Meta has become an AI success story over the past year with its decision to offer resilient AI models for free. Last week, the company released a powerful version of its enormous Llama language model, giving developers, researchers, and startups free access to a model comparable to the powerful paid model behind OpenAI’s ChatGPT. The company says all of its recent chatbots are based on the latest version of Llama.
And yet, Meta has struggled to find the right tone and niche for its own AI offerings. Last September, the company launched a series of AI chatbots loosely based on real-life celebrities. They included a fantasy role-playing dungeon master bot based on Snoop Dogg; a witty sports bot based on Tom Brady; and a daily companion inspired by Kendall Jenner.
But those bots didn’t meet with much success, and Meta withdrew them. Jon Carvill, a Meta spokesman, said the company had learned from earlier experiments. “AI Studio is an evolution,” he said.
There is ample evidence that users may find fully customizable bots more persuasive. A company called Character Artificial IntelligenceFounded by former Googlers who helped drive breakthroughs in artificial intelligence, the company has attracted millions of users to its custom chatbots.
Zuckerberg also touted other recent open AI developments from Meta at SIGGRAPH. The company has developed a recent tool for identifying image and video content called Segment Anything Model (SAM) 2. The previous version is widely used for image analysis. Meta says SAM 2 can be used to analyze video content more efficiently, for example.
Earlier in the day, in an onstage interview with WIRED’s Lauren Goode, Huang, the NVIDIA CEO, said he “absolutely” would like a “Jensen AI” that knows everything he’s ever said, written, or done. “You’ll be able to call that out, and hopefully something intelligent will be said,” he said. He could force analysts to pepper the bot—instead of him—with questions about the company. “That’s the first thing that has to go,” he said, laughing.