The mid-20th century ushered in the information age, a period defined by industry’s shift toward information technology. This era started with the shrinking size of computers and reached its peak with the creation of the World Wide Web, which brought information access to the masses. According to some technology executives, that period is now concluded due to the ascent of artificial intelligence, and a fresh technological epoch has started.
“We have transitioned from [an] information era to intelligence era,” stated Prakhar Mehrotra, PayPal’s senior vice president and global head of AI, at a conference earlier this month.
Mehrotra explained to reporter Sharon Goldman that this “intelligence era” is characterized by industries moving beyond the paradigm of simply storing and accessing data. Thanks to AI’s abilities, data can now be created more dynamically, with the long-term objective of attaining autonomy in certain workplace areas.
Businesses are competing to integrate AI—and its pledges of greater productivity and output—into their workplaces, yet the outcomes have been inconsistent. A study from August revealed that 95% of corporate AI initiatives aimed at the workplace did not achieve swift revenue growth.
“It’s going to be a journey…You have to go through this crawl, walk, and run,” Mehrotra commented. “I think that adage has been true 10 years back, is also true in this era.”
The future of AI factories
Marc Hamilton, Nvidia’s vice president of solutions architecture and engineering, who was also interviewed at the conference, stated that the future of workplace AI deployment involves investing in AI factories, located either on-site or in the cloud. This is because the data required to operate companies will be produced by AI rather than mainly being looked up by people or machines.
“When you go and say, ‘Generate a PowerPoint slide that says this,’ or ‘I’m working on this coding function, can you go in and generate code?’ It’s not retrieving it from the database, it’s taking a model and generating that data,” Hamilton said.
Mehrotra pointed out that for companies to successfully develop the computational capacity necessary to produce this data, a new fundamental unit must be valued by businesses: tokens. These are the basic elements of text that AI uses to comprehend and handle language. Tokens serve as both the pieces of information for training AI models and the content created by AI in response to a prompt.
“Every company has to think about their data in terms of tokens, because then [they] can derive that intelligence from it,” Mehrotra said.
As a measure of input and output, token generation has emerged as a crucial metric, especially for technology firms. In May, one company announced that its service, which operates on Nvidia’s hardware, produced over a certain number of tokens in Q1 of this year, representing a five-fold increase from the previous year. Such output metrics can assist AI companies in attracting investors and increasing their valuations, although some analysis indicates the link between tokens and actual demand or profits is not as strong as these companies might imply.
Mehrotra and Hamilton concurred that while many organizations now recognize the importance of tokens for enhancing AI, they are still evaluating the optimal way to incorporate them. This includes deciding which tokens to obtain externally, which to produce internally, and for which applications. Consequently, each company effectively operates its own type of AI factory, both consuming valuable tokens and producing them.
“I see it as just building that muscle,” Mehrotra said. “Like if all the employees start thinking in terms of tokens, in terms of generating process, then, yeah, it’s a different company.”