AI Briefing: Celebrity bots, ChatGPT gets new senses, and more

As Meta builds and scales its generative AI ambitions, a big question will be whether the company and other giants will learn from the mistakes of the Web2 era or if new generative AI tools will accelerate old problems. 

To develop its AI chatbots, Meta worked with various writers, producers, user researchers as well as diversity, civil rights and technical teams. It also provided new details for how it plans to develop AI responsibly to mitigate risks related to bias, privacy, misinformation and other issues. For example, typing “/reset-ai” into Meta AI will delete a conversation with the chatbot.

Researchers also warn of how “human-like” AI might actually be uniquely dangerous. A report published by the nonprofit Public Citizen looks at the the risks related using anthropomorphic design when developing AI systems, which details the ways user might be more easily manipulated by marketers but also more nefarious actors.

“Meta has just expected the public to trust them before and broke that trust,” Rick Claypool, a researcher at Public Citizen who wrote the report, told Digiday last week. “So I feel like that is a place that merits a lot of skepticism.”

Other AI news from the giants and beyond

• It was also a big week for other major AI players. OpenAI added new multimodal capabilities for ChatGPT — allowing it to “hear,” “speak,” and “see” — and days later it announced ChatGPT can finally browse the internet.

• Amazon is investing up to $4 billion in Anthropic, a generative AI company whose Claude chatbot competes with ChatGPT and Google Bard. (A day after the deal was announced, Amazon became a defendant in an unrelated antitrust lawsuit filed by the FTC.)

• IBM released updates to its WatsonX platform including new AI models and an offer to indemnify customers if they face any IP-related legal challenges while using IBM’s generative AI systems. The company also said it will publish the data sets used to train AI models, which could help improve AI transparency — something experts have called for companies to do.

• Spotify debuted a pilot for an AI Voice feature that will translate podcasts into other languages. For starters, the company is working with podcasters including Lex Friedman, Bill Simmons and “Armchair Expert” co-hosts Dax Shepard and Monica Padman.

• New stipulations for using AI are also part of the newly signed deal between major movie studios and the Writer’s Guild of America (WGA), which had been on strike since May.

AI stories from across Digiday

• Getty Images debuted its first generative AI platform, following its copyright lawsuit against Stability AI and its investment in Bria AI.

• Google also announced a new way for publishers to opt out of having their content be used for AI training but still show up in search results within Google Bard. However, publishers are worried about generative AI’s impact on traffic and IP protections.

• Snap is shuttering its enterprise augmented reality division, but sees potential for generative AI-infused AR.

• Some creators are using generative AI to make content on platforms like YouTube, TikTok and Snap. Meanwhile, Influencer agencies are also focusing more on generative content.

AI Hot Takes

It’s no surprise that AI was the topic du jour during nearly every conversation at Zeta Global’s annual conference Zeta Live, which took place last week in New York. Here’s a sample of what mar-tech founders, agency execs and CMOs said about generative AI during various on-stage talks:

• “Protecting your identity, protecting your data, protecting the information that surrounds you as a consumer is critical to us. And so it’s assumed that we do this right. The minute we do it incorrectly, it’s part of the 24-hour news cycle. And so, the better we get at utilizing these tools, the better we can deliver our promise of being on your side.” — Ramon Jones, evp and CMO, Nationwide

• “We’re seeing a lot of micro-use cases around AI, especially generative AI right now, around productivity and efficiency. I think we’re going to see a dramatic shift to what I call macro-use cases, which are around connecting the silos of the organization.” — Janet Balis, marketing practice leader, EY

• “Remember that [SAG-AFTRA] doesn’t just represent Tom Cruise. It also represents the guy who plays the voice over for the Chevy Cruise commercial.” — Steven Gerber, president and COO, Zeta Global

• “When we think about linear media, particularly, AI in the past has been too expensive to look at things like developing our own technology, our own software, and stuff like that. But particularly generative AI and the facilities to do coding and build programs and so on makes lots of stuff we wanted to do in the past quite possible now.” — Andrew Meaden, global head of investment, GroupM

• “The biggest challenge on the creative side is going to be the approval of some end clients. It sounds great to have a couple of thousand assets, but how do you approve 1,000 assets and how do you get governance around that?” — Dave Penski, CEO, Publicis Media

• “I have never seen the speed of technology permeate the boardroom that quickly, ever.” — Ajinkya Joglekar, svp of marketing, Sling

• “If the world is biased already, if it’s a bit weird already and these generative AI engines are crawling this weird, biased world we live in already, how do we avoid the situation where my daughter asks ChatGPT, ‘Give me 100 pictures of an amazing CMO,’ and then she gets back 100 white males because that’s the only thing that was indexed by ChatGPT? Not because the engine is to blame, but because that’s what they have…We have an opportunity to fix the world.” — Adam Singolda, CEO, Taboola