The Prompt is a weekly rundown of AI’s buzziest startups, biggest breakthroughs, and business deals. To get it in your inbox, subscribe here.
Welcome back to The Prompt.
With the arrival of AI, some businesses have been overwhelmed with applications from suspected North Korean operatives.
Getty
AI tools are helping North Koreans covertly apply for thousands of remote IT jobs in the US, Forbes reported. Companies large and small are being flooded with job applications from thousands of suspected North Korean operatives, who earn hundreds of millions of dollars and send the money back to the regime, where the U.S. government believes it’s used to fund its weapons of mass destruction program. With the help of AI tools, these workers are able to run multiple job profiles and apply for hundreds of jobs at once.
Now let’s get into the headlines.
REGULATION
This week, California legislators will vote on SB1047, a controversial bill that seeks to regulate the most advanced and powerful AI models. If passed, the bill would require the developers of AI models whose training either cost more than $100 million, or required a specified amount of computing power, to implement safeguards and allow third-party audits of safety practices.
It also requires AI companies to outline methods for shutting down the AI model and effectively implement a “kill switch” for the technology if needed. The legislation would allow the state attorney general to take action against a developer if its AI model causes severe harm such as mass casualties or more than $500 million in damages.
Silicon Valley leaders are deeply divided on their positions regarding the bill: xAI and Tesla founder Elon Musk and Anthropic CEO Dario Amodei have come out in support of the bill, while leaders from OpenAI, Meta and Google have voiced concerns that the bill would stifle innovation.
TALENT RESHUFFLE
Three of the five cofounders of French AI startup H have left the company after “operational and business disagreements,” according to The Information. The departure comes just a few months after the startup raised a whopping $220 million seed round from billionaires like Eric Schmidt and Bernard Arnault to build AI agents for multi-step tasks.
AI DEAL OF THE WEEK
Coding automation startup Cursor AI raised $60 million in Series A funding at a $400 million valuation, CEO Michael Truell told Forbes. The company’s AI tools are popular among developers at leading AI startups like OpenAI and Midjourney, where they are used to write, edit and predict parts of code. But Cursor isn’t short on competition— the market is flooded with similar AI coding assistants like Codeium, which launched an engine capable of digesting 100 million lines of code, and Cognition Labs, which is valued at $2 billion and created an AI software engineer called Devin. Tech giants are also developing their own AI programming tools in-house; Amazon CEO Andy Jassy said that its AI assistant, called Q has helped save the company $260 million and 4,500 years worth of time in terms of software development.
DEEP DIVE
The idea of America is big business on Facebook. The social network has hosted more than a hundred pages that have adopted American patriotism as a theme, boasting names like Proud American, Proud To Be An American, American Story, and We Are America.
But a large swath of those pages — despite their names — aren’t American at all. Instead, they’re run by foreign click farmers, many of whom are based in Macedonia, who use AI to pump out a near-endless ocean of clickbaity soup. Posts sharing prayers for American soldiers, rewritten tweets, memes and pictures of old Hollywood pin-up girls link out to AI-generated articles, against which the click farmers can sell advertising.
Headlines like “A Father’s Heroism: The Tragic Story of Phil Dellegrazie And His Son Anthony” tease short, uninformative articles on websites plastered with often sexual advertisements. The pages promoting them fake Americanness because they get paid every time someone clicks on one of their links, and in the advertising world, American clicks are some of the most valuable.
A Forbes review identified 67 Facebook pages — now taken down — that identified themselves as champions of American news, culture or identity, but were actually based overseas. As of August 20, they had more than 9 million followers combined — more than the Facebook pages of the Wall Street Journal or the Washington Post. Thirty-three of them were run from Macedonia, with others spread out across 23 different countries, including Canada, France, Morocco, Venezuela and Vietnam.
Click farmers, especially those from Macedonia, have a long history on Facebook. During the 2016 presidential election, teenagers in the small Eastern European country pushed fake news to millions of Americans on Facebook, making tens of thousands of dollars in ad revenue. In 2019, similar Eastern European pages ran the same playbook — this time, reaching nearly half of all Americans on the platform.
Now, AI has given those same operations the capacity to produce near-infinite volumes of low-quality (or outright fake) news — and in at least some cases, this AI-produced slop is breaking through. The pages have begun using generic AI-generated imagery (bald eagles, stars and stripes, camo soldiers and the occasional Statue of Liberty) to appeal to American Facebook users — and in at least some cases, it’s working. One post made last week by the Canada-based page American Patriots featured an AI-generated photo of an American soldier and his children, and received more than 100,000 likes and 35,000 comments. The American Patriots page, like most of the others, directed people from Facebook to click farms featuring low-quality articles.
Read the full story on Forbes.
WEEKLY DEMO
Do you want to practice a tough workplace conversation or get tips on how to negotiate a raise? Companies are increasingly deploying AI-powered career coaches as an alternative to expensive human counselors that can cost up to $240 an hour, Forbes reported. But people who have interacted with these AI-based career counselors note that these chatbots often lack nuance and can sometimes offer confusing advice. “I’m already confused about my career. AI [only] throws me in a bigger loop,” one third-year law student said.
AI INDEX
Two years ago, the Biden administration passed the CHIPS Act to incentivize the development of semiconductors and chips within the United States, as the country battled with China on developing AI models. But red tape and a grueling application process has largely kept funds out of reach from smaller firms that need it most, Forbes reported.
Less than 7%
Applicants that received funding from the 380 firms that submitted applications.
9 out of 23
Semiconductor manufacturers who were approved for the funding were smaller companies.
$4 billion out of $134 billion
Amount of grants and loans awarded to smaller companies; the rest went to chip giants like Intel, TSMC and Samsung.
MODEL BEHAVIOR
American rapper and singer Will.i.am is launching an AI-powered radio station called Raidio.FYI, which will allow listeners to listen to songs and news and ask questions to the host through a chatbot app built on OpenAI’s large language models, according to The Sunday Times. The rapper is reportedly an investor in OpenAI and Anthropic.