The AI stock boom that began two years ago passed another test this month, as enterprise software companies reported big results.
Salesforce.com and ServiceNow (right) both reported great quarters. I expect Adobe will do the same. These companies produce the software big companies use to compete. They’re saying customers are buying AI-based applications with both hands.
The question for Generative AI has always been whether customers are getting enough value to pay for the chips, software, and energy AI clouds need to keep growing. Chip companies like Nvidia are the front end of the boom, but without a back end, measured in productivity or new revenue, the boom will quickly fizzle.
So far the biggest gains are being seen among software developers. They claim they’re generating more code, faster, and getting it out the door bug-free than before AI systems could write the first drafts.
Outside tech, we’re mostly dealing with workers expecting big things and worrying about their skills. We’re still in the training stage, something that continued with the Internet long after the dot-com boom fizzled. I remember giving many speeches during the late 1990s saying there was no such thing as “e-commerce,” just commerce, suggesting what became known as “omni-channel” strategies. I also predicted the boom would bust.
We need to admit the same about AI. Despite all the fancy bells and whistles, AI is still just an outgrowth of database programming. It can create new kinds of output, it can operate in the background, but LLMs are mainly just well-trained databases. Stripping out the mystery around AI is essential to making it part of our lives.
But that won’t happen until growth slows and the stock boom busts. Before we all make use of AI, the industry will be on to something else.
Moore’s Law of Training
This delay between bringing in new tech and adapting to it is what I have called “Moore’s Law of Training.” We learn only as fast as we learn, we change only as fast as we change. Adaptation takes time, and technology doesn’t change that. (Image from ChatGPT.)
For example, I needed a haircut recently and found the shop had finally mastered the tech of 10 years ago. They were only seeing appointments. These were only available online, each barber named and pictured on the Web site, their background available at a click. Once I set the appointment, I got regular email and text reminders of it. When I arrived, however, the barber knew who I was and what I wanted. They weren’t taking cash. I just tapped my card and added a tip on the screen. All this used to require an additional worker and time from the barbers.
It was a nice haircut but a sterile experience. The old style shop, with people sitting around waiting their turn, chatting up the barbers, was community, a public act. This was a private act.
Technology remains buggy without the human element. Last month I got a call from my car dealer about a recall. I made an appointment, was reminded of it several times by text and e-mail, then I took out much of my day to drive across town for it. Turned out, after waiting in a line, with other cars, for a half hour, the work had been done months ago. The appointment was unnecessary. No one checked the tech before it brought in the customers. Imagine if I had to come from a job.
AI Will Take Time
The point is that technology filters down, from the biggest enterprises to the smallest, over a lot more time than you think. The mistakes filter down, too.
AI promises a revolution, but it will be an evolution. There will be fits and starts. There will be mistakes at all levels. It may be a decade before some people see any advantage from it. They will also see some disadvantages.
This is being ignored, as it has been ignored in every technology excitement since the mainframe.
