Techcrunch blog!: Your seafood protein alternative is now ready

Friday, February 16, 2024

Good Morning! It's a bright Friday morning, and what better way to end the week than with exciting developments in AI? OpenAI's latest model is apparently even more capable than folks initially thought — a paper from the company's researchers reveals its video generation model can simulate entire digital worlds, à la video game engines!

We're also excited about what seems to be crypto's emergence from a lengthy downturn, and are keeping an eye on key regulatory updates from the U.S. involving deepfakes and digital rights. We also have a startup that is using AI to help protect key infra, and another that is pursuing a novel way to create alternative animal proteins. To work!

– Alex

TechCrunch Top 3

OpenAI's Sora could remake video production: OpenAI's new video-generating model took X by storm yesterday. We already knew it could make cinematic 60-second clips from simple prompts, but a new paper from OpenAI indicates that it is also able to simulate digital worlds — a prompt led it to render Minecraft, including its UI, graphics, and physics. Does this mean I'll be able to create a version of Starcraft 2 that I am actually good at?
India could ban ProtonMail: After its service was used in a domestic bomb threat, privacy-focused Swiss software company Proton has been informed that it may be blocked in India. The company called the potential block an "ineffective and inappropriate response" to the situation. Proton makes email, storage, and VPN services with a focus on privacy.
Crypto leaves winter behind: Following quarterly results that trounced expectations and a rosy revenue forecast, Coinbase shares are trading up about 14% pre-market today. The U.S.-based crypto company named stablecoins, its Base blockchain, and regulatory work as key corporate projects for 2024.

Image Credits: OpenAI

Don't miss these

Putting Apple's Vision Pro through its paces: Brian Heater's series on his regular use of Apple's new face computer continues. This time, he's digging into an art form that he particularly loves: comics. The verdict? The headset is "fine" for reading comics, but it doesn't do anything revolutionary. (Bonus: If you are experiencing motion sickness from using a VR/MR headset, head here for tips on how to mitigate the issue.)

Meta's EU privacy gambit criticized: After years of litigation, Meta decided late last year to offer its users a choice in Europe: either pay for its service or consent to its advertising system. Now, more than two dozen digital and democratic rights groups, NGOs and not-for-profits, including privacy advocacy group noyb and Wikimedia Europe, are arguing that the EU should reject the setup.

India delays Paytm Payments Bank's demise: Indian fintech giant Paytm's banking setup, which the company uses to facilitate its payments tech, is on its last legs. A regulatory decision gives the bank an extra 15 days before it is "barred from accepting deposits and facilitating credit transactions," TechCrunch reports. Still, there's not much hope of any larger concessions by the domestic government.

OpenAI won't get the "GPT" trademark: U.S.-based AI giant OpenAI has been denied a local trademark for the term "GPT," which stands for generative pre-trained transformer. Why not? The U.S. Patent and Trademark Office said that the application was denied "because the applied-for mark merely describes a feature, function, or characteristic of applicant's goods and services." When you go back and reread what GPT stands for, that does make sense. Still, I doubt that OpenAI agrees with the ruling.

Anthropic tests misinformation-limiting tech ahead of elections: This year will see a fair few number of elections of note around the world, and how AI technology is used or abused during these democratic events will help set the tone for future AI regulation. It seems Anthropic wants to get started on mitigating the negatives before it gets told to. The well-funded startup is testing something it calls "Prompt Shield," which detects political questions and then sends those users to trusted information sources. If nothing else, the feature should ensure that if Anthropic is later dragged with its AI peers before Congress, it will have a good-faith security effort to point to.

U.S. government may modify deepfake rules to protect individuals: Speaking of AI regulation, the U.S. Federal Trade Commission may modify a rule that "bans the impersonation of businesses or government agencies to cover all consumers," Kyle Wiggers reports. The government could go as far as making it illegal for services to offer capabilities that they know are being used to harm folks through impersonation. This sounds like a good-faith effort, but I wonder if it will hinder people from making harmless, fun bits of digital art if a blanket ban is used to combat a more tailored risk.

Changes to kid-focused digital law in the U.S. fail to quell criticism: The Kids Online Safety Act (KOSA) has a real shot at becoming law in the United States, and privacy watchdogs are still not happy with it. Concerns raised by LGBTQIA+ groups have been partially ameliorated, but the EFF calls KOSA a "dangerous and unconstitutional censorship bill which we continue to oppose" despite recent changes to it. (This aged joke remains undefeated.)

Neara wants to protect critical utilities from severe weather: If you've ever lost power after a storm, you know how difficult it can get. Your food can go bad! Wi-Fi becomes a ghost! Neara is building AI-imbued tech to help utility companies and energy providers model what might happen to their power networks during nasty weather. This allows utilities to assess risk and plan ahead without needing to wait for a storm. Given that the world is heating quickly, which is expected to trigger more extreme weather events, Neara could sit at the meeting point of climate tech, AI, and climate change.

Everyone wants to build a satellite network: Starlink stole a march on the world by building a network of Internet-providing satellites that bring connectivity to far-flung reaches of the Earth, and Amazon is building a competitor that expects to begin commercial service this year. Rocket Lab wants in on the game, Aria Alamalhodaei reports, and is considering building its own satellite network that would help it create a steadier revenue stream than rocket launches.

Image Credits: Brian Heater

Before you go

Your fermented fish alternative is now ready: Meat and fish are tasty, but bringing them to your plate is hard on the planet. And that's even harder on the seas, where us humans have fished with such indiscriminate gusto that plummeting sea animal stock is more than an idle worry today. That's why fake fish is not such a bad idea. Pacifico Biolabs has a plan to create "muscle structures using fermentation" to limit our species' need to drop a net in the ocean. I will continue to despise all seafood, but this sounds pretty planet-friendly.