Tech execs looking to get the most value for their organizations from generative AI need to understand the basics of the technology, according to a report released Tuesday by Forrester Research.
“The tech world has been disappointed by several recent bubbles promising much but delivering no real value, but generative AI is already improving content creation, software development, and knowledge management at enterprises,” the report noted.
“However, hype begets bad information and misunderstanding,” it continued. “Tech execs need to know some basics like what generative AI is, how it can be used, what the future holds for generative AI, and what to do with it in the short term.”
To get a handle on what generative AI is, tech execs need to dismiss some of the misconceptions about the technology.
“It sounds banal, but the biggest misconception that I encounter time and time again is that generative AI and ChatGPT are not the same things,” observed Rowan Curran, a Forrester analyst and one of the authors of the report.
“When executives look at these things, it’s important to look at them as a broad technology that has just happened to capture our imagination through a chatbot interface,” he told TechNewsWorld.
“ChatGPT is an application wrapped around the GPT-4 or GPT 3.5 turbo model,” he said. “Tech executives need to look at the models in addition to the application.”
Not as Smart as It Sounds
Generative AI is a large language model, which means it is very capable of anything related to language, explained Sagi Eliyahu, co-founder and CEO of Tonkean, a Palo Alto, Calif. maker of a process experience platform that includes AI-enabled features.
“Since we as humans communicate and even think in words, LLMs now look as if they are capable of anything,” he told TechNewsWorld.
“But even though they appear capable of ‘thinking,’ language models are ultimately constrained by the data they’ve been trained on,” he said. “Like any technology, it is only as useful as how you leverage it into the existing culture.”
“People think because it sounds smart, it is smart,” added Daniel Castro, director of the Center for Data Innovation, an international think tank studying the intersection of data, technology, and public policy.
“People should not rely on it for facts or as a substitute for human expertise,” he told TechNewsWorld. “Instead, they should use it as a tool to generate ideas and augment human skills. Generative AI has many important use cases, but it is still a long way from artificial general intelligence.”
Mistaking generative AI for artificial general intelligence — a kind of AI that can perform any intellectual task that a human can perform — is another misconception, maintained Rob Enderle, president and principal analyst with the Enderle Group, an advisory services firm in Bend, Ore.
“AGI is still years off into our future,” he told TechNewsWorld.
“What generative AI is, is a large language model that can converse with you,” he said. “It is the beginning of a new user interface based on voice and appearance that is by design more human-like in use.”
Wide Variety of Use Cases
The use of “chat” in a generative AI like ChatGPT can also confuse execs who are nimrods to AI. “They confuse generative AI with simple chatbots commonly used for customer service on websites,” observed Mark N. Vena, president and principal analyst with SmartTech Research in San Jose, Calif.
“Those chatbots are not gen AI-based, as they derive their responses from a finite universe of common questions that are generally specific to a topic,” he told TechNewsWorld. “Gen AI curates its materials, in theory, for all content on the internet, so it’s much more real-time from a relevant content standpoint and can respond to a huge array of queries.”
While acknowledging that generative AI is still relatively immature, Forrester noted that tech execs can capitalize on a wide variety of use cases, including:
- Increasing developer productivity through text-to-code generation tools;
- Enabling visual designers to iterate and ideate quickly with text-to-image generators;
- Empowering marketers to create product descriptions matching their preferred brand language and tone; and
- Scaling the presence of executives by allowing synthetic avatars of themselves to appear in videos without having to record themselves.
“One of the most underappreciated aspects of generative AI is its ability to enable more people to create software than was ever possible,” observed Bob O’Donnell, founder and chief analyst with Technalysis Research, a technology market research and consulting firm in Foster City, Calif.
“There have been no-code, low-code development tools available for years, but you still need to be very technical to get them to work,” he told TechNewsWorld.
“One of the more interesting applications of generative AI is the ability to create code from descriptions,” he continued. “That means someone with an idea, without programming expertise, can do a lot of interesting things. That’s going to be incredibly impactful for businesses.”
From Excitement to Magic
Forrester noted that while generative AI is exciting today, the applications of tomorrow will seem like magic.
For example, a future analytics platform with embedded generative AI capabilities could allow a user to submit a query like: “Create an infographic of our past year’s sales revenue, operational expenses, and customer satisfaction and include an explanation for the trends summarizing our last three quarterly reports.”
“AI now allows end users to leapfrog from research to something far more useful — resolution,” Eliyahu said.
“And not just any kind of resolution, but resolution that’s differentiated, rapid, personalized, and context-aware,” he continued. “At the end of the day, that’s what people truly want and need out of technology — for their tools to understand and promptly resolve their requests, questions, and problems.”
Forrester concedes that problems have plagued generative AI. Text generators can produce coherent nonsense, as well as recreate harmful biases baked into their data, it noted. Questions about copyright and intellectual property are also unanswered.
“Beyond the capacity for hallucination, which we’ve seen in a lot of these models, AI isn’t going to be a solution to everything,” said Will Duffield, a policy analyst with the Cato Institute, a Washington, D.C. think tank.
“There’s always a risk of trying to overfit a new technology to solve problems it’s not ready to solve yet,” he told TechNewsWorld.
Seek Gen AI Vendor Input
Nevertheless, Forrester encourages tech executives to experiment with generative AI over the next six to nine months.
“It’s really important for organizations to start experimenting in this space and start engaging with their vendor partners to understand what they are doing,” Curran advised. “Most vendors have something on their road map about how they’re going to deliver generative AI capability.”
He also recommended that tech execs take a broad look at the vendor landscape. “It’s much bigger than some of the players that have gotten all the attention over the last several months,” he said.