Lab Weekly — 06/21/2024

Social Media vs. Parental Backlash & AI Disruptions, Plus, the latest news on video-generating AI, Netflix House, new Antropic model, and some cool stats

Want this newsletter in your inbox every week? Sign up for our mailing list here

Social Media At a Crossroads of Cultural Backlash & Generative AI

Social media has been in flux for a while, with major social platforms from Twitter to TikTok were confronting existential-level threats while bracing for the impact that generative AI promised to unleash. This year, however, social media companies have a new formidable challenge to overcome: convincing parents that they are not bad for kids. Could generative AI be the savior or its downfall? Let’s take a closer look at social media at a crossroads.

In case you missed it…

Apple Intelligence All The Way Down

In this recap of Apple’s WWDC announcements, let’s dive into Apple’s “safe yet impressive” approach to AI integration, the headline-grabbing partnership with OpenAI, software updates for Vision Pro, and more must-know tidbits for brand marketers.

The Backlash to Google’s AI Search, Explained

Why Google Search telling people to eat rocks is just the tip of the iceberg that’s about to hit online media ecosystem

Outlook 2024: APAC POV

With the era of passive growth over, here’s how brands can navigate the downstream effects and ensure sustained growth in APAC. Written by Sharon Soh, Chief Planning & Audience Officer, UM APAC

Episode 158: Apple Intelligence & Other WWDC News

As usual, Apple’s latest WWDC event brought forth many software updates, but the star of the show was Apple’s major AI update — Apple Intelligence, which will enable LLM-powered new features across Apple’s entire ecosystem. In this episode, Adam and Ryan took a deep dive into Apple Intelligence, the ChatGPT integration, its impact on the AI competitive landscape, and more.

If you enjoyed the episode, please consider giving us a five-star review on Apple Podcast. Follow the Lab on LinkedIn and on Medium for our latest insights. Thanks for listening!

Runway Unveils New Hyper Realistic AI Video Model Gen-3 Alpha [VentureBeat]

The past week has been a busy week for text-to-video I models. Besides the announcement of Runway’s new Sora competitor, Gen-3 Alpha, with new control and customization features, there is also Dream Machine developed by Luma AI that launched with open access to the public and “got slammed with traffic,” as well as Hedra offering a research preview of Character-1, their foundational video model that lets you create animated speaking characters. Combined with the likes of Google’s Veo and Kuaishuo’s Kling AI, OpenAI’s Sora now has some serious competition. Notably, OpenAI has not specified a date for making Sora available to the general public.

Since its founding in 2018, Runway has been making inroads into the media business, building relationships with TV and film studios whose staff use its AI-powered editing tools. While OpenAI tries to leverage ChatGPT’s name recognition to pitch Sora to Hollywood, Runway, in contrast, is taking a more grassroots approach by engaging with AI-curious creators and filmmakers. The company has organized two AI film festivals organized by Runway, receiving thousands of submissions from around the world and giving away tens of thousands of dollars in prizes for the best AI short films. The latest Tribeca Film Festival screened a series of shorts made with Sora and Runway.

For now, Hollywood professionals remain cautiously optimistic about video-generating AI, partly because they do present some cost-saving shortcuts that could benefit resource-strapped crews, and also because they are not yet good enough to fully replace human production teams. That said, hostility towards AI tends to arise whenever it extends beyond being an assisting toolkit — earlier this week, the London premiere of a movie with an entirely AI-generated script, provocatively titled “The Last Screenwriter,” was canceled last minute following intense backlash. For Hollywood, protecting their IPs (recognizable characters and house styles) from being sucked into the AI remix machine should be a top priority.

Related: Luma AI launches a Sora-like AI video generator called Dream Machine, with open access to the public [VentureBeat]; London premiere of movie with AI-generated script canceled after backlash [The Guardian]

Netflix to Open Two Entertainment, Dining and Shopping Complexes in 2025 [Variety]

The expanding experience economy keeps expanding, with Netflix set to take over two dead department stores in Dallas, Texas and King of Prussia, Pennsylvania (located 15 miles outside of Philadelphia) to open two multi-purpose entertainment complexes under the “Netflix House” branding. While detailed plans are still under wraps, Netflix says these entertainment centers will feature regularly updated immersive experiences and unique food and drink offerings. The exterior will showcase eye-catching sculptures and murals featuring popular Netflix characters, aiming to attract fans and intrigue all visitors.

