(C) Daily Kos
This story was originally published by Daily Kos and is unaltered.
. . . . . . . . . .
The Ten Nanofarad Diode [1]
['This Content Is Not Subject To Review Daily Kos Staff Prior To Publication.']
Date: 2025-07-24
This diary was prompted by a post on the Mastodon Fediverse, illustrating what happened when someone asked an “AI” for a design for an “autofire circuit.”
Unfortunately, in order for the average reader to appreciate what’s being talked about, I first need to give you a very basic introduction to some electronic components, what they do, and how they’re measured/rated.
Pre-EE 101
This is super-basic, and leaves a ton of stuff out, but should give you just barely enough to get the punch line.
Diodes allow electricity to flow in one direction, but not the other. Diodes are rated by the maximum amount of current they can pass, their forward break-over voltage (how much forward voltage you need to apply before electricity will begin flowing), and the maximum reverse voltage they can withstand before they self-destruct. And if you’re thinking, “Wait… LED… Light Emitting Diode… Do LEDs behave that way, too?” Yes, they do. An LED will light up if you apply current in one direction, but not the other.
Resistors are just what their name implies — they resist electrical current flow, and they do so equally for both constant and changing electrical signals. Resistance is measured in Ohms.
Inductors are essentially coils of wire that consequently generate an electromagnetic field. Because of how this field interacts with the wire generating it, they resist (or “impede”) rapid changes in current flow. The more rapid the change, the more the apparent impedance. Inductance is measured in Henrys.
Capacitors store an electric charge for later use, and are kind of the opposite of inductors. They permit rapid changes in current flow — the more rapid the change, the lower the apparent impedance. Capacitance is measured in Farads. However, this unit is gargantuan. Capacitors most commonly appear with ratings of microfarads, nanofarads, and picofarads (millionths, billionths, and trillionths of a Farad, respectively).
“Draw The Rest of The Horse”
Got that? Okay. Now, have a look at this. This small electrical circuit was created by something being marketed as a “generative AI,” and which the user is presumably paying money for:
If you’re now thinking to yourself, “I don’t get it,” congratulations, you have arrived at the same conclusion as every EE graduate and electronics hobbyist. We have a transistor (“1N4148”, which is actually a designation for a diode) where one of the three legs isn’t connected to anything. The “DE-9” connector (not its proper name), in addition to being unnecessary and not actually connected to anything, has “8”s in place of some of the pin locations. The 555 timer (large rectangle on the left), a chip which has only one RESET pin, is here shown with two. The 1k (1,000 Ohms) resistor at the top of the 555 has only one of its two legs connected.
But the dead giveaway, the thing that prompts one’s jaw to drop rapidly to the floor, the thing so preposterous that I felt compelled to relate this story, is the component under 555 chip — the little black arrow with a horizontal line through it.
That’s a diode.
The annotation next to it describes its rating as “10 nF.”
Ten nanofarads.
For a diode.
If this were a joke along the lines of the turbo encabulator, it would be mildly funny, and perhaps amuse some EEs as they tried to work out what it actually does. However, there’s a far deeper issue at play here, which I think this example of LLM hallucination makes starkly clear: There is no point in trying to understand what the circuit does, or what the LLM was trying to do. It is worse than a flawed design. It is no design at all. It is completely ignorant of electrical and electronic principles. It is not a circuit. It is a child’s drawing, imitating a half-remembered glimpse of a real circuit they saw months ago… Maybe.
Unfortunately, the clarity and obviousness of this complete failure is only evident to people with domain-specific knowledge. To anyone else... “Well, it looks like other circuits I remember from that high-school science book...”
A Complete 180?
In 1994, a bug was discovered in the Intel Pentium floating point unit — the infamous FDIV bug, that caused about one in 9,000,000,000 calculations to be slightly off. After trying to cover it up, Intel issued a massive recall costing them half a billion dollars. Today in 2025, these LLMs, frequently generating obviously false statements, and just as frequently failing basic arithmetic, are the hottest thing in tech.
What the hell happened?
It astonishes me that so many people who are otherwise not-dumb are incorporating these bullshit generators into their lives, and the lives of their employees, users, and customers. They insist, “Don’t worry; they’ll get better.” But unless a fundamentally new approach is developed, LLMs will remain parrots, chundering up phrases and sentence fragments and computer code and now, apparently, electronic circuits with no understanding of their meaning, much less their impact on the reader.
I fear this will not end well.
[END]
---
[1] Url:
https://www.dailykos.com/stories/2025/7/24/2334910/-The-Ten-Nanofarad-Diode?pm_campaign=front_page&pm_source=trending&pm_medium=web
Published and (C) by Daily Kos
Content appears here under this condition or license: Site content may be used for any purpose without permission unless otherwise specified.
via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/dailykos/
