Template:M intro technology robomorphism: Difference between revisions

From The Jolly Contrarian
Jump to navigation Jump to search
VisualWikitext
Created page with "Matt Bradley made an interesting point about the gallop towards AI: we should be careful of anthropomorphizing robots. What they do is a bad proxy or derivative of human intelligence. It is not the same, and we should not let habits of language conflate the two Ther’s another way of looking at it, too: we should not see what humans do as a proxy for how robots behave, either. Not only should be be mindful of lazy computer metaphors for human behaviour, but we should b..."
Tags: Mobile edit Mobile web edit
 
No edit summary
 
(14 intermediate revisions by the same user not shown)
Line 1: Line 1:
Matt Bradley made an interesting point about the gallop towards AI: we should be careful of anthropomorphizing robots. What they do is a bad proxy or derivative of human intelligence. It is not the same, and we should not let habits of language conflate the two
{{quote|
“If to call a man a wolf is to put him in a special light, we must not forget that the metaphor makes the wolf seem more human than he otherwise would.”


Ther’s another way of looking at it, too: we should not see what humans do as a proxy for how robots behave, either. Not only should be be mindful of lazy computer metaphors for human behaviour, but we should be wary of evaluating humans by reference to machine qualities, much less optimising our criteria for human contributions for numerical processing. For if there is a sure way to technological redundancy, that is surely it.
:—Max Black, ''{{pl|https://web.stanford.edu/~eckert/PDF/Black1954.pdf|Metaphor}}'' (1965)
(Technological redundancy, in its place, is no bad thing — we should not lament the demise of manual comparison over deltaview, any more than we should lament the demise of hand weavers over the Jacquard loom).  
}}
But technological redundancy has its place — clearing out the tedium and bureaucratic sludge in well-understood, low-risk, standard processes — but that seems not to be the aspiration of the thought leaders.
{{qd|Robomorphism|/rəʊbəʊˈmɔːfɪz(ə)m/|n|
“Any sufficiently advanced technology is indistinguishable from magic” says Arthur C. Clarke — the jury is out whether AI is different, but it is not unreasonable to proceed on the assumption it is not, and foolish to do otherwise.
The interpretation of human behaviour, activity, or interactions in terms better suited to a [[Turing machine]].}}
The main use cases for machines in any industry from the beginning of civilisation are these: power, speed, accuracy, efficiency, economy.
 
Machines do things easier and cheaper and quicker than humans.
{{drop|R|ecently, Matt Bradley}} made an {{plainlink|https://www.linkedin.com/pulse/why-humanise-machines-matthew-bradley-adgce|interesting point}} about our gallop towards [[AI]]: whatever we do — however tempting it is — we should be careful not to anthropomorphise when we talk about machines.
They therefore work best in constrained, predictable environments that have, as far as possible, been preconfigured to eliminate unknowns and minimise waste. The factory production line is the paradigm example. Human intervention is minimised and, where possible, eliminated.
 
Humans have been largely absent from production industries — this has been a great driver of the colossal pivot to the service industry. It isn’t that we don’t make things any more: we just don’t need humans to do it. We have optimised and configured the production industry so it can work by itself.
Machines don’t ''think'', and they don’t “''hallucinate''”. Hallucinating is a special, [[I am a Strange Loop|strangely-loopy]] phenomenon. Generative models don’t “see things that aren’t there”: they generate text based on statistical patterns in their training data. When they produce incorrect information, it’s due to limitations in their training data and architecture, not because they're experiencing false perceptions like a human might. No one yet has explained how human perceptions work, but we remain confident that machines don’t perceive and they are not self-aware.
The service industry is going through a similar process. Where it can it will eliminate expensive, manual processes that can be done automatically.  
 
Where this can be done easily, it has been: deltaview. Email obviates the fax room.
But the converse is just as important: we ''really'' shouldn’t ''evaluate'' humans by standards suited to machines — we shouldn’t ''robomorphise'' ourselves, in other words.  
But matters requiring human interaction have been harder to fully automate. The typical reactions have been to triage — delay the intervention of a human as long as possible — or to “self serve” — a rather cheeky means of getting the consumer to do part of the work, or absorb its expense — for you. Ryanair are masters of this — they even charge customers for getting their booking wrong!l (penalising humans for their own human frailty is the ultimate in chutzpah) but it happens within and without organisations.  
 
