Minimum viable curiosity
I am dissatisfied when I ask ChatGPT or Claude.ai to write something for me. The writing has no life, no flair. It’s repeating patterns it’s been trained on, and the result is a pretty good imitation, but the voice is tinny and robotic. Anyone exploring AI has a similar experience; they test the robots on topics where they are the expert and quickly find it’s not creative, but impressively derivative. No art, no flair.
This captures exactly how I feel with AI writing. It does the thing that I ask it, and no more. It’s not capable of creating prose that resonates, that drives a point home, that iterates on what came before in a way that captures the imagination. It’s only capable of doing something based on what it saw before.
Arguably this is what we do too, a predictive model that spits out one word after the other. But I have to believe there’s something extra, some humanity, some ingrained ability to connect that’s unique to the learning process we go through to piece words together over time.
AI writing has never surprised me. It’s always just been, and sometimes that’s enough to ship. But it’s always condensing, never leading, never pushing.
Will this change? Maybe. But in the meantime we should continue to push and go through the process of learning to write for the benefits it brings to the creative process of thinking.
For every task we’re asking the robots to perform, there was an essential initial step where the robots were trained on data generated by hard-working humans so the robots could perform the task. It’s called machine learning. They need to learn from the hard work of our learning, except it’s not learning, it’s mimicking and repeating patterns.
The robots have learned from our output, but not our processes. That’s the final gem we hold back, and can use for ourselves. By struggling through the process we can create our own beauty and add it to the world.
I’ve made a career being a human terrified by becoming irrelevant long before AI showed up to drive my car. You bet I am poking every bit of AI that I can. Daily. I am trying to figure out what it can and can’t do, and this article aside, I am optimistic, just like I’ve been for the last three decades, that revolutionary innovations will knock your socks off in the next few years. It’s still early days for AI. Really.
...
AI does a shockingly good job at programming and other structured tasks we thought were the domain of hardworking engineers, but AI is not curious. AI is trained, but it does not learn.
...
The work is the trying, trying again, failing, finding inspiration in the lessons of the failure, and going one more time. Only to fail once more. Being curious. “Why am I failing?” is required reflection. You ask yourself, you ask your friends, and then sometimes a lightning bolt strikes and you realize, “This is the lesson. I understand now. I know how to improve.”
My career has shifted so many times in the last decades that this new threat of AI is only a sequel in a long line of sequels setup to try and tell me that I’m less than I can be.
Having pushed through and fought against challenge after challenge, and learned daily across a variety of media, I’m tackling AI as the next challenge and trying to find a way through it.
The barriers to creating have never been lower; all you need is a mindset. Curiousity.
One of my favorite non-fiction books of the last decade about Leonardo da Vinci drives this point home. I want to be a curious person who is always questioning, piecing together the world, and finding ways to add a little more art to what existed before.
Via Rands in Respose.