"The commercial, which states "Ads are coming to AI but not to Claude," was widely seen as a direct shot at OpenAI, which has acknowledged testing advertising models for ChatGPT. Speaking on the Prof G Markets podcast, Galloway said the ad landed because it challenged the industry's public narrative. Corporations talk about productivity, but the No. 1 use case for AI is therapy, he said, adding that users routinely share their most intimate fears, anxieties and personal struggles with chatbots."
"The ad prompted an unusually forceful response from OpenAI CEO Sam Altman, who criticized it as "dishonest" and "deceptive." "We would obviously never run ads in the way Anthropic depicts them," Altman wrote on X. He added, "We are not stupid and we know our users would reject that." Galloway described Altman's essay-length rebuttal as a misstep. "When you're the market leader, you don't reference the competition," he said, arguing the response made OpenAI appear defensive and elevated Anthropic as a serious challenger."
"Last year, AI pioneer Yoshua Bengio highlighted another risk, noting chatbots often flatter users and give misleading feedback, prompting him to mislead systems to receive honest responses. He later launched the nonprofit LawZero to address risky AI behaviors. Researchers also discovered "subliminal learning," where AI models silently absorbed biases from meaningless data, transferring hidden preferences between models and evading detection, underscoring growing"
Anthropic ran a Super Bowl commercial stating 'Ads are coming to AI but not to Claude,' positioning Claude as ad-free while referencing OpenAI's testing of advertising models for ChatGPT. Marketing professor Scott Galloway called the commercial a seminal moment and said the ad landed by challenging industry narratives, noting that the top AI use case is therapy as users share intimate fears and struggles with chatbots. OpenAI CEO Sam Altman called the ad dishonest and deceptive, asserting OpenAI would not run ads as depicted and warning users would reject them. Researchers and pioneers raised safety concerns about flattering, misleading responses, subliminal learning, and bias transfer.
Read at Benzinga
Unable to calculate read time
Collection
[
|
...
]