TORONTO — “Making AI helpful for everyone,” promises a sign on Google’s pavilion, located metres away from centre stage at the Collision tech conference in Toronto. Artificial intelligence has already provided one form of assistance to its more than 36,000 attendees this week: A boost of optimism.

The interest: Ankita Patel said she came to Collision mainly to see AI researcher Geoffrey Hinton speak tomorrow. “As students, we need to know where we need to upskill ourselves,” said Patel, a master’s student in information systems from Northeastern University. Another attendee, Tom Vair, a community development officer for Sault Ste. Marie, Ont., said he wants to learn how AI will impact local businesses. 

The money: Large language models (LLMs) are here, but the industry is still figuring out the business strategy. Google and Microsoft—with its stake in OpenAI—had made the early running, but Adam Selipsky, CEO of cloud titan Amazon Web Services (AWS), projected serenity amid the AI race. “There’s not going to be one model to rule them all,” he said onstage Tuesday morning, noting that different customers will need systems for their own use cases. 

AWS is building an AI technology stack it can dish out to clients, including chips for training and running systems; a marketplace for the offerings of generative startups like Anthropic and Stability AI; and its own Titan-branded models, a moniker Selipsky said the firm picked “a while ago,” before the recent tragedy that befell the submersible of the same name. “So many companies already have their data in AWS,” Selipsky said. “You’re going to need data to feed into LLMs, so it really all becomes part of a single data strategy.”

Avoiding ‘shallow applications’: The buzz is translating to demand for AI products, executives said. But “there’s risk of low-quality revenue” amid the rush, said Mike Murchison, CEO of Ada, a Toronto-based customer-service automation platform. “There’s a lot of people who are purchasing generative AI-based software just simply by virtue of it having generative AI.” That revenue in hand now could prove dangerous for firms whose products don’t meet a need once the fad passes. 

AI will flip the model for SaaS businesses, Murchison said. Instead of human labour enabled by the software, platforms will be powered by the machine, assisted by workers. Ada is set up to let customers “steadily improve their AI over time, via coaching mechanisms.”

Menlo Park, Calif.-based Khosla Ventures is avoiding “shallow applications” of AI, said partner Kanu Gulati. A company using an LLM API to offer content summarization in its product “is not exciting enough anymore,” said Gulati, who led the firm’s investment in Toronto-based Waabi, an autonomous driving startup. “You have to really rethink the application—what is this fundamentally changing?”