Does all this sound familiar? It should. In many ways, these early days of artificial intelligence look and feel a lot like the birth pangs of the Internet. It isn’t quite apples-to-apples — there’s a lot that is different this time around. But there’s also a lot that is strikingly similar.
That’s good news for gov tech leaders. It means there’s precedent from which to learn. Those who recall the early days of the Internet say that those experiences can help tech executives to navigate effectively as large language models and generative AI come to the fore across a range of government operations.
THE HYPE CYCLE
The early 1990s were a time of hope and promise in the IT world. “I remember the first time I saw a browser, I guess it was in 1992, and it really sparked me,” said David Fletcher,* Utah’s former chief technology officer.
“I was the director of general services for the state of Utah, and when I first saw Mozilla, it really struck me that it was going to change everything that we did,” he said. And the present moment seems at least similar. “When I first was exposed to ChatGPT and generative AI, I thought that would be the next real major change.”
Given how the Internet has revolutionized everything from commerce to social interactions, it may be hard to recall that in its earliest days, it was all about communication. “Email was a way in which we could communicate with our state counterparts, with other local governments, as opposed to phones,” said Alan Shark,* former executive director of the Public Technology Institute, now teaching at George Mason University.
Netscape’s valuation at its IPO in 1995 was in the billions, ‘astronomical’ for the time. ‘OpenAI is worth more than a hundred times that, and it’s not even public.’
“It makes me a better editor, a better writer. It makes me more productive in every single thing that I do — in terms of making a presentation, brainstorming an idea,” Shark said. “For any local or state government individual, when it comes to personal productivity and creativity, this is a game-changer.”
In the big picture, the hype around AI feels familiar to New Jersey Chief Innovation Officer Dave Cole. As the Internet arrived on the scene, “I remember a sense of exploration, a sense of possibility, and also a sense of sort of disorganization and chaos,” he said.
Back then, he said, people were excited to share information, to make it available online. Today, “there’s that same sense of curiosity and interest.”
THE SAME, BUT NOT QUITE THE SAME
A professor at The Wharton School, Kevin Werbach worked with the Federal Communications Commission in 1994 doing Internet policy. While the present moment in AI resembles that time, he said, there are material differences.
“The hype and the financial activity, the startups and the company valuations and the investments” — all that feels familiar, he said. But the scale of the disruption has changed. Netscape’s value when it went public was in the billions, which Werbach rightly calls “astronomical” for the time. “OpenAI is worth more than a hundred times that, and it’s not even public.”
The Internet revolution was substantial, in its day. “But in some ways it was very, very small compared to AI today, where one company, OpenAI, already has 800 million weekly users, just three years after launch,” Werbach said. “We had nothing like that with the Internet, even though the Internet was very fast-growing.”
Shutterstock
RELUCTANCE VS. ENTHUSIASM
That was fast, and this is even faster. The supercharged rate of growth in AI-driven companies and AI-enabled processes promises to deliver a very different adoption curve this time around.
Back in the early days of the Internet, there were people who said, ‘This Internet thing is just a shiny toy and real IT at enterprise and large institutional scale will continue to be something completely different.’ They were wrong.
Today, “the technology far outpaces us in terms of what it can do,” he said. “That is a major swing. We were frustrated and impatient waiting for either in-house development or a vendor to solve the problems that we identified. Today, AI has moved so fast that we as humans are playing catch-up.”
Cole recalls the excitement in state government as the early Internet helped drive new levels of information sharing. “Suddenly the old constraints no longer applied,” and that helped boost adoption, he said.
As fast as the Internet improved communications, AI is delivering the goods even faster.
“We’ve seen some immediate use cases around data summarization — conflating different data sets to define matches and things like that — that have been pretty close to ready-to-go,” Cole said. As a result, adoption today is overcoming reluctance even faster than it did in the early days of the Internet.
That makes sense, considering the Internet was not so very user-friendly in its infancy. “You had to connect to the Internet, you had to buy a PC, dial-up was slow. Broadband expansion took many years,” Fletcher said.
With AI, on the other hand, “you have everything at your fingertips,” he said. “There’s an immediate ROI, and there’s not any new huge investment that has to go in front of it for people to start benefiting from it.”
Just a couple years out from the launch of ChatGPT, it’s already delivering value. “For smart people who know how to ask questions and know what they want solved, they can apply this awesome tool to real-world problems,” Fletcher said.
Given the potential for immediate ROI, it makes sense that AI adoption has hit the ground running. But if experience is any guide, it still makes sense to temper expectations.
“Back in the early days of the Internet, there were people who said, ‘This Internet thing is just a shiny toy and real IT at enterprise and large institutional scale will continue to be something completely different.’ They were wrong,” Werbach said.
“That being said, the people in Silicon Valley who said, ‘Everything’s going to change overnight and everything’s going to be replaced’ … they were wrong too,” he recalled.
It’s likely that this time, too, reality lies somewhere in between.
WILL THE BUBBLE BURST?
The dot-com frenzy sent the stock market soaring — and then plummeting. This part certainly feels familiar, and it has big implications for IT strategy.
