AI originality: squeeze out weird ideas analysis

I keep a folder of bad ideas on my laptop. Not bad in the noble, misunderstood genius sense. Bad as in half formed, embarrassing, probably unusable, and still somehow more interesting than the clean versions that survive a meeting. Lately, when I watch teams ask AI to summarize, polish, rank, and reframe ideas before anyone has had the chance to be strange, I wonder if the folder is becoming less a personal quirk than a cultural safeguard. Could AI squeeze out the weird ideas that change the world, not by censoring them, but by making them look inefficient too early?

The productivity bargain is not the creativity

bargain MIT Sloan Management Review frames the productivity debate through Nobel laureate Daron Acemoglu, whose argument is that AI’s real impact depends less on technological breakthroughs alone and more on economic incentives, institutional choices, and human centered paths. That is a useful distinction because productivity and originality are not the same game. Productivity asks whether a task got faster. Originality asks whether the work moved somewhere the team could not already see. A model can be excellent at turning ambiguity into structure, and that is exactly why it deserves caution during ideation. If the first artifact in the room is already fluent, the group may skip the awkward stage where unusual premises are allowed to breathe. The danger is not that AI gives teams bad ideas. The subtler danger is that it gives them sensible ones before the reckless ones have had a fair hearing.

When everyone gets help,

everyone may sound alike MIT Technology Review’s Rhiannon Williams describes the central paradox clearly: generative AI can make individuals feel more creative while also homogenizing and flattening collective output. The publication notes that these models make it simpler and quicker to produce text, images, video clips, and audio tracks. That speed is a gift for many workers, especially when the alternative is a blank page and a calendar full of meetings. But culture changes when everyone reaches for the same kind of assistance at the same moment in the process. A lone marketer may get unstuck. A whole department may begin to speak in the same rounded edges. Forbes Communications Council member Luciana Cemerka, a Global CMO at TP, describes a similar pattern in marketing teams: after the first rush of using generative AI for social posts, email drafts, and early blog versions, the work began to feel more similar and predictable. Not useless, not wrong, just less surprising. That should matter to builders as much as it matters to writers. Product discovery depends on weak signals, odd customer phrasing, and insights that do not yet sound like strategy. If every raw observation is immediately compressed into a tidy brief, teams may gain speed while losing the texture that made the observation valuable.

Protect the ugly stage Forbes’ account of AI in marketing points toward

a practical rule: use AI intentionally, not automatically. The first question should not be whether a tool can draft the post, plan, pitch, or prototype. The better question is when the tool enters the room. Timing is the hidden interface of creative work. One useful pattern is to split ideation into two phases. In the first, humans generate messy fragments without asking AI to rank or rewrite them. In the second, AI helps compare, expand, stress test, and translate the surviving fragments into something usable. This keeps the model in the role where it is strongest, as an accelerator of reflection, not the default source of taste. Teams can also preserve originality by saving what feels inconvenient. Keep the customer quote that does not fit the roadmap. Keep the feature idea that sounds too niche. Keep the metaphor that makes the slide worse but the room more awake. Weirdness is not automatically wisdom, but it is often where discovery starts before it learns to sound professional.

Treat AI like infrastructure, not imagination

The Knight First Amendment Institute’s essay on AI as normal technology offers a helpful way to lower the temperature without lowering the stakes. The authors argue that treating AI as normal does not minimize its impact, since electricity and the internet were also transformative general purpose technologies. The point is to stop treating AI as a mystical creative authority and start treating it as infrastructure shaped by incentives, defaults, and governance. That framing is liberating for teams. If AI is infrastructure, then creative process design matters. Who writes the first prompt matters. Whether dissent happens before or after the generated summary matters. Whether the model is asked for the most likely answer or the least obvious alternative matters. The tool does not just produce outputs. It reorganizes attention. So the next phase of AI adoption should not be a contest between purists and prompt engineers. It should be a craft conversation about where human friction is productive and where it is merely waste. Let AI remove the drudgery, translate the notes, create the variants, and expose assumptions. But before it makes the idea smoother, ask whether smoothness is what the idea needs. The builders who benefit most may be the ones who refuse to confuse polish with progress. They will use AI heavily, but not reflexively. They will protect the strange draft, the awkward customer insight, the unfashionable product hunch, and the sentence that does not yet know what it means. If the next world changing idea first appears as something inefficient, who in your workflow is still allowed to notice it?