A useful tool should help us think more clearly about what we want to create. That is one reason Janelle Shane’s playful look at artificial intelligence is worth watching. Beneath the strange results and surprising examples lies an invitation to become more deliberate: What are we asking for? What would a good result actually look like? And who remains responsible for deciding whether the result serves people?
Janelle Shane Makes the Question Memorable
In her 2019 TED talk, The danger of AI is weirder than you think, researcher Janelle Shane explores the gap between a task’s stated objective and the result a person intended. Her examples include unappealing generated ice-cream names and simulated machines finding unexpected ways to meet their assigned goals. The humour makes an important design question easy to remember: have we described the problem we actually want solved?
The talk belongs to its time. Its descriptions of the capabilities of contemporary AI refer to systems discussed in 2019 and should not be treated as a current comparison of every model. Our interest here is the enduring human question it raises. The practical suggestions below are our editorial reflections on working with tools while keeping purpose and judgement close at hand.
Write the Purpose Before the Request
Before opening a tool, write one plain sentence about the outcome you want. For a creator, that might be helping a new reader understand a difficult idea. For a shop owner, it might be explaining a product accurately enough that someone can decide whether it suits them. These are clearer starting points than a vague demand to make something impressive.
Then ask what success would look like from the reader’s side. Would they know the next step? Could they tell which details are established and which are your interpretation? Would they feel invited rather than pressured? A request becomes more useful when it includes the experience of the person who will eventually receive the work.
This preparation also gives you something to compare the output against. Instead of accepting a result because it sounds polished, return to your original sentence. Did the tool help achieve that purpose? If it did something else beautifully, that may still be the wrong result for this particular job.
Make Your Standards Visible
Many of our expectations remain unspoken because they seem obvious to us. We want an article to be clear, an image to suit the page, and a description to avoid promises the product cannot support. Write those expectations down. Even a short brief can name the audience, intended use, tone, necessary information, and limits.
Imagine requesting a description for a digital book. A useful brief could specify that it is a downloadable PDF, identify the actual topics covered, and require the price and format to remain clear. It could also rule out invented testimonials and guaranteed results. You have now described a communication task with boundaries instead of asking for excitement in the abstract.
Examples can help communicate style. Choose a paragraph of your own writing that feels natural and explain why: perhaps the sentences are direct, the language is warm, and the reader receives a concrete idea. An example accompanied by reasons gives you a clearer basis for reviewing the new work than a long string of adjectives.
Keep a Small Human Review Habit
Build a review step into the work before it reaches an audience. Read the finished piece as if it came from someone else. Check the names, links, numbers, and central claims. Where evidence matters, open the original source yourself. A sentence that feels confident still needs to earn your confidence.
It can help to review in separate passes. First ask whether the main idea is useful. Next check factual details. Then look at tone and readability. Finally, inspect the page or image in the place where people will encounter it. A paragraph that reads well in an editor may be difficult to scan on a phone.
This is also a chance to notice what has disappeared. Did your personal example become generic? Was an important qualification removed? Did the draft replace a thoughtful question with an exaggerated promise? Restoring these details often brings the work closer to your intentions and makes it sound more like a person speaking to another person.
Choose Measures That Reflect Your Values
Creators often count what is easy to see: views, followers, clicks, and posts published. These numbers can be useful, yet none of them alone describes the quality of a relationship with an audience. Decide what else you want your work to accomplish before allowing one visible number to dominate every choice.
For example, alongside clicks, you could keep track of recurring reader questions. Alongside publishing frequency, you could review whether each piece offered something specific. Alongside sales, you could pay attention to whether customers understood what they were buying. These observations bring the people behind the numbers back into the conversation.
A simple weekly note can help. Record one thing that worked, one thing that confused someone, and one change you want to test. You do not need a complicated dashboard to begin learning. You need enough attention to notice a pattern and enough honesty to respond when the pattern challenges your first assumption.
Leave Room for Your Own Contribution
Your experiences, preferences, and sense of responsibility give your work direction. Make room for them early in the process. Write down the story you want to tell, the problem you have actually encountered, or the question your followers keep asking. These details provide substance before any polishing begins.
When reviewing a draft, look for a place where your own observation belongs. Perhaps you tried something and found it harder than expected. Perhaps a reader’s question changed how you explained a topic. A modest, truthful detail can make the work more useful than a sweeping claim about transformation.
Keep the final decision connected to your name and your audience. You are the one choosing what to publish, recommend, or sell. Treat that choice as part of the creative work itself. Taking responsibility can be empowering: it means the tool contributes to a direction you have consciously chosen.
A Clearer Way to Begin
For your next small project, try a three-part brief: who this is for, what it should help them do, and what it must get right. Complete the work, review it against those three points, and make one thoughtful revision. Notice whether the clearer starting point changes the quality of your own decisions.
Shane’s talk gives us an entertaining place to begin this reflection. The larger opportunity is practical: become more specific about purpose, more attentive during review, and more willing to keep learning. Technology is most interesting when it invites us to bring greater care to the things we choose to make.
Watch the original: Janelle Shane — The danger of AI is weirder than you think, TED (2019). Talk details at TED.
Related reading: Your Attention Deserves a Front Door.
