As companies debut AI offerings at a breakneck pace, consumer concerns grow

IT experts are also greeting generative AI with a mix of excitement and worry. In a new survey of IT leaders conducted by Salesforce, 86% of 4,000 respondents said GenAI will have a prominent role in their organizations — up from 57% in March — but 64% share ethical concerns associated with the tech.

Despite concerns, consumers might see marketing uses of the revolutionary tech as more palatable: 38% of consumers surveyed by Gartner said they were “very comfortable or somewhat comfortable” with GenAI used in marketing while another 27% took a neutral stance.

More companies are testing new AI tools, but most are still focused on internal uses, noted Gartner analyst Nicole Greene. She said many are also still assessing their tech stacks and cleaning data to better understand its true potential, adding that others are still concerned about protecting data and IP.

“I feel like we’ve kind of been waiting for this moment when the conversation would shift away from just ChatGPT and toward productization of this,” Greene said. “There’s a huge tension there with how companies getting access are going to use this to influence consumer decision-making and if consumers are ready to receive it.”

A flood of developments

Last week, it seemed like every major tech company made news last week related to generative AI. Microsoft debuted new AI-powered enterprise tools across Bing Chat and Microsoft Co-Pilot including new ways for marketers to summarize calls, capture feedback and write follow-ups.

A day later, Salesforce debuted new generative AI tools for its own platforms, SAP announced new investments in three key GenAI startups, and Qualtrics unveiled a new platform with generative tools for enterprise management.

Apple also is rumored to be developing a framework for a competing large language model while Google is also reportedly readying to roll out new extensions to compete with ChatGPT’s plug-ins. As for OpenAI, it released more features for customizing responses based on who a user is and which answers might be most relevant.

Across the bevy of breakthroughs, perhaps the biggest news last week came from Meta, which debuted a new large language model called LLaMa 2 that is free for research and commercial use. (The giant will still require companies that have more than 700 million monthly active users to ask for special licensing from Meta to use the open-source AI model.)

Training LLaMa 2 on publicly available content and opening it for commercial use has some advertising agencies eager to experiment. One ad executive who spoke with Digiday said the LLM’s commercial availability beyond just research purposes puts it on the “short list” for use. Others pointed out that the updated LLM also comes with features that will help developers save time by not having to do more training.

“LLaMa 2’s collaboration with Microsoft makes it an industry standout, as the model is already available in Azure,” said Dor Leitman, senior vice president of product and R&D at Connatix. “This enables companies to easily integrate and train the models in-house, addressing privacy and legal concerns. It also reaffirms Microsoft and Azure as leaders in the AI cloud space.”

Worries still abound

The latest AI announcements come amid growing calls to responsibly develop AI. On Friday, the White House met with several top AI players — including Amazon, Google, OpenAI, Meta and Microsoft — to discuss ways to avoid many of the concerns experts have long warned about. New “voluntary commitments” related to safety, transparency and trust that companies are working on include a “watermark” to help consumers know content is AI-generated, plans to have AI systems tested by a third party before they debut, and agreements to report AI systems’ limitations and potential risks around bias and privacy concerns.

Beyond governments, other organizations are also developing their own frameworks. In collaboration with hundreds of AI experts, the World Ethical Data Foundation has released a new list of 84 questions to help responsibly train, build and test AI models.

“Today we’re seeing many doomerism narratives that play into hysteria and obfuscate the structures of AI instead of focusing on clear, accountable communication that enables honest discussion and assists decision making,” according to an open letter from the WEDF signed by a number of AI experts. “The creation of AI can be a technically involved process but the standard approaches to teach machines and the outputs created can be monitored and validated in this simple framework that enables everyone to have a seat at the table regardless of a person’s data science background.”

Meta’s move to open-source signals an “all-in moment” that could lead to the “first real mass adoption movement” of generative AI, said Matt Wurst, managing director for North America at SEEN Connects. He also pointed out it’s the third time in a month that Meta “came out swinging” in an emerging media space, following its VR partnership with Roblox for Oculus Quest and the debut of Threads.

“The fact that a huge platform like Meta is bringing an open-source LLM to market for commercial use has huge implications not just for AI tech startups, but the broader industrial complex,” Wurst said. “This latest release vaults them to the forefront of the artificial intelligence technology landscape. Ease of adoption is also a Meta specialty, so it’s going to be interesting to see how that — plus Meta’s massive dataset — will give Llama a big pool of content with which to train.”