With the recent success of ChatGPT, the power of AI has quickly reached the mainstream. However, AI wasn’t an overnight success. For a similar breakthrough to occur in AR, a foundational shift must happen, explains Josh Pendrick, CEO and co-founder at Rypplzz.
During the fourth quarter of 2022, the now ubiquitous ChatGPT came into the public consciousness. Since then, ChatGPT and AI, in general, have flooded media headlines, conferences and industry events, social media posts, and business conversations around the world. AI was not a new technology by any means â€“ in fact, it had been hyped for years by business leaders and others. However, ChatGPT was a major tipping point for AI because of the simplicity of typing in any question or request and getting a human-like response within seconds.Â
The power of the technology was really easy to grasp. While people realize that the large language model is anything but perfect, they also realize the implications of the technology and the potential offshoots that may come from it. Safety aside, business leaders must admit that the release of ChatGPT and its subsequent impact is forcing them to think differently about AI and ways they can benefit most from the technology.
At a distance, ChatGPT may seem like an overnight success story to many, but the key to its success is the underlying technology that came before it.Â
The Transformer Architecture
In 2017, a group of research scientists at Google Brain and others co-authored a paperOpens a new window about the transformer architecture, a deep learning model for natural language processing (NLP). The Transformer architecture introduced the concept of self-attention, a mechanism that allows the model to attend to different parts of the input sequence when making predictions. This mechanism proved to be highly effective for NLP tasks, such as language translation, language modeling, and text classification.Â
The GPT architecture builds on the Transformer architecture by pre-training the model on massive amounts of text data using a language modeling objective. This pre-training enables the model to learn a rich and robust representation of language, which can then be fine-tuned on specific downstream tasks with much less labeled data. The success of GPT has had a significant impact on the field of NLP, pushing the limits of what’s possible with language models and paving the way for new applications of NLP technology like ChatGPT.
Augmented Reality is in a very similar position to where AI was in 2017. If you remember, six years earlier, in 2011, IBM Watson had just come off of a dominant performance against two of Jeopardy’s all-time champions, Ken Jennings and Brad Rutter, which was a major moment that underscored the sheer power of AI. However, many years went by without another seminal moment for the technology until, of course, ChatGPT in 2022. When it comes to Augmented Reality, there are two big breakthroughs that come to mind for many: Google Glass and PokÃ©mon Go.
Google Glass, of course, was too early for its time and posed some significant questions about safety and privacy. Earlier this year, Google announced that it was officially discontinuing sales of the second-generation Google Glass Enterprise Edition, putting an end to one of the tech industry’s most fascinating projects.Â
PokÃ©mon Go, on the other hand, stole the public’s attention in the summer of 2016, putting AR in a position to be the next big thing in technology. While there have been some notable AR applications since then, none have risen to the level of PokÃ©mon Go. Although there is no one reason for this in particular, it has become clear that AR has hit a wall of sorts and may have peaked in its current form. The technology still holds great promise, but its version of the Transformer architecture has yet to come about.
The Current Limitations of AR
Today, AR applications are predominantly supported by GPS technology and camera vision. Though GPS is a very helpful technology for long car rides, it has limitations when it comes to navigating in small spaces. As someone who lives in Los Angeles and spends a lot of time in my car, I can tell you that I’ve missed my fair share of exits because the GPS wasn’t 100% clear if I should take the next turn or the one a few hundred feet ahead of it.Â
The problem is exacerbated when trying to find a specific location on foot. AR applications currently suffer from this limitation because they are predominantly supported by GPS technology, which has 10-20 feet accuracy and can’t map things by height. For AR to have its breakthrough moment, the infrastructure supporting its applications must be dramatically enhanced. Imagine trying to build next-generation applications on Windows 95. That’s what AR app developers are currently dealing with â€“ they just don’t know it yet because it’s the best they have to work with.
Looking Ahead for AR
According to TracxnOpens a new window , there re 1,027 AR startups in the United States. Furthermore, Statista estimatesOpens a new window that by the end of 2023, there will be 1.4 billion AR device users, with that number set to increase to 1.73 billion in 2024. And, just recently in June, Apple announced its plans for the Apple Vision Pro, a mixed reality headset that will be available for purchase some time in 2024.Â
The interest in AR is there. The world’s leading technology businesses are investing in it and consumers are demanding it. Yet, outside of the occasional Snapchat filter, it’s not cutting through the clutter of a very noisy world. For AR to have its ChatGPT moment, the infrastructure that underpins the technology will need to enable more precision, deliver a smoother user experience, and be more reliable than what’s currently available today. Done right, and it will completely change how AR is viewed, including the potential applications that can be built and its benefits for all industries. Until that happens, people will continue to wonder why AR’s potential remains largely untapped.
Image Source: Shutterstock