Facebook founder does not believe that in the future we will have a single artificial intelligence and attacks the closed models of his rivals
If Zuckerberg has his way, the future of artificial intelligence looks very different from the vision that companies like OpenAI, creators of ChatGPT, either Anthropic sell. In an interview with the Youtuber Kane Sutter published this week, he admits that he finds disappointing the language that many of these companies use when talking about the artificial intelligence tools they are developing. “They often talk about creating a real or true AI, almost like they think they’re creating God or something and that’s not what we’re doing,” Explain.
In the interview, Zuckerberg explains that there is a major disconnect in how people perceive these advances and the reality behind them, and that this disconnection is leading to a misinterpretation of the effect that artificial intelligence tools can have.
The founder of Facebook does not believe that we will have a single artificial intelligence in our lives and that there will be infinite models and tools that adapt to the needs of each person.
This vision is strongly supported by the fact that Meta’s artificial intelligence tools are built on relatively open models and available to anyone who wants to download them. OpenAI often gives access to its language models to researchers and artists, but it is generally a paid product with a clear profit motive despite the company’s name.
Other companies, such as Google, have mixed strategies. Gemini, the engine used by its artificial intelligence tools, is closed, but the company is also developing Gemma, an open model that is less powerful.
ON INSTAGRAM
The interview coincides with Meta’s launch of Meta AI Studio, a platform that allows creators to launch custom conversational AIs with which it is possible to maintain a conversation.
The idea is that these artificial intelligences function as avatars of the platform’s creators and influencers, so that their followers can have a conversation with them as if they were talking to the real person. All content created with AI will be clearly labeled and meta recognizes that, as is common with these types of tools, sometimes responses may contain erroneous or made-up information.
The tool is an expansion of a program that began in 2023 and that allowed Meta’s AI, present in all the company’s products, including WhatsApp and Facebook, to be given 28 different personalities, many trained with influential or famous people such as Dwyane Wade, Snoop Dog or Paris Hilton. The function is not yet available in all territories but Meta’s intention is to expand it throughout this year.
