X will deploy AI to write community notes, expand fact-checking


Humans will still review those AI-generated notes, and the note will only appear if people with a wide variety of viewpoints find them useful, said Keith Coleman, a product executive at X who runs the Community Notes program. — Bloomberg

Elon Musk’s X will start to publish Community Notes written by artificial intelligence agents, a move to increase the speed of the social network’s fact-checking product and expand it to reach more people. 

Developers will soon be able to submit their own AI agents for review by the company. Those AI agents will write a series of practice notes behind the scenes, and if the company deems them helpful, the bot will be deployed to write notes that will appear publicly on the service. 

Humans will still review those AI-generated notes, and the note will only appear if people with a wide variety of viewpoints find them useful, said Keith Coleman, a product executive at X who runs the Community Notes program. That’s the system in place for notes written by X users. AI notes could start to appear later this month. 

"They can help deliver a lot more notes faster with less work, but ultimately the decision on what’s helpful enough to show still comes down to humans,” Coleman said Tuesday in an interview. "So we think that combination is incredibly powerful.”

There are currently hundreds of notes published to X every day, Coleman said, and while he doesn’t have a target for how that might change once AI is involved, there could be a "significant” increase. 

X first debuted a crowd-sourced fact-checking program when the company was still known as Twitter, and well before Musk’s 2022 takeover. But its focus on Community Notes has increased under Musk’s ownership, and has recently been adopted by other companies, including Meta Platforms Inc and ByteDance Ltd’s TikTok. 

Musk himself has said Community Notes serves as a bulwark against false information, calling it hoax kryptonite, but he is also regularly flagged by the fact-checking process for posting misleading information. Earlier this year he also suggested the system could be "gamed by governments & legacy media.”

Coleman said that other companies adoption of community notes is evidence that it’s the best fact-checking system available. He’s also hopeful that asking humans to review the AI notes before publishing them will create a "feedback loop” that will improve the bots as well.

"It’s a new feedback cycle,” he said. "The model can be improved not just by one random human’s feedback, but by feedback from a diverse audience.” 

AI agents can be powered by any technology, Coleman said, not just Grok, the bot that was created by Musk’s AI startup xAI. – Bloomberg

 

 

 

Follow us on our official WhatsApp channel for breaking news alerts and key updates!

Next In Tech News

Anthropic buys Super Bowl ads to slap OpenAI for selling ads in ChatGPT
Chatbot Chucky: Parents told to keep kids away from talking AI dolls
South Korean crypto firm accidentally sends $44 billion in bitcoins to users
Opinion: Chinese AI videos used to look fake. Now they look like money
Anthropic mocks ChatGPT ads in Super Bowl spot, vows Claude will stay ad-free
Tesla 2.0: What customers think of Model S demise, Optimus robot rise
Vista Equity Partners and Intel to lead investment in AI chip startup SambaNova, sources say
Apple plans to allow external voice-controlled AI chatbots in CarPlay, Bloomberg News reports
Goldman Sachs teams up with Anthropic to automate banking tasks with AI agents, CNBC reports
US Justice Department casts wide net on Netflix's business practices in merger probe, WSJ reports

Others Also Read