Spotlight on AI is Good for the Metaverse

Subscribe To Download This Insight

2Q 2023 | IN-6971

Artificial Intelligence (AI), led by generative AI, has created a new hype cycle that has seemingly pushed metaverse to the backburner. AI, however, is a critical enabling technology of the metaverse and should be viewed as such. Further the focus on AI also brings much needed development and attention to other areas like computational resources, content creation, data and workflow management, and privacy and security—all applicable to the buildup to a future metaverse.

Registered users can unlock up to five pieces of premium content each month.

Log in or register to unlock this Insight.


It's not AI or Metaverse, but AI and Metaverse


After OpenAI’s ChatGPT exploded into the public consciousness, it drove Artificial Intelligence (AI) to quickly capture the tech spotlight and, in the process, unceremoniously dethroned the metaverse as the “next big thing in technology,” or so it would seem. Media headlines now suggest a passing of the torch has occurred between the metaverse and AI, one hype cycle giving way to the next, but this is an oversimplification of the maturity of these markets and their interconnectedness. A future metaverse, for instance, will need to leverage AI from the edge to the cloud enable applications/services, manage content/data, etc., with the edge representing multiple layers which includes devices, telecom edge, and Content Delivery Networks (CDNs).

News like NVIIDIA hitting a US$1 trillion market cap, driven in part by the potential for AI, is certainly a boon for the AI value chain, but it is also a foundational piece for the future metaverse. AI and ML are critical components of content creation, digital humans, perceiving the real world (e.g., Extended Reality (XR)), handling of data and workflows, and, more generally speaking, automating processes that would otherwise be unmanageable. Collectively the use of AI and shift from 2D to 3D interfaces will create new demands on computational resources that exceed what would be projected through organic growth. The use of digital humans for example can require up to ten times the computational resources compared to a typical chatbot, similar computational requirements extend to generative AI derived searches. While the cloud is scalable, it can only scale so far; AI processing and computing more generally, therefore, will still need to be distributed across devices and networks.

These future demands on compute make other news like Intel’s unveiling of details about its next generation Meteor Lake processors (mobile PC devices first) quite noteworthy, as it includes a Vision Processing Unit (VPU) to complement the Central Processing Unit (CPU) and Graphics Processing Unit (GPU) within the System on a Chip (SoC). Despite the established use cases for AI and the expected growth, there remains some skepticism when considering the grander visions, in large part fueled by “failed” past technology hype cycles. There is no question that AI has a key role to play within the technology markets, as it is already the case, but questions start to arise if AI will live up to the promises made within the early stages of this hype cycle. While the longer-term impact is certainly debatable, the impact of AI on markets, including the metaverse, is more readily definable.    

AI in the Cloud and at the Edge - A Driver for the Metaverse


In many respects, the shift in attention to AI should be welcomed by metaverse advocates since it moves the focus from a comparatively distant future to one of the core building blocks of the metaverse. While some of the hype cycle does point to longer term visions as well, the excitement, like that seen with the metaverse, is grounded on pre-existing trends and real-world use cases that point to this longer-term potential. Aspects of these longer-term visions can engender some skepticism, meaning it is best to consider the broader environment than focus on each case individually.

Intel’s Meteor Lake announcements for example have led some to question if there are enough applications to merit embedding the VPU onto the processor. Intel highlighted, among other applications, video conferencing (e.g., virtual backgrounds, eye contact correction, etc.) and creative markets, but for some this still felt like it offered limited value to mainstream users, while others viewed it as a “build it and they will come” prospect. The market environment, however, would suggest otherwise. Qualcomm, for example, claims to have shipped over two billion AI products (its Hexagon processor at the core of its AI Engine), while others like Imagination Technologies (neural network accelerators), AMD (Ryzen AI), and Apple (Apple Neural Engine) have similarly brought together a dedicated AI processor with the CPU and GPU.

Further, Intel has included AI processing in its hardware prior to Meteor Lake: Intel’s Gaussian & Neural Accelerator (GNA) block, for example, is used for audio processing (e.g., eliminating background noise) and has been part of the Core lineup since the tenth generation Ice Lake processors. Additionally, the VPU is Intel’s third generation solution since it acquired Movidius in 2016, so Meteor Lake is part of a progression of steps to bring AI to x86.

Within the context of the metaverse, bringing AI processing to devices addresses many potential needs and use cases:

  • Content Creation: From images and video to 3D content, AI can bring accessibility to content creation. Generative AI is already being used to create 3D assets from 2D images, saving time and resources for enterprise and industrial users. A broader accessibility of content creation tools will create more opportunities within the consumer space.
  • Managing and Processing Data: Beyond the more immersive elements of the metaverse data is at the core of this vision, by analyzing and collecting data about the real world through machine and computer vision. Not only in terms of collecting data from more places (sensors, external data sets, simulations, etc.), but how these integrate to create a more comprehensive representation of the whole, be it a factory, building, city, supply chains, individuals’ personal data clouds, or Earth itself.
  • Powering Large Language Models: AI processing can be used to power digital humans, virtual assistants, and user interfaces. Some of these large language models will be made suitable for processing on device, with more complex queries or tasks shifting the cloud.
  • Privacy and Security: The metaverse speaks to decentralization of control and this often extends to user data. Regulations and general consumer sentiments also favor stronger controls for users over their personal data. Pushing AI processing to the devices allows data and any personalization to remain local, giving users control over their data in how it is used, stored, and accessed. Relatedly, some companies, either due to requirements for latency and/or controls over data and IP, will require local/edge processing and storage of data.  

Interoperability, Content, and Services Need to be Top Priorities


For the metaverse, focus needs to be placed on the core enabling technologies, like AI, rather than what the long-term vision means for today’s market, because actions and decisions made today and on the buildup to the metaverse will ultimately shape and determine what that future looks like. Focus also needs to be placed on interoperability and the content and services that will fuel these experiences—technology alone is often not adequate to drive adoption. Standardization and regulatory environments will need significant work to ensure the future metaverse is fair, accessible, and capable of addressing all the needs and demands across industries and breadth of users.

In the context of content creation, significant work is needed to determine content usage and rights with AI generated/assisted content. Companies like Adobe, which has extensive libraries of content, can train models with these assets to ensure proper rights are maintained, but most companies do not have access to these content libraries nor should users be forced to only use tools and applications from companies with these types of resources. Work is needed to ensure generative AI models produce content that is both usable (unencumbered by copyright) and, presuming there is enough human creative input, capable of establishing ownership or rights to created content. The Content Authenticity Initiative is a good example here that is currently working to add a layer of verifiable trust to digital content (initially images and video) that preserves metadata and history about that asset from creation through alterations.

Ultimately the metaverse is a process and progression of steps towards a future that includes more immersive use cases and 3D interfaces. AI has the spotlight, but it is an essential element and catalyst of the metaverse and should be viewed as such, rather than a replacement or alternative.