Facebook is relying on far-flung advances in artificial intelligence and machine learning, as well as a large number of new hires, to build its version of the metaverse.
On Wednesday, at a company event now acknowledged as Meta, CEO Mark Zuckerberg disclosed Project Caraoke, an attempt in getting AI to a “human-level” of intelligence where it can learn, forecast, and act on its own.
Zuckerberg stated that in the metaverse, AI is required to aid people in navigating between the virtual and physical worlds, and since these virtual worlds are changing continuously, AI will be required to understand the context and grasp as humans do.
Facebook is building AI with the ability to learn on its own
The project, which is pronounced like a sing-along activity, is nearly identical to Google’s DeepMind machine-learning assistant. Facebook is heavily investing in machine-learning technology for advancing its AI capabilities. The company announced that it had begun to implement self-supervised learning, a tool that enables AI models for recognizing patterns on their own.
According to Zuckerberg, algorithms in a regular AI learning model, such as those used by Facebook, learn from labels and tags fed as inputs by humans, such as Facebook content moderators. The AI then makes decisions based on the labels. Until now, Facebook’s SSL tests have outperformed previous AI models for text and video.
he added that SSL is still evolving, but we believe it will be a crucial tool for the metaverse. The amount of information in the metaverse will be far too large to be captured by labeled data sets.
For Facebook’s metaverse to materialize as per Zuckerberg’s claims, an alluring virtual world full of entertainment and effective applications will be the next version of the mobile internet. It requires technology with less inconvenience for deploying and operating. Self-learning artificial intelligence is a step in that direction.
In a different segment, Joelle Pineau, a managing director for the organization’s AI-research group, discussed the company’s work in robotics, including machines with human-like skin. Pineau also stated that the company was utilizing AI for testing robots in real-world settings such as homes, apparently for allowing them to learn like humans.
Pineau stated that learning in the physical world via trial and error, rinse, and replay is much swifter, deeper, and transferable.
Project Caraoke to enhance language potential of Facebook
Caraoke is also a part of the organization’s plans in broadening its language potential, which Zuckerberg described as a foundational issue for the metaverse.
During the event, two projects were announced in assisting Facebook to enhance its language potential: No Language Left Behind, an AI-established translating procedure for the internet that is expected to learn each language without the utilization of English as a translation base, and Then Universal Speech Translator, a tool for translating information and communication into any language while in augmented reality (AR) or virtual reality (VR) experience. Neither has been utilized.
All of the executives who spoke at the event accepted that the level of AI technology required for any of Facebook’s metaverse aspirations to turn into a reality — whether in the virtual or real world — was at least many years away.
Meta’s plans to employ additional AI researchers for building the metaverse
As the event came to a close, Zuckerberg and other executives openly invited people to apply to work for Meta in AI and other departments involved with the metaverse. The organization has recently struggled in retaining and recruiting talents given its numerous issues with harmful content, political misstatement, and other social concerns.
Even though Meta is hiring AI scientists, self-learning algorithms have a long path ahead before they can be useful to the organization.
As AI expert Yoshua Bengio put it amid the conference, we are still far off from human-level AI.
