Continuing its streak of the biggest weekly hits on AI Tangle is Google, its latest hit looking to usurp OpenAI's Sora spot as the most advanced AI video generation model (that we know of) with Veo 2. Other key highlights include:
- xAI upgrade its flagship Grok-2 model with an update, brings the "Grok button" to X/Twitter
- OpenAI researcher-turned-whistleblower found deceased in his San Francisco apartment
- Tom Hanks' latest film "Here" receives stark criticism from Lisa Kudrow for endorsing AI in the film industry
Join us at AI Tangle as we untangle this week's happenings in AI!
|
It seems Google is not allowing OpenAI or its other competitors to enjoy the holiday spotlight whatsoever, as its DeepMind research department debuted Veo 2, an upgraded GenAI video model that is capable of, allegedly, creating 4K resolution videos over two minutes long. While still limited to 720p and eight-second clips in Google's VideoFX, its theoretical capabilities far outweigh OpenAI's Sora. On top of that, Google has also upgraded Imagen 3, enhancing its ability to generate detailed, stylistically diverse images.
What else about Veo 2 is better now?
Veo 2 builds upon its predecessor's video-generation abilities with upgrades in realistic fluid dynamics and light rendering while adding sharper textures and improved "camera movements," though things like "character consistency" remain an issue, as DeepMind has admitted to itself. Meanwhile, Imagen 3 upgrades include richer detail, faithful prompt adherence, and improved stylistic outputs, now accessible on ImageFX. Both systems are using DeepMind's SynthID tool for watermarking purposes to address misuse concerns as AI video and image generation evolve. For now, however, access remains limited to VideoFX, which is behind a waitlist, though a broader rollout is on its way.
|
As announced in its blog post, xAI has begun rolling out an upgraded version of its Grok 2 chatbot, described by the startup as "three times faster" with improved multilingual and instruction-following capabilities. Free users are limited to 10 questions every two hours, while Premium subscribers get higher limits. Additionally, a new "Grok button" on X/Twitter provides real-time event context and trending insights. Other improvements include updates to the API, such as better efficiency, reduced pricing, and the upcoming addition of the Aurora image generator, which has already debuted within Grok, even if it did not last for a while.
After having worked at OpenAI for four years, former OpenAI researcher who turned into one of its loud critics, Suchir Balaji, aged 26, was found dead in his San Francisco apartment on November 26, with authorities ruling his death as a suicide. As late as October, Balaji was interviewed by the New York Times, where he publicly criticized OpenAI's data practices and its alleged violations of copyright law, having left the company in August to work on personal projects. Balaji's whistleblowing came at a time when OpenAI faced multiple lawsuits from publishers and authors, putting it under a spotlight over concerns about its data usage in training AI models.
American actress Lisa Kudrow recently publicly criticized Tom Hanks’ new film "Here", calling it "an endorsement for AI" during a discussion on the Armchair Expert podcast. Directed by Robert Zemeckis, the film uses Metaphysic Live, an AI-driven tool, to de-age and face-swap actors in real-time, which Kudrow raised concerns about regarding AI's impact on acting careers and broader societal implications, questioning future work for humans. While Metaphysic's technology has been widely praised for its realism, AI remains an omnipresent issue in the industry, with concerns about the exploitation of AI tech fueling recent Hollywood strikes.
Meta has pushed an update to its Ray-Ban Meta smart glasses that includes a bunch of new AI features, including real-time video processing, and live speech translation between English and Spanish, French, or Italian. As part of the v11 update, the update will allow users to receive real-time language translations via open-ear speakers or phone transcripts. Though only available to the US and Canada at first, the Ray-Ban Meta smart glasses will also be able to now identify songs via Shazam. The update is currently available for members part of Meta's Early Access Program.
YouTube recently introduced an optional setting for creators to allow third-party companies, such as OpenAI, Meta, Microsoft, Adobe, and many more, to train their AI models using their videos. The default setting is off, but creators can opt in via YouTube Studio and even pick and choose which third-party companies they want to give the green light to. YouTube emphasizes this is to support creators in the AI era and unauthorized scraping remains prohibited. Google itself already uses YouTube data for AI tools under its existing terms of service.
|
Durable - Durable makes setting up online businesses quick and simple with its AI-based no-code platform together with SEO, marketing tools, and review automation.
Hoop - Made for the busy professional, Hoop is an AI task management platform that captures and prioritizes your tasks with a global task list across your teams.
Manaflow - Turn long, repetitive spreadsheet and software tasks into automated, natural end-to-end workflows with Manaflow - build your dream AI operations team.
Collato - Collato is an AI meeting assistant and notetaker that transforms meetings into easy-to-read documents with ease, allowing users to focus on the meeting, rather than jotting down notes.
|
Smartphones & AI - Does The Average Person Even Want it? (2-min read)
Though AI has seen a massive uprise in enterprise and the business world as a whole, a recent survey regarding AI features in iPhones and Samsung devices has found that the overwhelming majority of users sees little to no value in them, which begs the question - has the best simply yet to come, or is there more to it?
|
| What did you think of this newsletter? Let us know! |
|
|
|
|
|