The Golden Trash Can of Artificial Intelligence

Do you know all the time you've been spending developing Artificial Intelligence skills? I have bad news. They are going to the trash. Because AI is a golden trash can. One which will eventually hold all the “AI skills” we’ve been working so hard to develop. Your prompting strategies for Midjourney? Useless. Your impressive collection of models for Stable Diffusion? Pointless. Your creative ways to bend ChatGPT to your will? Zzzzzz... Your catalogue of apps for every imaginable task? Trash.

Don’t cry, though. It wasn't all in vain.

With the evolution of both the capabilities and interfaces between these bots and us, the convoluted processes of today will indeed be redesigned, replaced by automatic skills and natural conversations. And that’s fine. For when we think long term, what matters most to us, explorers of this new world, isn’t how to use the current tech, but how to reinvent everything else, based on the potential of these new instruments.
Machine Learning — the real discipline behind what we call AI — is not just a new skill for computers. It’s a new kind of computing altogether. One that developed from machines that followed instructions to machines that listen, learn, interpret context and make unsupervised decisions.*

With the launch of large language models, machines learned to communicate, which dramatically accelerate its ability to evolve. Now engineers are teaching it to use tools too. They’re showing them how to understand what an external API does, how to use it and how to apply the result.**

If the natural progression continues, it won't take long until AI can use tools to make other tools. Which means they will be much closer to self-evolving, just like our own species has done.

That’s when AI will truly spike.

At that point, the only thing that will last from our early days in this field is going to be our ability to imagine that upcoming reality. The understanding of the possibilities for our future, not the skills of the present, will determine who the architects of the new world will be.

The question is: will this vision com from one of us… or one of them?

Spooky.

———

* "Programs that make decisions" is the theme of a recent post about how trying AI tools is like jumping from riding a bike to riding a horse.

** If you are curious about “machines learning to use tools,” I suggest reading about this paper called Toolformers. This link here from Jesus Rodriguez is a good start.

#artificialintelligence #ai #machinelearning #toolformers