The question of the best AI is like the question of the best car. If there was a single best car, we wouldn’t need so many different categories of cars. The same can be said of the best laptop or the best camera.
Of course, we can come up with a set of criteria and then determine an overall winner across all offerings. This may be interesting for professionals and nerds. But the practical value is limited.
Why? Because it always depends on your own needs.
The AI market has evolved rapidly since the “big bang” of ChatGPT. At the touch of a button, you can create images, video and audio in addition to text.
There is a lot of money flowing into AI because venture capitalists are going crazy about it. Everyone dreams of funding the next super hit.
This investment bubble is bound to burst. Until it does, everyone is trying to get a piece of the action.
It will become increasingly important to clearly differentiate yourself from other offerings. A unique selling proposition (USP) is required. The USP should make it clear who a product is aimed at and what makes it special.
The central question is: Why use X and not Y? AI tools increasingly have to answer this question as well.
Of course, there are benchmarks: standardized tests that are supposed to show, for example, whether and how well a text AI can handle specific tasks.
But these benchmarks can be manipulated, and they are at best one of many ways to form an opinion.
To find the best AI for your use case, it must be clear what it is supposed to do in the first place. For example, it may be that OpenAI’s GPT-4 is the best offering for a variety of tasks. But a much smaller AI can still outperform this “best” AI in individual disciplines.
Performance is also not the only important criterion. It can also be about how flexible and customizable an AI is, whether it can be used in compliance with data privacy regulations, whether it can be linked to internal documents and databases, and much more.
We see this effect already playing out in the relatively young field of video AI. Do you need very long clips? Is it important that people are portrayed realistically? Is coherence between clips important? Each of the offerings boasts different strengths.
In this respect: The question of “best AI” makes little sense without more details. This is already the case today, and will become even more so in the future.