Roger Watt

I think I should join the 'wrong prior -> wrong inference' club.

At the risk of being v patronising...this is perfect. Simultaneously:
i. a critique of attempts to resuscitate long-defunct theories
ii. an explanation for why those attempts are hopeless
iii. a better theory of autism than any of them
All within a few dozen words.
& apologies if I've misunderstood.

and the mud itself is evidence of humanity?

Think I’ll sit back and take that as a compliment, unwarranted as it is!

Asymptote ?

I think the world gives way too much credence and importance to the views of retired professors. Expertise - in the real fluid sense - is very much in the hands of younger people.
I am a retired professor.

Sorry - certainly not questioning you!
I'm a trumpet player and it's very easy to make too much physical effort which is detrimental to performance.

the laziness principle of least effort

Does it mention all those muileann around lost townships?

scientist in balloon basket, looking at twisted balloons above

#AcademicSky #PsychSciSky

Psychologists working on the #ReplicationCrisis

And judging the wait is what musicianship is all about.

Perfection.

Years ago I showed that the time intervals between repetitions of the scale degrees in music was a Gamma distribution with α varying with scale stability. Before you can have another unstable note, a lot of stuff has to happen.
You can have C or G whenever you like, but a B or F has to wait.

I used to demonstrate this is lectures on my trumpet. You can play Amazing Grace with three stable notes (C E G) but it's deadly dull. Add in two more, less stable ones, (D A), and it's better. Add more (F B) and it's a bit too much. Go fully chromatic and no-one recognises the tune.

As well as earthquakes, by the way, music provides a good analogy of the same phenomenon. Lots of predictable notes and every so often an unpredictable one that triggers a lot of newly predictable notes and changes what the music does to the listener.

Maybe you simply can't change the rate of production of new ideas**.

**Just to be clear I don't think individuals really have ideas out of the blue in the heroic sense we like to imagine. I think ideas happen as a consequence of an environment (a book, paper) triggering them in an engaged mind.

And, thinking about time and productiveness, going back to earthquakes, main shocks need to be spaced out in time because a lot of pressure has to build up.
Ditto, maybe, for new ideas. Maybe new ideas appear only when the (intellectual) environment has built up a lot of pressure.

Which can AI help you with? I suspect that AI is quite good at doing replies - gaining followers - but is entirely useless at changing minds.
So, AI might help scientific replies, but not scientific starting points.
And in science, replies get you followers which help for a career but nothing else.

So now we ask the question about efficiency/productivity. Now we have to decide what the purpose of posts is. Is it to gain followers or to change minds? If followers, then the advice would be to make lots of replies. The second? Well, express ideas that are good, clear and unusual.

So, two distributions: time-intervals between starting points (large time-constant); time-intervals between replies (small time-constant). At the risk of derailing everything, I'm of the view that starting points are more interesting than replies.

The replies follow the original post quickly and either agree or disagree: typically a consensus is reached. The number of replies to a post depends on the 'prestige' of the OP and on how provocative the material is. I'd guess there's a fairly sharp exponential function on number of replies.

Sitting here, watching the aurora and thinking about AI science/research efficiency/productivity. Here's where I've got.
It's actually almost exactly analogous to posting on social media - and earthquakes. There is a certain small proportion of starting points and many more replies (aftershocks).

So I guess, if I wouldn't normally trust you to perform good science because you're primarily chasing a productivity goal, I sure wouldn't trust your AI-augmented science however more sophisticated, flashy, polished it looks. Faster is definitely not better. You can mass produce your poor judgment.

Agreed. So the worrying bit is that, on those terms, AI would do well in the REF, in grant applications and would do quite well for job applications and for promotion.
I have long wished that the REF (which drives the others in the UK at least) asked for each person's best single paper.

And yes the frame is just wrong.

If only this tiny part of the graphic was actually to scale…

Textbook abusive relationship

Kindness is a radical act in a violent & selfish world.
It’s our greatest, and most underutilized, superpower.

The bit about the non-existence 'typical trajectories' through health care caught my eye. In higher education (when I was part of it) the thought was to use analytics & engagement to find typical routes and flag up students who were not doing well. It just doesn't work because student life is messy.

Christ²