It's a hype cycle of artificial intelligence Robert Bleumoff sees it too often. Business leaders hear anecdotes about early AI breakthroughs, mistakenly mistaken it for a mature use case, fearing that they are missing out, and plunge head-on into adoption, resulting in an implementation that doesn't bring forth expectations.
“That's the chain: AI success, theater, FOMO, and some form of failure,” Akamai's executive vice president and chief technology officer said at the Emtech AI Conference, MIT Technology Review. “I've seen it happen many times.”
In his presentation, Blumofe used Akamai's AI evolution as an example of how business leaders can break the hype cycle and build AI flow ency across the organization.
Employee involvement should be at the heart of this strategy, he said. A survey conducted by Pew Research earlier this year found that about one in six U.S. workers use AI to do their job in some way.
Blumofe, who holds a PhD in Computer Science from MIT, believes that AI adoption is low. “Most jobs at this point can benefit from AI,” he said. “It's about which tasks will benefit the most and how [using] Which format of AI. ”
He provided four tips to business leaders striving to help their organization achieve AI flow.
Don't overinvest in large language models
If in fact a dedicated AI model is suitable for dealing with a particular task, too many people are thinking of AI with just large language models trained with trillions of parameters. Akamai, for example, deploys many custom models to identify and analyze potential cybersecurity threats.
“In many ways, LLM is an incredibly expensive way to solve a particular problem. It's rare for businesses to cover every event in history,” says Blumofe. “One lesson from Deepseek, he said, referring to Chinese startups where AI models are cheaper and less computing than their US competitors.
Don't LLM make your judgments successful
LLM is good for classifying emails and more, but it's a “success theater,” Blumofe said. Most enterprise issues require more complex solutions besides creating clever prompts to organize the basic format of structured data.
In most cases, LLMS is part of a “technology ensemble” that comes together for dedicated solutions. In addition to the cybersecurity threat hunting model, Akamai has developed a chatbot that answers employee questions about moving customers to a new platform, as well as a tool to write responses to vendors and requests for customer suggestions.
Explore the AI world beyond LLMS
Maturing beyond the one size approach of using LLMS means looking at technology more comprehensively. This means knowing that approaches like deep learning (to recognize patterns) and symbolic AI (to create logical responses) are better bets. “There's a whole world of AI beyond LLMS,” Blumofe said. “I argue that in many ways these models are likely to provide enterprise value in the long term.”
Let's experiment with our employees
Akamai has built an internal sandbox to “make everyone do their thing and play with AI,” Blumofe said. That approach contrasts with the company's approach of selecting a small number of AI pilots from a list of dozens of proposals. Akamai's IT team may scream at some point in time considering bandwidth usage and computing costs, Blumofe admitted, but until that happens, he doesn't feel that he needs to evaluate each possible AI usage case.
At the end of his speech, Blumofe asked members of the audience about companies such as Shopify and Duolingo, which are asking employment managers to prove that AI cannot do their job before AI can hire humans for their roles.
Such companies “grab the tail in front of the dog,” Brumov said. “The burden of proof should go against it. What is the problem you are trying to solve? What is the right technology? If it's AI, that's great – but I should have the burden of proof.” do not have ai? ”
Leading an AI-led organization
Directly on MIT Sloan
Sign up now
