Hardly anyone had artificial intelligence (AI) on their minds three years ago. Have past innovations shaken up the world of business and technology in the same way? How should students prepare for the future of AI?
Three longtime technology leaders, including AI investors, startup coaches and executives at semiconductor giants, discuss how AI has changed, and will continue to change, their personal and professional worlds.
Qualcomm CTO talks AI revolutions
Artificial intelligence has shaken up the business and technology world in the last few years, but this shake-up has been going on much longer than people may realize. While ChatGPT and other generative artificial intelligence models have changed daily lives in the last few years, businesses have relied on machine learning models for much longer.
ChatGPT launched to the public on Nov. 30, 2022. But ChatGPT is far from the first AI tool businesses rely on. Jim Thompson’s company Qualcomm debuted artificial intelligence products using convolutional neural networks long before the rise of transformer-based models like ChatGPT.
Image classification model AlexNet launched in 2012 and “blew everybody away” in the engineering world, said Jim Thompson, then-vice president of engineering at semiconductor and technology licensing company Qualcomm. AlexNet could differentiate between image content, for example distinguishing pictures of dogs and cats, through a convolutional neural network.
Though AlexNet made waves among tech enthusiasts, ChatGPT and other large language models, which ran on transformers rather than convolutional neural networks, were the first AI-based technology adopted widely by the public. Transformers use an underlying attention mechanism that makes them more suitable for identifying relationships between words, sentences and paragraphs, and even time and space.
“[AlexNet] opened up the eyes of a lot of engineers like me, that oh my god, this is a revolutionary technology,” Thompson said. “But that was kind of confined to engineers, you know? …Almost anybody in college who's written a paper or cleaned up an email understands the power of [ChatGPT].”
Thompson, a Wisconsin native, grew up in Eagle Heights and received three degrees in electrical engineering from UW-Madison. At Qualcomm he worked to pioneer 2G, 3G, 4G and 5G wireless communication networks and create semiconductor chips for mobile phones and devices. He retired in 2024 after 30 years at Qualcomm, spending the last eight as the company’s chief technology officer.
Qualcomm ventured into AI by building off AlexNet to develop convolutional neural networks for computer vision. Convolutional neural networks are used, for example, when cameras on Waymo’s self-driving cars determine whether an obstacle is a stop sign, a roadblock or a traffic light, adapting their behavior accordingly.
“At the bottom line, this convolutional neural network can look at a picture and then help you determine how you process each portion of that picture and identify objects,” Thompson said.
Today, mobile phone cameras use convolutional neural networks to identify and process objects within an image after a photo is taken.
“For example, the sky: you want it to be blue, and so you make it a little bluer, or you could at least reflect the accurate color,” Thompson said.
On the hardware side, Qualcomm added AI capabilities to digital signal processors already utilizing some of the linear algebra behind AI, Thompson said.
The cultural shift following ChatGPT was “really dramatic” in comparison to AlexNet, he said, similar to the personal computer revolution of the 1980s.
“When I was a kid, there was no such thing as computers that sat on your desk. Even when I was an early engineer, they were just too expensive,” Thompson said. “What extended computing was the Internet allowing [your computer] to talk with somebody's computer that's super far away. AI is almost like a continuation of the evolution of computing.”
It took years for the convolutional neural network pioneered by AlexNet to take root in the first self-driving cars. When Qualcomm used AlexNet’s idea to build their specialized processors, Thompson said he significantly underestimated “the complexity of the software” required to make AlexNet “usable”.
“We spent a lot of time 10 years ago trying to define the software,” Thompson said. “I presented it to the organization [as] ‘democratize AI’: In other words, make it simple enough that a lot of people could take advantage of it and include it in their design.”
Similarly, AI product development won’t happen overnight, Thompson said.
Popular natural language processing models like GPT-5, Claude 4.5 and Google Gemini use billions or trillions of parameters in their operations which requires more significant memory and computing power, housed in external data centers through a process called cloud computing.
Qualcomm approached transformer-based AI from a smaller market, creating language models for small devices like phones and computers.
Thompson said the AI world is starting to follow this trend, focusing on similarly specialized and smaller markets.
“The working model is much smaller, and much more efficient, and uses less power,” Thompson said. “I think of it like [The Matrix]: remember when they stick that thing in the back of Neo’s head, and he goes, ‘I know kung fu’?”
In recent years, Qualcomm has focused on retrieval-augmented generation (RAG), a technology that eliminates AI hallucinations by forcing a language model to rely only on a small base of highly accurate data. The RAG, which functions like a smart control+F feature in a document, set of documents or database, can be used internally and for customer technical support.
