During an interview with CNBC’s Deirdre Bosa on Tuesday, Alphabet CEO Sundar Pichai stated that Google would take action if it discovers that Microsoft-backed OpenAI utilized YouTube content to train an AI model capable of generating videos. Pichai’s remarks follow OpenAI’s technology chief Mira Murati’s uncertainty, expressed in March to the Wall Street Journal, regarding whether YouTube videos were included in the training data for OpenAI’s Sora model, introduced earlier this year.
Murati mentioned that OpenAI utilized both publicly accessible data and licensed data. Later, The New York Times disclosed that OpenAI had transcribed over a million hours of YouTube videos. When questioned about potential legal action against OpenAI for violating Google’s terms of service, Pichai did not provide detailed specifics.
“Pichai indicated, ‘I believe it’s up to them to address.’ He continued, ‘I don’t have further comments to provide. We maintain clear terms of service. Typically, we engage with companies to ensure their understanding of our terms. We’ll handle it accordingly.’ Pichai emphasized that Google has established procedures to assess OpenAI’s compliance with regulations. Newspapers like The New York Times have already scrutinized OpenAI for purported copyright infringement and training models on their published articles.”
Pichai’s interview came after his keynote address to developers at Google’s I/O conference, during which executives unveiled new AI models, including Veo, capable of generating synthetic videos. Access to early versions of these models will require approval from Google. OpenAI made a preemptive move ahead of the Google event by introducing an AI model named GPT-4o. They demonstrated how users of their ChatGPT mobile app could engage in lifelike voice conversations, interrupt the AI assistant, and prompt it to analyze objects in front of a smartphone camera. On the subsequent day, Tuesday, Google showcased similar forthcoming features.
“Pichai remarked, ‘I don’t believe OpenAI has released their demo to users yet. It doesn’t seem to be included in the product.’ OpenAI announced in a Monday blog post that ChatGPT Plus subscribers will soon have access to an early version of the new voice mode. Pichai also mentioned that Google’s Project Astra multimedia chat features will be integrated into its Gemini chatbot later this year.”
“Pichai affirmed, ‘We have a well-defined strategy to tackle this, and we’re committed to executing it effectively.’ He mentioned that Google has significantly decreased the expenses associated with deploying AI models in web searches by 80% since providing a preview last year. This achievement was made possible by leveraging Google’s custom Tensor Processing Units (TPUs) and Nvidia’s widely used graphics processing units. During the keynote, Google also announced the rollout of its AI Overviews in search results for all users in the U.S.”
Apple’s Worldwide Developers Conference is scheduled to take place in Cupertino, California, in June. According to Bloomberg’s report in March, Apple was considering integrating Gemini into the iPhone. Pichai mentioned to Bosa that Google has had a strong partnership with Apple over the years. In court last November, a Google expert witness stated that the company shares 36% of its search advertising revenue from the Safari browser with Apple.
“We’ve prioritized providing excellent experiences within the Apple ecosystem,” Pichai emphasized. “This is a commitment we take seriously, and I’m confident that we have numerous methods to ensure our products are easily accessible. We’ve observed the popularity of features like AI Overviews on iOS during testing, and we’ll persist in integrating Gemini and other innovations into that environment.”