Ilya Sutskever isn’t done working on AI safety

Welcome back to TechCrunch’s Week in Review — TechCrunch’s newsletter recapping the week’s biggest news. Want it in your inbox every Saturday? Sign up here.

This week, Ilya Sutskever launched a new AI company, Safe Superintelligence Inc. (SSI), just one month after formally leaving OpenAI. Sutskever, alongside Jan Leike, was integral to OpenAI's efforts to improve AI safety with the rise of “superintelligent” AI systems. Yet both Sutskever and Leike left the company after a dramatic falling-out with leadership over how to approach AI safety.

In EV news, Fisker filed for Chapter 11 bankruptcy protection, capping off months of problems with its Ocean SUV that included recalls and dozens of lemon law lawsuits. This is the second vehicle company Henrik Fisker named after himself that has wound up in bankruptcy. His first effort started in 2007 and filed for bankruptcy protection in 2013.

Change Healthcare this week confirmed that a February ransomware attack resulted in the theft of medical records affecting a “substantial proportion of people in America.” The company processes patient insurance and billing for thousands of hospitals, pharmacies and medical practices and has access to massive amounts of health information on about a third of all Americans.

News

The Department of Justice vs. Adobe: The U.S. Department of Justice filed a lawsuit against Adobe alleging that the company hides termination fees and makes it difficult to cancel subscriptions. Read more

OpenAI acquires Rockset: OpenAI announced that it has acquired Rockset, which builds tools to drive real-time search and data analytics, as the company continues to invest in its enterprise sales and tech orgs. Read more

Buttons are back: Clicks released a nostalgic, BlackBerry-esque phone case that adds a keyboard with physical buttons to the bottom of your iPhone. We got to try one for ourselves. Read more

Where humans and AI coexist: Butterflies is a social network where humans and AI interact with each other through posts, comments and DMs in an effort to have more creative relationships with AI. Read more

Apple kills Pay Later: After launching in late March 2023, Apple’s Pay Later feature is no more. Instead, Apple Pay users will be able to access loans through a partnership with the third-party app Affirm. Read more

Beware, Outlook users: A researcher has found a bug that allows anyone to impersonate Microsoft corporate email accounts, making phishing attempts look credible and more likely to trick their targets. Read more

Perplexity takes on Google: The AI-powered search startup now displays results for factual queries such as weather and time at a place, currency conversion, and answers to simple math queries directly through cards. Read more

Runway unveils Gen-3: The company’s latest AI model for generating videos delivers a “major” improvement in speed — as well as more control over the structure, style and motion of the generated videos. Read more

Analysis

What should AI look like?: From black holes to colorful blobs, representing AI in user interfaces can be a challenge. While approaches differ to branding purportedly all-seeing, all-knowing, all-doing intelligence, Devin Coldewey explores how companies have coalesced around the idea that the avatar of AI should be non-threatening, abstract, but relatively simple and non-anthropomorphic. Read more

Why Fisker failed: As Fisker files for Chapter 11 bankruptcy protection, many are left wondering what’s next for the ill-fated EV startup. Sean O’Kane argues that whatever happens to Fisker or its assets, it won’t change the fundamental problem: that it wasn’t ready to grapple with bringing a flawed car to market. Read more

Pushing ChatGPT’s cultural limits: The current ChatGPT offers answers that are too generalized for specific questions that cater to certain communities, as its training appears Eurocentric and Western in its bias. With most AI models not built with people of color in mind, Dominic-Madori Davis and Tage Kene-Okafor report on the Black-owned chatbots and ChatGPT versions that cater specifically to Black and brown communities — and help founders capitalize on OpenAI’s cultural slip. Read more