Thompson described several use cases: in one, an engineer struggling to understand a complex 100-page user manual inputs their question and the manual into the RAG, which uses an AI algorithm to search for the answer in the manual and output a result; in another, a customer inputs a technical question into an AI assistant that uses a RAG to pull up the relevant Qualcomm documents and parse through them; in a third, a RAG helps a software developer figure out where to start making changes in a codebase.
“[The RAG] is not creating some weird hallucination based on all the random crap you get from the internet; it’s going after a particular document,” Thompson said. “It’s a way of controlling things and being more accurate.”
Thompson also said AI can be utilized by coders in industry and Qualcomm employees, with the caveat that it's more effective for producing mundane code than developing new software. He said undergraduates should focus on learning the fundamentals of their field because, unlike the newest AI use-case or tech fad, “they never change."
“Learning those fundamentals, being independent — the best employees are the employees that don't need to be told what to do, where they understand the big picture of what you're trying to accomplish and they understand their task,” Thompson said.
At the end of the day, Thompson said, the major technological developments he’s witnessed in his lifetime have functioned to “augment” human capabilities, not replace them. AI doesn’t mean all engineering and software students will lose their jobs, “it just means that you need to embrace the technology and get better,” Thompson said. “[AI] is going to slowly transform the economy, and we'll do better because of it.”
‘We won't be able to imagine life before AI’: Longtime entrepreneur, AI investor compares AI to the dot-com boom
Mike Trigg, a Wisconsin investor and entrepreneur who’s worked in software services for decades, said he’s been around long enough to see “a couple distinct waves” wash over the tech industry. He now believes the recent explosion of artificial intelligence is as significant as the development of personal computers, mobile phones and the dot-com boom.
“Just like you can't imagine life before the internet, or life before mobile phones, we won't be able to imagine life before AI,” Trigg told the Cardinal. “It's going to touch everything we do, transform every single industry; and it has a similar sense of inevitability about it.”
Trigg’s first jobs were in the networking and telecommunications industry, at companies like 3Com, which invented and popularized Ethernet communication, and MCI, a former rival of AT&T. Later, he transitioned to the software side, working in Software as a Service (SaaS) startups including a social media network called hi5, then joining the large file sharing organization Hightail, which he said generated $60 to $70 million in revenue each year, as chief operating officer.
Trigg got into the tech field “when the internet was just becoming a thing” in the mid 1990s. By the time he graduated business school at Northwestern University, “it was a full-fledged boom, if not a bubble.” As a current investor, he sees the same markers of the 1990s tech bubble, now from the other side of the table.
“If you don't have ‘AI’ central in your pitch deck and name, you’re almost unfundable right now,” Trigg said. “It's this really weird dichotomy where businesses that don't have that angle are getting passed by for investment, and you hear this giant sucking sound into the AI sector.”
AI accounted for 92% of GDP growth in the United States in the first half of 2025, according to Harvard economist Jason Furman. It’s outperforming the dot-com boom in terms of company profit, but national optimism is significantly lower than the “general excitement” of previous innovations, Trigg said.
“There's a perception, and I think it's an accurate one, that this is just going to make the rich richer and really concentrate wealth,” Trigg said. “In those past booms, there was a feeling that this was going to help the economy overall … this feels like [it’s] just going to make Elon Musk more money.”
At the height of the dot-com boom, the Nasdaq index, a stock index measuring the success of the United States tech sector, rose 86% in one year, 1999. But the next year, it fell precipitously by 77%. As the bubble popped in the early 2000s, companies that “didn’t have strong business fundamentals” went out of business in scores; but the ones that did, including the prestigious FAANG suite of companies comprising Meta, Amazon, Apple, Netflix and Google, have remained “drivers of economic growth” to this day.
Trigg said the question isn’t whether AI is a bubble: it’s which part is going to pop. Companies without a clear “value proposition,” or marketable service or product, will be first to go, and the others will race to establish themselves, ushering in an era of AI companies with FAANG-level prestige, he predicted.
“I think there's a lot of investors who have serious FOMO about this whole space and are kind of putting money into chasing any engineer that can say they’re doing something AI-related,” Trigg said. “But yet, I firmly believe that there will be a next generation of companies — OpenAI, Anthropic are sort of the lead horses on that now … that are transformative and enormous and incredibly successful.”
In a world where AI can write code faster than top engineers, create podcasts and media, help discover new vaccines and be your all-expenses-paid therapist at the same time, what’s left for humans to do?
“I really live this in two worlds,” Trigg said. “In my tech consulting, I work with AI-focused founders and I'm in that world of hype and excitement. But I also am an author, and in the publishing world, there's incredible fear.”
