Dieter Rams created fantastic designs of his own, and also inspired some of the most famous industrial designers of our day. His motto was famously “Weniger, aber besser” — which translates to “less, but better.”

Greg McKeown adopted Rams’s motto for his book about living a more deliberate life, Essentialism. The point of the book is that you’ll live a much more fulfilling life, and have a much more fulfilling career, if you do less, but focus more on the things that are important. Less, but better.

I’ve recommended this book to tons of people and every person that reads it recognizes the truth of the sentiment. Our lives are overrun by … stuff! How many emails do you get a day? How many meetings do you have? How many requests and projects? How long is your to-do list? In your personal life, how long is your list of “books to read”? How about movies or shows to watch?

ChatGPT and “More but Average”

This is what astounds me about how some of the biggest companies in the world (especially Microsoft and Google) are pitching their “AI Assistants.” Microsoft shows a ChatGPT integration with some publishing software that fills a page with words when prompted “write a paragraph about minimalistic design” — the assumption being that someone will then publish this somewhere for other people to consume. They create instagram reels automatically from photos. They write emails for you, and summarize responses.

Microsoft’s vision (and Google’s, to a lesser extent) of the AI enabled future is all about MORE. More emails, more meetings (that you don’t actually have to attend because they’ll be summarized for you), more blog posts, more white papers, more videos, more social media posts.

Is MORE what we need?

ChatGPT, and other LLMs, are statistical models. They are incapable of providing insight because they don’t actually understand. What they provide is what they believe a generic human is most statistically likely to reply with when asked for something.

So it’s not just more, it’s more of the same. More, but average.

Is ChatGPT terrible?

This may come off sounding like I’m anti-LLM, but that’s not the case. I think they’re incredibly powerful tools that have some amazing potential use-cases.

But the use-cases that are getting the most attention are pretty much the worst use-cases for them. So let’s use a simple filter to determine which use cases are actually beneficial.

Is this a case where “less, but better” would be appreciated? Then it’s not a good use case.

Think of email, marketing posts, social media, meetings … we all would prefer “less, but better” for those. I understand that marketers might not agree with that, but anyone outside of marketing would.

Those are all instances where chatGPT and other LLMs will make people’s lives worse. Even though ChatGPT promises to ease consumption by providing summaries of overlong meetings, marketing copy and emails, they are also the ones producing an excess of meetings, marketing copy and emails. We’ll soon be sending AI-assisted emails so that they look longer and more important, that someone on the other end will use an AI to summarize back down to something consumable. In the end our communication will become games of telephone with AI players in the middle.

But is there a case where “more, but average” is sufficient?

Writing code is a perfect example of this — I think most coders are happy writing average code, and LLMs can help coders produce more code. More, but average is fine (for a lot of coding, not all coding — I don’t want “more but average” in my Linux kernel).

Responding to more support calls with average service will be — well, fine for any company that responds to too few support calls with below average service right now. I feel like that’s basically the whole telecom industry.

So there are cases where “more, but average” is perfectly sufficient and those are the LLM use-cases we should be going after. We shouldn’t be looking to AIs to help us write and read more emails and attend more meetings. Less, but better is what we want with those things.