DAILY RECORD / 2022
37 published entries in 2022.
NOTES
The question for 2023
The question is whether AI will make more people more capable of acting on what matters.
The year of the new interface
In 2022, ordinary people could experience intelligence as a conversation.
The source matters
As AI becomes part of research and decisions, sources become more important.
A model cannot repair a broken process
AI on top of a bad workflow only makes the confusion faster.
AI needs infrastructure
The invisible infrastructure will decide who builds lasting products.
Systems fail at the edges
The real test is what happens with messy questions and the consequences of being wrong.
The future of education is personal
AI could give every learner a patient, available, adaptive guide.
A quiet technological shift
Some technological shifts begin quietly, through millions of private conversations.
Context is the real product
The best AI systems will understand enough about a goal to make the next step clearer.
The cost of an answer is collapsing
The cost of a first answer is approaching zero.
Every company will become an AI company
Every company will need to decide where intelligence belongs in its work.
The confidence problem
The most dangerous answer is the plausible one that makes you stop checking.
Intelligence has a trust problem
The future of AI depends on knowing when to trust it, question it, or ignore it.
More answers, better questions
AI will make answers abundant, which may make questions scarce.
The next generation learns differently
People growing up with an AI tutor may experience information differently.
The chatbot is not the system
ChatGPT is a window into something larger.
Power needs an interface
The technology that matters makes its power usable without creating dependency.
Helpful is not harmless
The more natural AI feels, the easier it will be to trust it too quickly.
The speed of explanation
AI could lower the fear of not understanding quickly enough.
Intelligence and attention
The opportunity is for AI to help people focus on what matters.
The model does not know me
A useful assistant will need more context, but context makes privacy and trust real.
The output problem
As output becomes cheap, taste and judgment become more valuable.
The first personal operating system
People may use AI to think through decisions and organise fragmented parts of life.
The research assistant
The first thing I would want from AI is help seeing the map of a question.
A tutor before a teacher
A patient system could explain something ten different ways.
The copy is not the point
More people can turn an unfinished thought into a first draft.
Everyone is testing the edge
The first use cases are chaotic, but that is how new interfaces reveal themselves.
The answer is not the point
The interesting part is what happens after the first paragraph.
The first mass AI interface
People can now open a page and test AI for themselves.
Software that talks back
Most products give people options; this one gives people a response.
The question is the skill
The interface is becoming language.
A different kind of interface
A computer now feels like something you can think with.
Intelligence without context
AI can generate something convincing without understanding the person receiving it.
A new kind of literacy
The people who understand AI first will not necessarily be the best programmers.
The computer still asks too much
Most software still makes people adapt to its structure.
The model is not the product
A powerful model is not automatically useful.
The models are waiting for an interface
AI image models and coding tools feel impressive, but still slightly distant.