Is Responsibility the Biggest Lesson in AI
As everyone settles in to the realities of managing token costs, context, skills, and agents in this new AI-first world, I can't help but wonder if we're missing the point.
I see most discussions circle the topic of what it is capable of, how to best use it, or what the future with AI will look like. While interesting, this has been the same discussion for several years and there is something more important lurking underneath most of these threads.
I'll focus on the tech industry for a moment since that is where I'm most comfortable. The discussion around AI in tech is unique in that the development capabilities that OpenAI, Claude, and others bring are novel and compelling. At least, for a moment.
For everyone espousing how wonderful they are, there are 10 more complaining about the "slop", production issues, floods of pull requests, and token spend.
If we zoom out though there is something else going on. There is discussion about ethics, security, data centers being built at great expense, concerns over water usage, and the insanely backwards economics of AI.
To put it simply, the tech industry has believed that building more tech is always the answer. It's true for developers all the way up to the AI companies themselves. We all believe we can build ourselves out of the problems we have.
What is different now is that we can move a lot quicker with this mindset than before, and note I didn't say move better.
When AI produces bugs in code, we blame AI. We don't blame the person running the agents or the person reviewing the code or the person who wrote the tests that missed it. When we produce systems that crash, leak data, suffer security breaches, we shrug and say that's how it goes.
Prior to AI we moved a lot slower and the rate of change was slower too. This naturally limited the blast radius of what our irresponsibility could do, but now those limits are gone. The same types of problems are getting produced 10x as much as they did before.
It's not AI's fault, it's ours.
We are an industry that has always loved the act of creation and the dreams that come with it. We want to build the next new thing using the next new tools. We don't want to fix the problems in what we created. We never have.
Consequently, we never got good at fixing them or avoiding them.
So what happens when our tools and technologies allow us to create faster when we were never good at doing it safely? We are pouring gasoline on a fire.
Here's the thing though. I don't think we're doomed to be stuck with this. I don't think the plague of "AI slop" is a tool problem. It's an us problem, and it is fixable.
The challenge for the entire industry isn't what can you create, it's how will you create.
I don't think there will be some major crisis about this, any more than there were crises over quality and security in the past. There will be stress, late nights, and a quiet acceptance of slight improvement that is accepted as the new normal with layers of process, management, and tools to protect everyone from our mistakes.
Except for the rare few who learn to answer the challenge. They will be prized and sought after. They will become the ones folks trust to get something done right instead of fast. They'll be the ones who take responsibility for what they and their tools produce. They'll be the ones that are 10x better with AI, and whose economics make sense.