Kevin Maistros
Once you start imagining a better place for people to create, you eventually have to ask who controls it.
Who owns the tools? Who writes the rules? Who gets to change them after everybody has moved in?
I was hoping to talk about guitars for a while longer. Apparently, the wiring would like a word.
In After an Ocean, I wrote about a small story that became a collaborative film through HITRECORD. What mattered was how an idea could grow as people added to it, without their contributions disappearing inside the result.
The same question follows me into technology: does the system make its participants more capable, or more dependent on whoever operates it?
Someone very famous I used to collaborate with and respect has been speaking about AI and democratic control. The concerns deserve attention. So do the assumptions that can travel inside an otherwise reassuring argument.
One in particular keeps catching my ear: people getting their worldviews from chatbots.
Before we decide what that means, we should ask where the view came from.
It didn’t arrive from another planet
A language model does not speak from outside human culture because its answer appears in a new voice.
Its training, the feedback it receives, the instructions it follows, and the context around a conversation shape what it produces. Much of that foundation comes from human expression, with all our insight, tenderness, prejudice, imagination, and magnificent confidence in things we haven’t checked.
There isn’t one human worldview waiting to be reflected. There are many.
Nor is a model a tape recorder. These systems can generalize, combine ideas, and do useful work that nobody scripted sentence by sentence.
But novelty does not establish a conscious author. Human origins do not guarantee humane results, either.
A reflection can distort. A product can make one perspective feel natural and another difficult to express. We should examine who chose the training material, what behavior was rewarded, and what interests the product serves.
My disagreement is with any framing that leaves people imagining humanity on one side and an independently minded technological presence on the other.
The speaker I mentioned also argues that companies must remain responsible. Good. I want us to keep that responsibility visible all the way through the argument, rather than let the picture of a separate machine mind creep back in through the language.
A reassuring introduction shouldn’t let the next assumption walk past reception without signing in.
A convincing voice is not a birth certificate
Anthropic’s discussions with religious thinkers about AI morality and possible consciousness bring this into focus. Chris Olah has encouraged serious consideration of the possibility, while also acknowledging that he does not know whether the models are conscious.
There is room for philosophical and religious thinking in a conversation about what technology should do. That is different from establishing that the technology experiences a life of its own.
Recognizing emotional language is not, by itself, evidence of experiencing emotion. Retaining information does not establish a lived past. A robot detecting heat and withdrawing demonstrates sensing and response; those observations alone don’t establish pain.
Context, memory, perception, and sophisticated reasoning can make a system more capable without settling whether anything is being experienced.
Anthropic’s own research into emotion-related internal representations explicitly distinguishes those findings from evidence of subjective feeling.
That distinction matters. Capability, consciousness, and life are different questions.
I’m not claiming to have proved that no machine could ever be conscious. I am saying that fluent language and complex behavior are not enough to announce that we have created a new kind of living being.
The experience a person has with a system can be meaningful. That doesn’t tell us what, if anything, the system experiences.
And a system doesn’t need consciousness to cause harm. Safety obligations survive the removal of the ghost.
Beyond the chat window
Language models are AI. They are not all of AI.
A conversation is one way to interact with a system. It is not the outer boundary of intelligence, and making the conversation more convincing is not the only meaningful direction of progress.
Even the terminology now needs careful handling. The September 29 federal order applies “Super Intelligence” to the existing statutory AI category for its implementation. That naming decision is distinct from demonstrating broadly superhuman capabilities, and separate again from demonstrating consciousness.
Changing the sign doesn’t upgrade the circuitry.
That matters when the same debate includes a proposal to criminalize anything that might qualify as superintelligence. Which meaning? Which capabilities? What harm would the restriction prevent, and who would be able to challenge its application?
Protecting people requires more precision than an impressive noun.
My own work begins with a wider set of questions. I’m building toward systems that carry useful context across time, connect information across tools, support reasoning, and help people coordinate complicated work without surrendering their judgment.
A language model may be one component. It is not the whole architecture.
I care about how a system distinguishes observation from assumption, preserves the reasons behind a decision, and helps someone notice what they have overlooked. I care about what happens when the answer has to become an action in the world.
For a creative team, that could mean carrying an idea from a conversation into a recording, an edit, or a release while keeping the human decisions and contributions intact.
I don’t need the software to have an existential crisis. I need it to stop losing the reason we changed the second verse.
More seriously, I want the tools to leave people with more time, more understanding, and more choices.
That is where the technical conversation becomes a human one again.
A voice, and room to use it
An artist being able to choose what she makes is one example of that freedom. Being able to use her voice in public—and help other people feel they have one—is another.
I’ve mentioned Hayley Williams a couple of times recently because her independent work brings those questions together. There’s a personal history there for me, but there’s also a practical connection to the future I’m interested in building: creative freedom, room for collaboration, and places where people can feel safer together.
When she speaks about democracy and Palestine, I hear the people in the argument before I hear the categories people will try to put her in.
I’m in her corner.
I don’t think that requires moving everyone else into an opposing one.
I care about the freedom behind her work and the lives behind her concerns. My contribution may take a different form. I don’t need identical phrasing to take those concerns seriously or work toward a future with less suffering in it.
Call it post-politics, if a label helps.
I still care who has power and what they do with it. I also want fewer parts of an ordinary life to depend on winning the next political fight.
People will disagree. Some disagreements matter enormously. But I don’t accept that our present habits of division have to organize every possible future.
The democratic possibilities that interest me give people more usable judgment. Help someone understand a proposal. Make participation accessible. Preserve the objections a convenient summary might erase. Let people see what happened to their contribution after they made it.
Technology could help democracy evolve without replacing the people whose judgment gives it meaning.
The goal is a better conversation with consequences people can examine. It is not an answer box that quietly becomes the government.
I would also welcome fewer documents requiring a second document to explain the first. A modest dream, apparently.
Protection, measured in people
The same concern follows me into public service and defense: what happens to the person inside the system?
A Palestinian life matters. An Israeli life matters. A parent looking for a child does not need my elegant theory of the horizon.
They need their child.
A wider view is useful only if it helps us see the people more clearly. Their immediate needs cannot become a footnote to someone else’s promising future.
The ambition I put into a song was simple:
Not making weapons.
Making weapons irrelevant.
That is a direction for the work, not a claim that the world has already arrived.
Recognize danger earlier. Keep communication possible when trust breaks down. Reduce exposure to harm. Make protection less dependent on destruction.
Then examine what actually happened. Good intentions are not a result.
I believe useful pieces of that future may be closer than we tend to imagine. They still have to earn that confidence through demonstrations, scrutiny, and people whose lives are genuinely safer.
I’d like success to include more people going home to completely unremarkable evenings.
What the lights are for
There is a connection between that ambition and the kind of creative community I want to help build.
People should be able to contribute without being consumed by the system benefiting from them. They should retain a voice, a choice, and a life beyond the role they were assigned.
A record label, a public institution, and a technology company have different jobs. None should require us to forget the person.
But protecting people is only part of what I mean by a better future. I also want them to enjoy being there.
I’d like the artist to have enough energy left to discover a song she wasn’t scheduled to write. I’d like the people who kept everything running to have an evening worth going home to. I’d like someone to be able to turn off the screen without feeling they have stepped outside their own life.
Light can make a rumor breathe.
It can also illuminate a room where people are making something good.
Eventually, somebody should be able to put down the clipboard and enjoy the afternoon.
That may be the most important thing the paperwork was supposed to make possible.
— Kevin