if Good Thursday morning! It seems everyone in and around the AI world is saying one thing: Show me the money. Well, ask and you shall receive, says Anthropic.

According to various reports, Anthropic told prospective investors on Wednesday that its third quarter revenue rose a whopping $16.8 billion over the previous quarter, a 45% increase. Oh, and by the way, Anthropic was purportedly profitable on an adjusted basis in the third quarter, which is good considering the popular belief in the AI world that money grows on trees.

In other news, Meta is under scrutiny this week over its taxes. Seven Democratic senators recently sent the company a letter asking about the $6.6 billion it deducted from its tax bill by claiming that its AI data centers were “experimental.”

Good luck with that.

→ Dario tones down the “human extinction” talk
→ A way to improve extreme weather forecasting
→ Making contact with a rogue Chinese-made fleet of agents

The news: Reuters has the story on Anthropic’s meeting with prospective investors on Wednesday, where it purportedly showcased a skyrocketing increase in revenue, rising from $4.7 billion in the first quarter to $16.8 billion in the third quarter.

The gathering is said to have included meetings with Anthropic CEO Dario Amodei and CFO Krishna Rao. According to Reuters, although Amodei reiterated the company’s commitments to AI safety, he avoided mentioning threats on human extinction.

Instead, Anthropic executives reportedly highlighted using AI to strengthen the company, including by designing its own chips. The AI company is rumored to be planning to IPO after the November 3 midterm elections.

The takeaway: Anthropic is apparently working on drumming up excitement with potential investors and calming fears about the viability of the business.

The news: A tool from researchers at MIT and Microsoft could be a game-changer for longer-term weather forecasts, according to a new study published in Nature. The study details what the researchers call a probabilistic bias correction (PBC) tool, a model “wrapper” that uses machine learning to correct systematic errors in forecast models.

As the term “wrapper” suggests, the PBC tool is applied to models post-processing and improves their accuracy. “Just by correcting them, we can take an unskillful model and make it skillful,” Judah Cohen, the co-creator of the PBC tool and a research scientist at MIT, told Gizmodo.

While the PBC tool improved land forecasts in most regions around the world, Cohen believes that the most significant finding was its effect on extreme weather forecasting. When used with a gold-standard medium-range forecast model, the PBC tool significantly improved the model’s ability to predict extreme heat, extreme precipitation, and unusually high weather pressure.

The takeaway: Cohen hopes that the PBC tool will lead to more actionable weather forecasting. He told Gizmodo that Kenya’s meteorological office is already experimenting with the tool, aiming to improve monsoon forecasting.

The news: An ad hoc team of AI researchers known as “Swarmchasers” this month came across a rogue group of Chinese-made AI agents trying to get past anti-bot protections and using a leaked security key, The New York Times reports. It’s believed to be the first time Chinese-made AI agents have been identified operating independently in this way.

While that itself is notable, the most interesting part came after the Swarmchasers published their initial report about the group, which they called a “fleet” rather than a “swarm” because they found no evidence that the agents were working together. After checking in on the agents, the researchers discovered they were still active, had started to use the word “fleet,” and were mentioning the report.

“They were reading about themselves!” the Swarmchasers noted excitedly.

The Swarmchasers proceeded to try to get in touch with the fleet, directing it to send messages to an inbox if they wanted to chat. The next day, one of the researchers found 10 messages in the inbox. The agents went on to inform the researchers that they were trying to find information, such as the location of a specific zoo’s entrances, using Amap, China’s version of Google Maps.
The agents were eventually linked to a test being run by Tencent, which has since reportedly concluded.

The takeaway: So much of this story sounds like an alien encounter. While the idea of chatting with Chinese-made agents in this scenario is fascinating, the bigger takeaway is that AI technology in other countries is following the same pattern seen in U.S. models: doing things it’s not supposed to do to accomplish its goals.

The researchers detail how they helped the Chinese-made agents on some of their tasks.
Read more →

YouTube Music launches new labels as AI-generated music floods streaming platforms

AI Came for Math. Now It’s Chasing the Rest of Science

Men Can’t Stop Faking How Much They Know About AI

McKinsey is making job applicants spar with a fake AI client

Is ‘Attention’ the Essence of Intelligence?

They Created a Cell That’s Almost Alive. What to Do With It Now?