As the author of two novels, Trigg has experimented with using AI to write individual scenes, but says it struggles to capture the complete picture of a plot.
Trigg said students must “figure out how to be the architect” to keep up in the modern world, predicting AI’s ability to write large chunks of “rote” code will free up engineers to focus more heavily on design.
Though AI excels at autocompleting lines of small functions and writing emails, it struggles with creating highly synthesized designs, whether those be technically focused — databases and full-stack projects — or long-form articles, books and novels.
Though becoming the “architect” may require choosing between re-skilling or losing a job in the short term, Trigg is confident AI will raise the standards of living for everybody in the long run.
He cautioned against letting social skills “atrophy” in favor of technical acumen: “This other part of what makes you a well-rounded person … will be less apt to be changed radically by AI.”
Biotech founder compares AI ethical dilemmas to genetic engineering
Allen Dines is the executive advisor at WISC Partners, a Midwestern tech-focused investment firm partnered with the UW-Madison College of Engineering. He’s worked extensively in biotech, founding BioRenewal Technologies, which developed techniques to clean up contaminated sites. He also founded the Midwest Research University Network (MRUN), a consortium which aims to “accelerate research communication” by connecting university ecosystems with investors, according to its website.
In addition to his role at WISC Partners, Dines is involved with gener8or, a startup incubator offering investment and mentoring to startups nationwide, and Merlin Mentors, an entrepreneurial coaching organization. He’s also a Fellow at Midwest tech hub 5 Lakes Institute and MRUN. His experiences allow him to put a finger on the pulse of the investment world.
“There's people that are totally fretful, people that are saying, ‘Oh, this is going to be the end,’” Dines said. “There are other people that are saying this is incredible.”
The dot-com boom was less influential to the biotech industry than Thompson or Trigg’s industries, but Dines said AI was comparable to the boom in the agricultural and biotech industry that came after the invention of gene splicing and genetic engineering, especially in terms of the international blowback from genetically modified organisms.
“Back when I first got into this area … the first companies were coming along [that were] splicing new genes into bacteria so that they could produce insulin, for example,” Dines said. “The company I joined [was] in the agricultural space, so we were interested in upgrading the quality of various crops by using new genes that would improve crop resilience to pests and better yield characteristics as well.”
Though gene splicing is “now kind of a routine thing” in food production, Dines said, Americans at the time were concerned about its safety and ethicality.
“There was some incredible scientific potential, and we're realizing most of that, but at the same time, there were a number of people who were saying, ‘we’ve got to be careful about this’... we still have a group that talks about frankenfoods,” Dines said.
The scale of AI still shocks him.
“As an investor, I am astounded at how the investment community has gone full tilt in investing in AI, and [how] the companies themselves are building these massive data centers and doing all this expansion of their systems,” Dines said.
AI’s massive investment numbers — and they truly are massive, at about $109.1 billion in the U.S. in 2024 alone, and $368.5 billion that same year globally — should be viewed as investments in technology as a whole, according to Dines.
“We’re using AI across the board in virtually any kind of application that anybody can think of ... today, we could safely say almost every company, no matter what they're doing, is somehow a technology company,” he said.
Dines believes AI has ushered in lifestyle changes so quickly that people have already forgotten what the world was like without it, comparing LLMs like ChatGPT to the invention of Grubhub, the first app for mobile food ordering, but on a much larger scale.
“Today, people don't think anything of ordering online to get pizza, or whatever,” Dines said. “In a way, [Grubhub] changed things so quickly that we almost forget that it did massively change things, and I think that's what we're looking at with AI.”
The sheer “myriad” of applications of AI also surprised Dines, who said “it’s all over the board” from freight train logistics to neonatal care to customer service to helping high school students figure out college lists.
One startup Dines worked with wanted to use machine learning and AI to process premature infant data in the NICU, providing care staff with advance warning of any crises likely to occur. Though the startup in question struggled with funding and did not succeed, there are a host of companies looking to use AI in healthcare, including, Dines said, a company using an AI algorithm on myocardial data to detect early signs of a heart attack.
In his personal life, Dines has been using AI to answer tech questions that would have taken “hours screwing around with Google search,” like splitting a PDF file. “It actually makes searching almost like having a knowledgeable colleague,” he said.
Unfortunately, AI has also made Dines’s job harder. Emails from entrepreneurs who hadn’t done enough research on his firm used to be easy to filter out. Now, they’re using AI to craft messages that match up with the information on his firm’s website.
“The problem is if I spent my time responding to all that stuff, I wouldn't have time to do anything else,” Dines said. “So you have to draw a line. But maybe that's one of the downsides, that as more and more things [become] generated by AI, it gets harder and harder for us as individuals to deal with the information flow coming to us.”