Internet enabled outsorurcing service to the customer or consumer. Typing pool!
{{quote|“To a man with a computer, everything looks like a computer.”<Ref>[[Neil Postman]]’s similar observation: “To a man with a computer, everything looks like ''data''.”</ref>
:— JC}}
 
If we benchmark humans against computers, we will lose. Machines do predictable things more easily, cheaply and quickly than we can. They always have done. They work best in constrained, predictable environments configured to eliminate [[uncertainty]] and minimise waste.
 
====Why humans don’t work in manufacturing any more ====
{{Drop|I|n the west}}, humans have been largely absent from production lines for decades — hence the colossal pivot to [[bullshit job|service industries]]. It isn’t that we ''don’t'' make things any more: it’s just that, usually, we don’t need ''humans'' to do it. We have optimised and configured the production to work by itself.
 
The service industry underwent a similar process as it ballooned with refugees from manufacturing. We automate expensive, manual processes wherever we can.  
 
But matters requiring human intervention — judgment — have been harder to fully automate. The typical reactions have been to [[triage]] — delay human intervention as long as possible — or to “self-serve” — a rather cheeky means of getting the consumer to do part of the work, or absorb its expense — for you.  
 
Ryanair is the master of this — it even charges customers for getting their booking wrong!  
(penalising humans for their own human frailty is the ultimate in chutzpah) but it happens within and without organisations.  
 
Internet enabled outsourcing service to the customer or consumer. Typing pool!

Latest revision as of 10:21, 11 November 2024

“If to call a man a wolf is to put him in a special light, we must not forget that the metaphor makes the wolf seem more human than he otherwise would.”

—Max Black, Metaphor (1965)

Robomorphism
/rəʊbəʊˈmɔːfɪz(ə)m/ (n.)


The interpretation of human behaviour, activity, or interactions in terms better suited to a Turing machine.

Recently, Matt Bradley made an interesting point about our gallop towards AI: whatever we do — however tempting it is — we should be careful not to anthropomorphise when we talk about machines.

Machines don’t think, and they don’t “hallucinate”. Hallucinating is a special, strangely-loopy phenomenon. Generative models don’t “see things that aren’t there”: they generate text based on statistical patterns in their training data. When they produce incorrect information, it’s due to limitations in their training data and architecture, not because they're experiencing false perceptions like a human might. No one yet has explained how human perceptions work, but we remain confident that machines don’t perceive and they are not self-aware.

But the converse is just as important: we really shouldn’t evaluate humans by standards suited to machines — we shouldn’t robomorphise ourselves, in other words.

“To a man with a computer, everything looks like a computer.”[1]

— JC

If we benchmark humans against computers, we will lose. Machines do predictable things more easily, cheaply and quickly than we can. They always have done. They work best in constrained, predictable environments configured to eliminate uncertainty and minimise waste.

Why humans don’t work in manufacturing any more

In the west, humans have been largely absent from production lines for decades — hence the colossal pivot to service industries. It isn’t that we don’t make things any more: it’s just that, usually, we don’t need humans to do it. We have optimised and configured the production to work by itself.

The service industry underwent a similar process as it ballooned with refugees from manufacturing. We automate expensive, manual processes wherever we can.

But matters requiring human intervention — judgment — have been harder to fully automate. The typical reactions have been to triage — delay human intervention as long as possible — or to “self-serve” — a rather cheeky means of getting the consumer to do part of the work, or absorb its expense — for you.

Ryanair is the master of this — it even charges customers for getting their booking wrong! (penalising humans for their own human frailty is the ultimate in chutzpah) but it happens within and without organisations.

Internet enabled outsourcing service to the customer or consumer. Typing pool!

  1. ↑ Neil Postman’s similar observation: “To a man with a computer, everything looks like data.”
Retrieved from "https://jollycontrarian.com/index.php?title=Template:M_intro_technology_robomorphism&oldid=92098"