human programming interfaces

Very interesting. (I witnessed) a long discussion on HPIs / human programming interfaces. More common in etiquette school, critical theory, and the social sciences.

The context seems to be that two parties from different cultural backgrounds (were) processing an objective signal ("the text" including graphics), based on different decoding assumptions.

I guess it's a meta-topic for "AI Enterprise Governance". Enterprises (as size:large capital moats) skew towards, high-context cultures, with many hidden/implicit assumptions. Maybe that points back to the meme, that as AI makes encoding/decoding of ideas very cheap ... we should expect to see an increase in variance of encodings in the wild, and an increase in utilisation of traditional training in classics and the humanities.

To "conservative decoders" this may be a reminder that there are useful resources in the wild, which are too easily discarded (false positives for a rejection test).

To "liberal encoders" this may mean that many decoders are not yet able to make sense of novel encodings. So it offers both, (1) an opportunity for encoding that goes under their radar and (2) a risk that information intended for communication may be rejected by a target recipient. 

🌞

AI : product information sheets

Re : https://www.linkedin.com/posts/chrisfongkh_aigovernance-responsibleai-airiskmanagement-ugcPost-7474641719481749505-zyW5/

Thin slicing, I believe the problem is somewhat deeper.

You seem to point out that there's a plethora of "formats" attempting to "summarise key biases, structural components, etc." ... which has thrashed around recently as much as the trendsalad of words used to describe the model architectures, which are bring brought to market in "innovative" rushiness.

I would simply posit that fundamentally the industry has not consolidated around a standards body, that issues governance language, gounded in standardised terminology.

Example 1 : business language is a mess, but folks like Gaartner provide a "point of view" 

Example 2 : Internet technologies are a mess, but IETF W3C provides a "point of view". 

There is no analogue for AI.

Maybe someday, they will end up with a Artificial Intelligence Engineer Task Force and Agentic Software Consortium.

( I have an embarassingly old interest in complaining about such messes. This is the exact problem I whinged on, about the liberal arts & sciences curriculum's lack of standardised meta-language, traversal graphs, and architecture ... as an undergrad. )

10:14 sleep:cowwork

Model A :

- 2.5 hours / 150 minutes for cowwork, twice a day

- 9 hours for desking and miscellany

Model B :

- 2.5 hours / 150 minutes for cowwork

- 4.5 hours for desking and miscellany

- twice a day

Cowwork : feeding, motor loading, grooming, medication