“There’s billions being poured into AI and a lot of these startups don’t have clear business models,” Fletcher said. “Some of those companies are going to go belly up. They’re just applying ‘dot-AI’ to their name and all of a sudden they’re a billion-dollar company. It’s clear that we’re in a bubble.”
The people who are leading need to continue to lead and not be reluctant to take advantage of the great capabilities that are out there. We have a really huge opportunity to change the nature of government in a massive way.
“How is the market going to support all these companies doing the same thing?” he said. With a shake-up likely, IT leaders should “look at any long-term contract tied to AI with extreme caution. And brace for rising prices. Everyone right now is trying to get their stuff into the market. When this crashes and consolidates, the price is going to go up. Put that into your planning.”
We’ve been down this road before. “I remember when AOL started charging,” he said. It started out free, “but it didn’t remain free for long. You would get the disc with the upgrade, and you’d be loading it and paying $20 a week, whatever it was.”
With that in mind, “play with all the free stuff while you can, so you understand what it can do, how it can help you, so that at a later time you can figure out which company you want to be working with,” he said. “We should take advantage of this moment in time to try everything, while all this stuff is largely free.”
One way to hedge against the bubble bursting is to balance in-house and vendor-supported capabilities.
“When we developed our New Jersey AI assistant, the internal chat interface for state workers, we wanted to define that balance. So we’re using a foundational model from a best-in-class private corporation. At the same time, we’ve developed the front-end application to be something that we can manage and maintain,” Cole said.
“If you have a commercial solution that’s in the hands of all the users and that commercial solution for whatever reason is no longer viable, that’s a massive migration,” he said. “Whereas if you maintain the control surface and the interface that people are accessing, and do so according to your data security and privacy standards, then if you need to change something on the back end, you’re only changing one interface. It’s just a more sustainable model.”
Another effective hedge: balance the use of well-established vendors and newer players. The startups may be shaky, “but there are also massive companies that we know are going to continue to be there,” Fletcher said. “You can make a solid investment that cuts across those: investing in some that you know are going to continue to be there, while still investing in the innovative startups.”
Shutterstock
WHERE IT'S ALL HEADED
The Internet has evolved. Dial-up modems are long gone, and if that Nigerian prince hasn’t gotten the money out of the country by now, he probably isn’t going to.
Of course, it hasn’t all been for the good. Now there’s ample bandwidth available for extremists to share their ideas and lightning-fast throughput for bad actors to use when trying to undermine elections.
As the Internet has evolved, capabilities and risks have grown in equal measure. It seems fair to ask: Will the AI revolution take us along a similar trajectory? And how should
IT leaders be planning for what comes next?
Cole points to certain similarities. “In the early days of the Internet, it wasn’t a particularly secure place. It was a very open place, a very expansive place, but best practices [around security] were not universally adopted,” he said.
“If you look at AI, they are certainly meaningful questions around data security, privacy, what is acceptable use for training information, what needs to stay walled off,” he said. IT leaders need to be focused on those issues, “especially in government, where you’re dealing with public trust.”
Looking beyond security concerns, there’s a qualitative shift with AI that has no real parallel in the early-Internet narrative.
“When I think of Netscape Navigator and dealing with AOL — this is different,” Shark said. “The difference today is that the intelligence factor is totally morphed into something that I find inconceivable.”
Where the Internet merely facilitated human activities, AI can do stuff. That means we need to approach it somewhat differently. With the Internet, we just wanted to go faster. With AI, Shark said, we need to ask questions like, “How do we know that we can trust what it’s giving us, aside from the fact it looks correct? Because these systems are trained to be like puppy dogs, they’re trained to please us.”
“That puts a lot of responsibility on humans to make sure that the output is indeed solving things,” he said. “If we come up with a logarithm for some kind of benefits program, we need to check it to make sure that no one is being excluded because of some error in the computation that we failed to catch.”
On the upside, state and local leaders already seem to have benefited from the lessons of the early Internet in one important way. Where Internet-related policy and strategy took a long time to materialize, “most states already have AI strategies, and it’s just been a couple of years,” Fletcher said.
With policy and strategy rapidly coming together, government can and should move quickly. “The people who are leading need to continue to lead and not be reluctant to take advantage of the great capabilities that are out there,” he said. “We have a really huge opportunity to change the nature of government in a massive way.”
The change this time around will potentially be even more profound. The Internet ushered in a new era of communication, a new vision of connectivity. AI promises to be even more deeply transformative, “because what it changes or substitutes for is intelligence, which is at the core of all human activity,” Werbach said.
“You no longer need the same number of people to do certain jobs. Or you can do certain jobs much, much more quickly and efficiently. It creates a very different environment, even more so than the Internet did,” he said.
WHERE DOES ALL THIS TAKE US?
“If you zoom out to 30,000 feet, I would say the same thing I would have said then. That is: If you are fundamentally in the business of providing services to people in your jurisdiction, that is going to be radically changed by AI, just as it was radically changed by the Internet,” Werbach said. “You do not want to be behind that change.”
