saturday’s spring game. The codes sent scanners to each player’s page on the university’s athletics website, where student-athletes could link to social media pages featuring branding opportunities. (College athletes can now profit off their “name, image, and likeness.”) More than anything, the move got writers like me to talk about UCF spring football, which we never would have otherwise talked about. So score one for head coach Gus Malzahn.
This is the web version of Data Sheet, a daily newsletter on the business of tech. Sign up to get it delivered free to your inbox.