The tone has changed
The way people talk about artificial intelligence has shifted over the past year, and the shift is easier to hear than to measure. There is less certainty in the room and a good deal more curiosity. Fewer sentences begin with a prediction. More of them begin with a question about something specific that somebody actually tried.
That is a healthy direction of travel. A field speaks in forecasts while it has little experience to draw on, and in examples once it has some. The arrival of the examples is itself the news.
Questions that travel well
The best questions being asked now are small and concrete. What did this change about the first hour of your day? What did you stop doing? What did you have to check afterwards, and how long did the checking take? Where did it quietly fail, and how did you notice?
Questions of that size can be answered honestly by somebody who is still working it out, which is nearly everybody, and an honest partial answer is worth more than a confident whole one.
Where the good conversations happen
Very little of the useful talk happens on a stage. It happens in the ten minutes after a meeting, between two people comparing notes on the same ordinary problem, with no audience and nothing to defend.
The setting matters less than the absence of an audience. Nobody is holding a position, so nobody needs the answer to come out tidy.
Those exchanges are hard to schedule and easy to miss. They reward being present often rather than being prepared.
Listening for the verb
A quick test of whether a claim has anything behind it is to listen for the verb. Talk about what a technology will mean, will change or will transform describes a place nobody has visited yet. Talk about what somebody did, tried, abandoned or repeated last week describes a place they have actually been.
Neither kind of sentence is wrong, and the first is often necessary. Only the second one can be checked.
What a careful listener carries away
Listening well across a season tends to produce a short list rather than a conclusion. A few places where the ordinary account and the ordinary experience disagree. A few tasks that turned out to be harder than anyone expected. One or two small successes repeated in unrelated places, which is usually the earliest sign that something is real.
October is a good month to write that list down, while the autumn is still young enough to act on it.
Frequently asked questions
How has the conversation about artificial intelligence changed?
There is less certainty and more curiosity. Fewer sentences begin with a prediction and more begin with a question about something specific that somebody actually tried.
What makes a good question about a new technology?
It is small enough to be answered honestly by somebody still working it out: what changed about the first hour of the day, what stopped, what had to be checked afterwards and for how long.
Where do the most useful conversations happen?
Rarely on a stage. Usually in the ten minutes after a meeting, between two people comparing notes on the same ordinary problem with no audience and nothing to defend.