Generative AI is a classic case of disruption. About 50 % of people talking about it mention how amazing ChatGPT is as an experience. The tool has a stunning user growth of reaching 100m users in 2 months. The velocity of Large Language Model (LLM) size increases much faster than Moore’s law. They also mention that AI will permeate everything like electricity does.

The other 50 % of people are saying that these LLMs make mistakes. Also they speak about how they don’t see the use cases for things like image generation. Some people believe that it’s mostly hype, and that the compute is too costly so the economics don’t work.

A few folks are undecided… is it real or is it hype, they ponder.

A metaphor to understand generative AI growth

It reminds me of classic Clayton Christensen and his steel mini-mills example. Incumbent companies tend to diminish the attractiveness of new entrants with different technology.

They tend to point out the features that new entrants fail at. For example, remember when the iPhone came out, there were many people saying they’d never give up their Blackberries because the iPhone didn’t have push mail. When’s the last time you’ve seen a blackberry? Also, they tend to dismiss these entrants as not sophisticated or mature enough to compete with the incumbents. Also, do you remember when social networks started, that many folks said they preferred the privacy of email as they felt uncomfortable sharing messages so publicly in a social network? How many people do you know today who don’t belong to any social network (LinkedIn is one!).

It’s hard to fit new technologies into existing structures, so one dismisses them. “It’s only a fad” one says.

If you add the speed at which Generative AI / LLM technology is evolving to what the experience it able to provide you today, and as a result you begin to imagine what these technologies will be capable of soon. You can see that some existing use cases are ripe with the technology as is. Just like college students were just right as a first use case for social networks. You build a beachhead.

You don’t have to win everything, rather you just need to win a few use cases, then a few others, and then a few others…