Zuckerberg's "AI Dream": It is not a single product but a customizable integrated tool

**Source: **Financial Association

Edit Zhou Ziyi

Image credit: Generated by Unbounded AI tools

In a recent podcast, Meta CEO Mark Zuckerberg shared his vision for future applications of artificial intelligence (AI). He hopes to integrate AI into various Meta products, including Facebook, Instagram, WhatsApp, and the "Metaverse" that Meta has been pursuing.

According to Zuckerberg's vision, AI will no longer be a single product, but a fully integrated utility. He discusses ideas for including AI in ad ideation; aiding in social interactions; and simplifying complex coding tasks in the Metaverse.

The LLaMA language model is the "ace product" of AI technology being developed by Meta. Unlike other large companies' AI products, Meta chose to open source the underlying computer code of the LLaMa system.

In addition to LLaMA, Meta also plans to deploy various large-scale language models (LLMs) to suit different needs and businesses.

Zuckerberg's Vision

Zuckerberg said that he is more inclined to a personalized AI experience, and the use of AI products currently on the market is too "one size fits all".

"We don't think there will be one right thing, we think there should be a lot of development in (AI)."

Zuckerberg emphasized the importance of personalized AI product development, which lays the foundation for a series of tools to enhance each product in Meta's portfolio.

Ubiquitous AI Applications

Zuckerberg envisions this AI being able to conceive ads based on user input: Advertisers can state their product, and the AI will craft an ad unit that connects them to their audience.

Zuckerberg said, "In the future, you just need to tell us that I am a dog walker, I am willing to help others walk dogs, help me find the right people, and create the most effective ad units."

As far as the Virtual Metaverse is concerned, Zuckerberg expects AI to play a role in simplifying complex tasks like coding: Users can describe the world they want to create, and the AI will generate the code for them.

Zuckerberg's vision for AI also includes social interaction. He believes that AI assistants can be a tool to strengthen social connections and improve communication. For example, AI assistants can remind your friend's birthday, or let some difficult-to-express ideas pass to others.

Interestingly, Meta also plans to bring generative AI technology to social platforms like Instagram. Riding on growing consumer interest in AI photo editors, Meta is developing a tool that will allow users to modify their photos with text prompts for sharing on Instagram.

In addition, Meta has a project that has not yet entered internal testing, and users can use text prompts to instruct the AI how to edit photos they plan to post to Instagram.

Open source AI system

In short, Zuckerberg believes that AI is a ubiquitous tool, not a single product, and he expects AI to be applied in various complex tasks.

A customizable LLM, such as Meta's LLaMA system, would support Zuckerberg's vision.

As the AI competition gradually entered the day-to-day stage, Meta resolutely decided to open source the code of its LLaMa system. According to Meta, anyone, whether a researcher, student or government official, can download the source code as long as they provide their email address to Meta and pass the company's review.

Meta believes that such an open source approach will expand the influence and make AI develop faster.

Yann LeCun, Meta's chief AI scientist, previously pointed out that "the winning platform will be an open platform" and said that it was a wrong decision for Google and OpenAI to become more and more closed.

This approach is bound to attract a lot of doubts. Critics believe that the open source method is dangerous because it is likely to be abused and generate some harmful speech. Regarding this, LeCun said, "You can't stop people from creating harmful information, but you can. Cut off the dissemination of such information."