The printing press, steam engine, electricity, mass production, the computer, the micro-chip, and the Internet are some of the biggest examples of general-purpose technologies (GPTs) over the last five hundred years, with an increasing number of them appearing over the last two centuries. Touted by economists, they define GPTs as those technologies that have impacted on many economic sectors and thus ended up changing the world, mostly for the good.

Some of the world’s most valuable companies have succeeded with these GPTs most recently on the Internet. The so-called magnificent seven — Microsoft, Apple, Google, Meta, Netflix, Nvidia, and Amazon — all succeeded on the Internet, and investors are looking for the next GPTs, most recently in generative AI of which ChatGTP just happens to include the same three-letter acronym as GPTs.

For instance, a Wall Street Journal article describing the failure of the metaverse, blockchain, and autonomous vehicles, primarily based on words from Harvard Business School professor and Andreessen Horowitz crypto research partner Scott Kominers, says,

 “They’re all ‘general purpose’ technologies that must fit in with all kinds of other things, including laws, infrastructure, and people’s expectations. And changing everything about a society — from how it regulates crypto and whether people are comfortable doing their work in VR, to laws over who is liable in the event of an autonomous vehicle crash — may take a lot longer than the billions already invested in these technologies can last.”

While economists might debate whether these technologies are GPTs or not, for us the bigger issue is that every technology must first succeed in a niche (some use the term killer application) in which the new technology is much better than the old one before targeting “everything,” something not mentioned in the article or by Professor Scott Kominers. I believe that this oversight is common in today’s world of hype and exaggeration.

Our conclusion from analyzing these and other recently purported GPTs (eVTOLs, hyperloop, virtual and augmented reality, delivery drones) is that is that today’s VCs focus far too much on vague goals 30 years down the road and not the here and now because it’s easier to raise money from investors through that approach than by targeting a useful niche. And raising big money has become the key issue for today’s VCs because that is the road to riches, not the old style of finding a niche and expanding to new markets.

Steam engines began their lives as a means to pump water out of coal mines, electricity to light city streets, mass production to produce bicycles, computers to do accounting, and integrated circuits to power missiles. Without the identification of these successful early niches, the diffusion of these technologies would have taken much longer.

I am not the first to point out the importance of niches. Everett Rogers, a professor of rural sociology at Ohio State University, synthesized research from over 508 diffusion studies to write his best-selling 1962 book Diffusion of Innovations in which he described those early niches and the evolution of them into major markets. In his best-selling 1991 book, Crossing the Chasm, Geoffrey Moore also described these early niches and their evolution for many successful high-tech products of the 1970s and 1980s.

Unfortunately, much of this common sense has been forgotten in a mad dash to scale up and hype new technologies without identifying useful niches and doing the product and market experimentation that is involved with finding useful niches. Determined to beat competitors, the question of who might most benefit from the metaverse, blockchain, autonomous vehicles, or eVTOLs is forgotten because the answer to this question is “Everyone!,” it’s a GPT!, and thus startups and incumbents must get the money, tout the products in the media through influencers, tech evangelists, and the tech bros, and then rapidly release products and keep the hype going in the media. Facebook, LinkedIn, and other social media outlets are a key part of generating the enthusiasm for these new “GPTs.”

One reason the startup system is so focused on hyping new technologies and then scaling them up is because that is how the big bucks are made. VCs take 2 percent of any monies gathered from investors so commercial success of the technology is not necessary for them to financially succeed.

A second reason is that the biggest bucks come from touting the privately held startups to new investors in a series of funding rounds — called Series A, B, C, D, E, and F — before going public. The key is to rapidly increase the valuations of the privately held startups beyond those of Unicorns ($1 billion) and Decacorns ($10 billion) to reach the penultimate level of Hectocorns, a $100 billion valuation, all before any information on profits and losses must be released to the public in an initial public offering (IPO).

No wonder early niches are forgotten in this race to be a Hectocorn! The best way to obtain a big valuation in an IPO is to obtain a big private valuation, and then convince the public that the company must be worth big bucks because a lot of very smart (chuckle, chuckle) private investors said so.

The problem is that these hypesters have not changed the economic rules of new technologies. Experimentation to find successful niches is still essential to a new technology’s success.

This is why I have looked so hard for successful “use cases” or so-called killer applications in my research on AI and other new technologies (See here and here) that go beyond cases for algorithms and Big Data in retail, search, and online fraud. Gary Smith and I have been trying to find those niches in which the economics of AI are the best, and thus AI will rapidly diffuse in this niche. A lot of our pessimism comes from finding more funny stories than successful use cases, a pessimism that makes us (and others such as Gary Marcus) strongly hated by the AI hypesters and even those who are usually quite logical in their arguments, such as Yann Lecun.

Finally, it is not just the startup system that has forgotten about the importance of niches. Big consulting companies such as McKinsey, Accenture, and PwC make outrageous forecasts ($16 trillion for AI by 2030). Furthermore, most of these consulting companies hire graduates from top MBA programs such as Harvard Business School, where Professor Kominers now resides. These business schools emphasize “disruptive technologies,” the rapid diffusion of them, and the importance of the business model far more than experimentation in products and markets to find the most useful niches.

I don’t believe that economists intended to create a monster when they coined the term “general purpose technology”, but they have. Business schools and big consulting firms jumped on this and other vague hype-filled terms, and the result was a startup system that has spent the last decade investing billions in technologies that don’t warrant the investments. As the Wall Street Journal article that is cited above says: “Some investors have warned us since 2015, at least, that the venture-capital industry and the companies it supported were in a place of irrational exuberance — and those predictions are finally coming to pass.” A nine-year bubble?