ChatGPT-4o is Very Cool, But What is it For Exactly?

ChatGPT-4o is very cool, but what do we need it for exactly?

Watching the latest OpenAI demos on ChatGPT-4o reminded me of an insightful essay by Benedict Evans.

He compares modern AI to the advent of personal computers in the 80s.

Computers were invented several years before but were not accessible to everyday people. This was not only due to the high cost, but also because it was unclear what everyday use they were designed for.

This changed when Dan Bricklin observed a professor using chalk to create a spreadsheet on a blackboard and realised that a software could accomplish the same task.

He developed VisiCalc, the predecessor of Excel, for the Apple II. When he demonstrated it to accountants, it blew their minds!

Although the Apple II was very expensive, $12k in today’s money, this specific use case was so strong that every accountant bought one. It could effectively save weeks of manual work.

ChatGPT, especially GPT-4o, can do a lot of things, almost anything really. But what can we use it for in practice?

The truth is that, apart from some experiments, we only use Ai for specific tasks. For instance, I used it to proofread the text of this post, create images for my blog, and analyse spreadsheets.

But contrary to what many demos suggest, I don't use AI for everyday tasks, like organising a vacation.

Benedict Evans suggests that we’ll embrace Ai as we embraced personal computers only when specialised apps will be built to solve specific problems. Like VisiCalc did for accountants.

Even today, laptops are nothing without the specialised apps that we run on them.

I would argue that OpenAi and others are already going in that direction, similar to Apple with proprietary apps like Safari for browsing or Garage Band for music. When you purchase a Mac, it already has native apps that can perform many of the tasks you would typically use a personal computer for.

Custom GPTs are the equivalent to native apps on a Mac. Indeed, I only started using ChatGPT regularly after custom GPTs were introduced.

Search is another strong everyday use case.

GPT-4o seems to be very good at it, but others like Perplexity are also gaining momentum.

The question is…

will OpenAI, or an equivalent, dominate the Ai app market, or will we see a rise of many small and large developers delivering their own solutions to specific problems, as happened with personal computers?