Bibs & bobs #5

 Bibs and bobs #5

Delegating work to nonhumans

This b&b is short. I’ve been spending way too much time mulling the big framings of automation versus augmentation. While I have a lot of sympathy for the argument put forward by Erik Brynjolfsson in this paper: Brynjolfsson, E. (2022). The Turing Trap: The Promise & Peril of Human-Like Artificial Intelligence. Daedalus, 151(2), 272-287. https://doi.org/10.1162/daed_a_01915, What needs a lot more attention is what actually happens when augmentation happens, the trade-offs, the capacity swaps between human and nonhuman. There is a lot of noise in AI around the neglect of this issue, e.g. the alignment problem and the other panics associated with giving large language models tasks. Much more to say and think about.

Abstractions

A wild, playful and insightful ride courtesy of Venkatesh Rao, @vgr. If you emerge from this unmoved then good luck with keeping those trusty mental routines spinning the way they always have.

Expectations

There is a lot of fun things the brain does when we tinker with things like expectations, placebos and the like. 

Reproducibility

Via Steve Stewart-Williams, 

73 teams tested the same hypotheses with the same data. Some found negative results, some positive, some nada. No effect of expertise or confirmation bias. "Idiosyncratic researcher variability is a threat to the reliability of scientific findings." 

link.

Humour

If you are in need of inspiration for a book dedication, a list.  And if you were a fan of Whose Line is it Anyway?, you might enjoy ToonProv. 

Bibs & bobs #4

 Bibs and bobs #4

Maps and their effects

A wonderful video by Johnny Harris about “the island of California” a period in human history that I found useful when thinking about scenario planning.  The video is also a neatly framed instance of fake news.

AI and education

There is a no shortage of commentary, hype, spin, doom saying and wishful naming [McDermott, D. (1976). Artificial Intelligence Meets Natural Stupidity. SIGART Newsletter(April), 4-9.] to be found in relation  to AI and education. This excellent post by Michael Feldstein gives a useful overview of the current state of what I think of as LLM wrangling. As I have noted and the focus of much of my thinking is concerned with the problem of delegating work to machines. It seems to be very much black box territory. You poke the LLM with text to see how it responds. 

This is exactly the logic that ought to inform thinking about how to deal with LLMs as they currently exist and their deployment in formal education settings. Instead of having educational panic  #971: OMG we can’t use a plagiarism checker to see if this was written by a student or a machine. I recall the time when software that generated crossword puzzles appeared. Many teachers were overjoyed, an app (called software way back when) that created busy work for students. Yay! There were however a few teachers who embraced the app differently. They had students use the app to produce crosswords. You can guess which students learned more about a topic built around a crossword.

Educational panics about the digital go back at least as far as the advent of electronic calculators a very long time ago. The opportunity to think through their use and ask more sensible questions, e.g. what complementary skills do students need in order to use these devices, was largely missed. Approximation skills anyone? 

Formal education requires a selective amnesia. It is illustrated by an almost manic capacity to preserve practices that have long outlived their usefulness. The origins of the practices are long forgotten. They were likely developed to solve a particular problem at the time, a problem that no longer exists. The practice lives on, ghostly, inexorably. Age-based schooling is an obvious example. In time so will the current madness around measurement.

So as we begin to see educational panic after educational panic over AI and formal education. It is reassuring to know that there are sane folk out there, e.g. the book by Mike Sharples & Rafael Pérez y Pérez, Story Machines, which illustrates an alternative approach to think about writing and LLMs. The Story Machines website is here.

I’m still of the view we seem to be too attached to a single and limiting view of AI which is why I like the argument in the post Venkatesh Rao wrote about AI as artificial time or super history It’s different, and IMHO a better way to think about AI as it is currently being developed. 

Neurotypification

a normal person is anyone who has not been sufficiently investigated - Edmond A. Murphy

Elegant post on measuring mental traits. A demolition job on the notion of normality. 

via @RosemarieNorth

Neurotypical syndrome is a neurobiological disorder characterized by preoccupation with social concerns, delusions of superiority, and obsession with conformity. There is no known cure   —Laura Tisoncik

So many spectra, so little time to find my spot on each.

Searching

This list of search options was compiled by the good crew at Recomendo.

Links

Stephen's Web ~ Education at a Glance 2022 ~ Stephen Downes  links to the OECD annual report.

HOME | OpenAcademics  well worth a prowl around.

ditto for Academic Chatter | Twitter, Instagram | Linktree

and Online Library and Publication Platform | OAPEN fo open access books

Bibs and bobs #3

 Bibs and bobs #3

Delegating work to nonhumans or machines

A detailed and thoughtful post by Peter Greene about using robots as replacement teachers. 

i’ve long held the view that the easily justifiable use of machines in education was to support students with various disabilities. It’s not a simple task. Whenever you have a machine do something for you, there is always an exchange that takes place between the human and the machine. This exchange is often bracketed as adapting to the interface of the machine. That is important and obvious but there are always more subtle capacities in play for the user. An illustration that I have used to clumsily make this point is that of a calculator. If you use a calculator there often will be complementary skills that are necessary for the user depending on the nature of the calculation. The most obvious one is approximation skills, i.e. you can look at a sum and quickly work out that the answer will be roughly .5 or 5,000 or whatever. There are other complements that I won’t point to. The idea of complementarity I trace back to Bruno Latour’s famous reflection on the sociology of a few mundane artifacts. The simple summary is this: it is not a simple consideration. Latour clearly demonstrates that with his analysis of an automatic door closer. To me all of the noise around using AI falls into this problem space: when you delegate you still have work to do, different from what the machine has done.

And then

DALL-E now open to everyone.

Then there was text to video from Meta AI. 

On Writing

The Uneven U notion for writing drafts is neatly explained. Well worth a read via Naomi Barnes.

Books

An open access pdf of Weller, M. (2022). Metaphors of Ed Tech. AU Press. https://doi.org/10.15215/aupress/9781771993500.01  available here. A useful commentary from Stephen Downes.

Researching

Metaphors and stories to talk and think about some routine research work via Stephen Downes. 

Managers, leaders and such

A useful, thoughtful piece on those who find themselves in leadership, management roles.

Humour

Social theorists as Jedi knights – A Twitter Thread

Universities pay staggering salaries to Presidents, Chancellors, VPs and provosts by the dozens, etc and in every administrative office there is a 57 year old woman named Peggy with a title like "Admin Assistant II" and that's the person who actually runs the university, via the Thesis Whisperer.