Welcome back to The Prompt,
Data analytics giant Palantir is suing the founders of a fledgling startup called Guardian AI for allegedly stealing its trade secrets in order to start an AI company that helps hospitals fight insurance claim denials, Forbes reported. The lawsuit alleges that the company’s cofounders Mayank Jain and Pranav Pillai incorporated their Y Combinator-backed startup days after leaving Palantir. “There is no mistaking Defendants’ work for anything but a rip-off of Palantir,” according to the suit.
Now let’s get into the headlines.
BIG PLAYS
Elon Musk’s artificial intelligence company xAI acquired social media platform X (previously known as Twitter) in an all-stock transaction that valued X at $33 billion, said Musk. Both those companies are owned by Musk, the world’s richest man. xAI was already training its models on the platform’s data. The deal means that X’s models, compute and talent will merge into xAI, which is now valued at $80 billion. X has been struggling with an exodus of users and advertisers after Musk bought the company in 2022, and it faces increasing competition thanks to the emergence of alternative platforms like BlueSky and Threads.
SHOW ME THE MONEY
Cloud compute provider CoreWeave went public on Friday, marking the first big IPO of the year. The company, which rents out access to its GPUs to AI companies like Microsoft, Mistral, Meta and Cohere, ended up pricing its shares at $40, below its target of $47 to $55, earning it a lower valuation than it had aimed for: $23 billion. Its share price closed at $52.27 today. Founded by three commodity traders who bought hundreds of Nvidia chips for bitcoin mining, CoreWeave pivoted to AI in 2019— a prescient decision that has made all its founders billionaires by Forbes estimates.
ENTERPRISE ENDEAVORS
Anthropic is pairing up with Databricks to sell its suite of AI models to that data analytics platform’s thousands of customers that includes Conde Nast, Comcast and Block. The deal, which is worth $100 million, will allow enterprises to build AI agents —systems that can carry out specific tasks in departments like customer support and engineering— using Anthropic’s technology.
AI DEAL OF THE WEEK
OpenAI closed a $40 billion funding round, which is said to be the largest private tech deal in history. But it comes with a catch: $30 billion of that amount will be released by the end of the year, but only if the ChatGPT maker restructures itself into a for profit entity, according to The Wall Street Journal. The investment, which is being led by SoftBank, values OpenAI at $300 billion.
DEEP DIVE
After OpenAI rolled out new image generation capabilities in ChatGPT last week, social media platforms like Instagram, Reddit and X were awash with AI-generated images that mimicked the style of popular Japanese animator Studio Ghibli, whose cofounder Hayao Miyazaki has famously called AI-generated art “an insult to life itself” and has publicly been opposed to the use of artificial intelligence in art.
OpenAI CEO Sam Altman and several other OpenAI executives have created and shared their own “ghiblified” images after the company has relaxed its policies over creating images in distinct artistic styles. These Ghibli-style AI-generated images have quickly gone viral as people have created thousands of portraits of themselves and their families. Others have used ChatGPT to depict movie scenes and historically significant moments like the 9/11 terrorist attacks and John F. Kennedy’s assassination. Even the White House posted a AI Ghibli version of a photo of a crying woman being arrested by an ICE officer.
The virality has drawn more people to ChatGPT, which OpenAI now claims has more than 500 million weekly active users. Altman posted on X that the trend has added 1 million users, resulting in GPU capacity issues. “Our GPUs are melting,” he said.
Many members of the creator community have voiced their concerns with AI companies like OpenAI training their models on copyrighted works including those from Studio Ghibli. “Authors and artists are getting increasingly angry with the large-scale theft that is happening,” Ed Newton Rex, CEO of Fairly Trained said.
But there are also significant privacy risks with uploading personal images and other sensitive data to ChatGPT. Rachel Tobac, a privacy researcher and CEO of SocialProof Security says some people who are using ChatGPT to create “ghiblified” selfies might not be aware that the images they feed into it can be used by OpenAI to train future models, unless they’ve actively opted out of training. There is also a chance that a data breach could expose the images to the public. “If you want to retain ownership of a photo, Ghiblifying it is not the way to go,” she said.
“Some people might think ‘oh, wait a minute. I didn’t realize they were going to be training the algorithm on the picture of my newborn. That kind of makes me uncomfortable.’”
WEEKLY DEMO
AI startup Runway, whose software is popularly used for field editing, has released a new AI system that it claims can generate videos of specific people, places and things that are consistent across different frames— something that AI models have long struggled to do. The underlying models are also better able to understand real world physics and represent them.
MODEL BEHAVIOR
Researchers from Dartmouth College released the results of the first ever clinical trials on generative AI-based therapy chatbot called Therabot. The study, which surveyed about 100 people diagnosed with depression, anxiety or an eating disorder, found that people diagnosed with depression experienced a 51% average reduction in symptoms and overall improvement in well being. That study comes as people increasingly turn to, and converse with, AI chatbots and virtual companion apps.
