One reason the U.S. government is so spooked by TikTok: AI

Welcome to AI Decoded, Fast Company’s weekly newsletter that breaks down the most important news in the world of AI. You can sign up to receive this newsletter every week here.

The real reason the U.S. government wants to ban a ByteDance-owned TikTok

TikTok could shut down in the U.S. as soon as January 19 if the Supreme Court doesn’t step in. The high court is now considering the constitutionality of a law passed by Congress in April mandating that TikTok find a new owner or face a ban. The government believes TikTok’s parent company, Beijing-based ByteDance, might share user data with the Chinese government. But Congress has failed to explain how such a data transfer could be so dangerous to the 170 million Americans who use TikTok.

The data threat has a lot to do with AI. Congress is trying to play the “long game” with TikTok, one Washington insider tells me. China has a history of vacuuming up any data it can get its hands on, from user data to supply chain analytics to intellectual property. Per this insider, lawmakers worry that a decade from now Chinese AI may be able to access and weaponize TikTok user data to move U.S. public opinion one way or the other on key issues like tariffs. 

TikTok’s algorithm is already eerily good at putting videos in front of users that match their tastes, values, and sensibilities. That capability will only improve as AI models grow more sophisticated. Because AI makes generating content cheaper, the Chinese government might be able to leverage what it knows about each TikTok user, and generate content for them that’s specially tailored to persuade them. 

Congressional offices have lately seen a tidal wave of calls and letters from TikTok users urging them to stop the ban. This has caused some Democratic lawmakers to change their position on the issue, according to the D.C. insider. But, ironically, it also served as a demonstration of the power of the platform to exert political force on lawmakers, which could have hardened the pro-ban positions of some lawmakers.

ByteDance says it is 60% owned by international investors, 20% owned by the company’s founders and Chinese investors, and 20% owned by employees. The Chinese government owns a “golden share” of one of the company’s Chinese subsidiaries, which is said to amount to 1% of the company. But in China, the law permits the government wide access to, and influence over, private companies. For U.S. lawmakers, anything short of a total divestiture by ByteDance leaves the door open for trouble. 

OpenAI, Microsoft take first steps into offering AI agents 

Generating and summarizing text is yesterday’s news. The latest hype is around AI agents that can work independently on our behalf. And OpenAI and Microsoft are giving AI optimists plenty to get excited about.

OpenAI launched a new beta feature this week called “tasks” that essentially lets users ask ChatGPT to perform actions on their behalf in the future. One might ask the chatbot for a daily news or weather update at a certain time or for a reminder to buy concert tickets when they go on sale. “Tasks” may not sound very ambitious, but it’s very likely a first step toward a more independent and proactive—that is, more agentic—AI experience. The new feature is rolling out to ChatGPT Plus, Team, and Pro users over the next several days.

Microsoft has already added its “Copilot” AI assistant features across many apps within its Microsoft 365 suite for business users. Some enterprises pay $30 per user per month for that service, which links the AI with worker files and company data (i.e. the “graph”). Some Microsoft 365 shops don’t want to pay for the copilots, but do want to start using AI agents. So Microsoft created a “Microsoft 365 Copilot Chat” plan for those customers, which provides a free (albeit non-graph-connected) version of chat and the ability to use (pre-baked or custom-built) agents on a metered basis. It’s clear Microsoft is convinced that agents will be a key part of worklife going forward, and it wants to get everybody started now. 

This short film might make you into an AI “doomer” 

For many months, I’ve followed the great debate over the long-term existential risk of AI between the AI accelerationists (Marc Andreessen and Yann LeCun) and AI doomers (Eliezer Yudkowsky and Geoffrey Hinton). The accelerationists say that tech companies will have ample time to train and restrain AI models from working against human interests. The doomers say we may only have one chance to constrain artificial superintelligence systems before they learn to pursue their own goals despite human interventions. I used to think that the accelerationist side sounded a bit more grounded and pragmatic, but after watching the recently released short film &t=1397s">Writing Doom, I’m not so sure.

Written and directed by Suzy Shepherd, Writing Doom portrays a team of TV writers charged with creating a new season of a show in which artificial superintelligence plays the “bad guy.” The film, which won the Grand Prize in the Future of Life Institute’s Superintelligence Imagined Contest, both enlightened and unnerved me. It deserves a wider audience.

More AI coverage from Fast Company: 

• Hinge will now use AI to grade your dating profile prompts
• AI slop is still distorting news about the L.A. wildfires
• Joe Biden’s new chip export rules strangle competition to ensure U.S. AI supremacy
• Is Art Basel’s AI embrace a turning point for the art world?

Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.

https://www.fastcompany.com/91261513/us-government-tiktok-ban-ai?partner=rss&utm_source=rss&utm_medium=feed&utm_campaign=rss+fastcompany&utm_content=rss