Netflix’s launch of permanent offline destinations represents a strategic move to expand its revenue streams and enhance user acquisition through offline, out-of-home entertainment. This approach leverages physical spaces to create immersive experiences that extend the brand beyond the digital realm. Still, the operational costs of running large-scale venues could be significant, and ensuring profitability will be crucial to the longevity of such ventures. Just ask Disney about their park and experience division.

Related: Sony Pictures buys Alamo Drafthouse [Variety]; Pluto TV brings back ‘free movie weekend’ indie theater promotion [NextTV]

Anthropic Debuts New Claude 3.5 Sonnet Generative AI Model [TechCrunch]

Besides OpenAI, Anthropic is probably the promising one among the AI startups. It is certainly one with the highest profile outside the big three (Microsoft, Google, and OpenAI). This week, the San Francisco-based startup has released its latest AI language model, Claude 3.5 Sonnet, which has reportedly impressed early testers with its ability to understand nuance, humor, and complicated prompts, as well as write in a more natural tone. Sonnet is the first release in the Claude 3.5 family, with two more models expected to come out later this year.

The update also introduces a new coding workspace called Artifacts, where users can see, edit, and build upon Claude’s creations in real-time, seamlessly integrating AI-generated content into their projects and workflows. For example, if Claude drafts an email, users can edit it directly in the app without needing to transfer it to another text editor. User-centric features like Artifacts enable more hands-on interaction with AI outputs, transforming it from a helpful chatbot into a more versatile and interactive tool.

Related: Anthropic claims its latest model is best-in-class [TechCrunch]; OpenAI co-founder Ilya Sutskever is starting a new venture, Safe Superintelligence Inc., aiming to create a safe AI system [Bloomberg]

Situational Awareness:

YouTube is experimenting with Notes, a crowdsourced feature that lets users add context to videos [TechCrunch]

Content moderation is increasingly being outsourced to users themselves, which is better than no moderation I suppose. But then again, who is going to verify those proliferating community notes to weed out misinformation?

Apple is discontinuing Apple Pay Later [9to5Mac]

Apple says it’s shifting its focus to the new installment loan features coming to Apple Pay sometime later this year. Experts say the numbers just didn’t add up for Apple to keep running this Klarna-like BNPL service.

Japan has passed a law that will require Apple and Google to open to third-party app stores [The Verge]

Japan joins the EU in forcing the digital gatekeepers to open up their app stores. Somewhat hypocritical though, they excluded Sony and Nintendo from the list despite their policies for their respective game stores.

McDonald’s will stop testing IBM’s AI to take drive-thru orders [The Verge]

Maybe they just finally realized that IBM doesn’t have the best AI technology in the market and decided to ditch them for a better LLM provider.

Michael Kors partners with Mastercard to offer customers the company’s AI-powered ‘Shopping Muse’ for personalized product recommendations [Business Wire]

Notably, the fashion brand claims that Shopping Muse generated around a 15 to 20% higher conversion rate than traditional search queries in initial tests. Curious to see if that holds up in wide release.

• People are skeptical about AI-generated news. 52% of US and 63% of UK respondents say they would be uneasy with news produced mostly using AI; 59% worry about false news content online, up 3% YoY, per a new report published by the Reuters Institute for the Study of Journalism, based on surveys of nearly 100,000 people across 47 countries.
• Variety reports that “House of the Dragon” returned to HBO and Max with 7.8 million viewers on Sunday night, putting the Season 2 premiere 22% below the 10 million viewers that the “Game of Thrones” prequel series launched with in 2022. Citing Nielsen data, Variety attributes the drop to a decline in linear viewing, as around 1 million fewer viewers tuned in to HBO’s cable channel on Sunday than did for the Season 1 premiere.

If you find our insights valuable and would like to have a deeper conversation on technology and media innovations, or need to sound smarter in a client meeting or a pitch, please feel free to reach out to Chelsea Freitas, our VP of Strategy, at chelsea@ipglab.com.

If you liked this edition of Lab Weekly by IPG Media Lab, why not share it?