Technology

➚ MOVE FAST: Open source. AI pioneer Kai-Fu Lee’s new Chinese startup, 01.AI, has built an open-source large language model that it claims outperforms Meta’s Llama 2. Lee said he wants to build 01.AI into a more transparent version of OpenAI. “It’s not US vs. China; it’s open vs. closed,” he wrote in a post on X.

➘ BREAK THINGS: Closed source. U.S. export controls are increasingly cutting off Nvidia from its Chinese customers, putting the semiconductor giant in a bind. One of the latest signs of trouble: Ernie chatbot maker Baidu decided to buy more of its chips from Huawei instead, according to Reuters.

London-based AI startup Luminance claims its proprietary large language model was able to negotiate both sides of a legal contract without any human involvement. “This is just AI negotiating with AI, right from opening a contract in Word all the way through to negotiating terms and then sending it to DocuSign,” a company executive told CNBC in an interview.

Investors are pouring millions into companies betting AI will help reduce the countless hours of paperwork lawyers need to sort through or draw up, Semafor reported earlier this year. That goal proved harder than earlier startups anticipated, but the advent of generative AI appears to be bringing it closer to reality.

GitHub’s AI coding assistant is a moneymaker

THE SCOOP

GitHub’s AI coding assistant, called Copilot, has gone from an experiment to a moneymaker, the company’s CEO Thomas Dohmke said in an exclusive interview with Semafor.

“We are happy about the growing positive margins the product has,” he said.

The revelation helps answer an important question about the business of generative AI: Because the powerful models require so much costly compute power to run, can companies make a profit on products that rely heavily on the nascent technology?

Dohmke’s answer was an unequivocal yes. “The cost of goods sold, which is all the costs to run Copilot per user, is lower than the price that we’re charging,” he said.

Copilot was the first consumer product to use AI models from OpenAI, the company behind ChatGPT. It helps software developers write code, the same way Gmail suggests the next word in a sentence while typing an email.

GitHub initially launched Copilot as a free, experimental product in 2021. Dohmke didn’t say when Copilot began posting positive margins, but he implied the company has been building infrastructure in anticipation of demand.

That is a capital-intensive investment, especially considering that AI models like OpenAI’s require graphics processors, which are not found in traditional data centers.

On Wednesday, GitHub launched several new products that significantly increase Copilot’s role in the company’s offerings. In December, it will make Copilot Chat generally available, allowing developers to treat it more like a colleague, moving it even further beyond its autocomplete roots.

Number of pornographic deepfake images uploaded this year to the top 10 websites for sharing such content, according to data compiled by industry analyst Genevieve Oh and shared with The Washington Post. Oh also found that 143,000 AI-generated videos had been uploaded in 2023 to 40 different sites, more than the previous seven years combined.

Since ChatGPT launched, a giant question has loomed over OpenAI: Is it a one-hit wonder, or a real business that could stay ahead in an intensely competitive AI land grab?

After all, it began as a nonprofit that turned into a startup whose CEO owns no equity in the company. On stage at OpenAI’s first developer conference on Monday, Sam Altman showed that even he’s in on the joke, calling ChatGPT a “low-key research preview.”

The event was not so much about the company’s technology, but showcasing its ability to quickly switch gears, from pure research to a blitzscaling tech startup trying to take over the world. OpenAI went from one of many firms that allowed businesses to tap into large language models to building a major ecosystem aimed at meeting the AI needs of pretty much every tech-enabled company.

On stage, Altman said 92% of the Fortune 500 was using OpenAI products. After the event, I asked him how much those companies were using it. He said the metric OpenAI cares about is “tokens generated per day,” but the figure was so high that “it sounds like a made up number.”

And Altman clearly expects it to go up. He said with each iteration of GPT, companies have found more ways to use it. “By the time we get to GPT-5, we expect it to work pretty well for most things we might want to build,” he said.

• Meta announced it will require political advertisers to disclose AI-generated images and videos starting next year. The new rules will impact the 2024 U.S. presidential election, which has already involved several deepfakes, including of Republican frontrunner Donald Trump.
• TikTok’s Creator Fund will stop operating next month in the U.S. and several European countries. The program promised to distribute $1 billion to creators over three years and was criticized by major stars for its low payouts. Its replacement, the Creativity Program, lets creators monetize videos that are at least 1 minute long.
• YouTube is testing two generative AI tools that will answer questions about a video and summarize what people said in the comment section. For some academic videos, users can also create quizzes to test their understanding of the lessons featured in them.
• Ghosters, rejoice! Instagram will finally allow users to turn off read receipts in direct messages. Mark Zuckerberg and Instagram head Adam Mosseri said the change is currently in testing; it’s not clear when it will roll out or if it will also be added to Facebook messenger.

• TikTok is telling advertisers that #Israel is trending, but not #Palestine.
• A new Democratic strategy group is out with its first polling — a warning that Biden’s focus on “jobs” is getting him nowhere.
• A company whose floating power ships switched off electricity to two of Africa’s poorest cities says it’s in talks with six more countries to expand across the continent.