wrote:
On 17/09/2026 1:24 pm, Jeff Liebermann wrote:On Thu, 17 Sep 2026 02:02:09 +1000, Bill Sloman <bill.sloman@ieee.org>>
wrote:
On 17/09/2026 12:00 am, someone wrote:All the email services are doing this now, or soon will be. We live in a>
surveillance state.
>
Instead of saying the capacitor might explode, say the capacitor might
experience an uncontained failure event.
Even artificial intelligence should be smart enough to translate
euphemisms.
Let's see if that works. Feeding "uncontained failure event" to
ChatGPT results in:
<https://chatgpt.com/share/6aab58aa-9954-83e8-a8d7-3a8631c001b9>
"An uncontained failure event usually refers to an aircraft engine
failure in which internal engine components break apart and escape
through the engine casing."
I would call that a miss by the AI.
Probably not. AI depends on large language models which contain a huge
amount of text. The implication is that the text string "uncontained
failure event" showed up most often in documents about plane crashes.
>
It's not a great euphemism for an explosion. "Uncontained dispersion
event" might be better.
Nope. A capacitor explosion is subsonic and therefore should be
called a deflagration. If it were supersonic (unlikely), it would be
called a detonation.
"An uncontained dispersion event is an accidental release of a gas,
vapor, or hazardous material into the open atmosphere without a physical
barrier to stop it. The cloud moves freely, shaped by the wind and air.
This can cause flash fires or spread toxic pollution"
>
You might have to go to an "energetic Uncontained dispersion event"
That might also describe shaking a soda pop bottle or opening a
champagne bottle. For an old paper dielectric capacitor, replace soda
pop bottle with confetti generator. (Those were fun when I was in
high skool).
"An energetic uncontained dispersion event is a highly technical, formal
euphemism used primarily in aerospace, military, and engineering
contexts to describe a catastrophic explosion, fragmenting failure, or
blowout."
None of that correctly describes the original "capacitor might
explode".
Back to the AI topic. The problem I'm trying to describe is the
inability of the AI to read our mind and guess our intentions. As you
mentioned, when faced with multiple interpretations, the AI is most
likely to assume that the most common use case is the correct
interpretation. Misuse a phrase often enough, and it becomes the
default interpretation. This is further aggravated by an LLM database
that is polluted by spelling errors, verbiage, creative definitions,
jargon, slang, homonyms, etc.
<https://en.wikipedia.org/wiki/Homonym>
I consider it a minor miracle that an AI can extract a correct
interpretation in the presence of all such "noise".
--
Jeff Liebermann jeffl@cruzio.com
PO Box 272 http://www.LearnByDestroying.com
Ben Lomond CA 95005-0272 AE6KS 831-336-2558
Les messages affichés proviennent d'usenet.