AI Drools, PI Rules
Techno Talk
Have you heard of pseudo-intelligence (PI)? Over the past few
months, I’ve been introduced to a smorgasbord of PI-enabled tools
and technologies, some of which I’d like to share with you here.
I hail from Yorkshire, which,
for our international readers, is widely
acknowledged to be the best county
in England. (Don’t listen to anyone
from other counties who says otherwise; they are foolish folk and you
cannot trust them.)
Increasingly, I find myself identifying with the characters in Monty
Python’s classic Four Yorkshiremen
sketch (you can watch the video at
https://youtu.be/ue7wM0QC5LE).
Especially the part at the end where
they say, “…you try an’ tell the young
people of today that, and they won’t
believe you…”!
When I tell young engineers that I
spent my formative years as an engineer in a world without the internet,
they look at me as though I’m a time
traveller who has just dropped in
from the Victorian era.
A wealth of information
I can hardly believe the wealth of
information that’s now available at our
fingertips. These days, for example, I
can locate and access data sheets for
virtually any electronic component
on the web in a matter of seconds.
Things weren’t always this way.
Prior to the introduction of the
internet, engineers maintained their
own mini-libraries in their offices.
If you were lucky, the company you
worked for might boast its own library. Often, you were required to
contact local distributors and request
that they share data books and application notes for components of
interest.
Even after the internet impinged
itself on our collective consciousness circa the early 1990s, there was
little information of use on it. Even
when useful information did begin to
wend its way onto the system circa
the mid-1990s, that information was
not easy to find.
Yahoo started life as “Jerry and
David’s Guide to the World Wide Web”
in 1994, but it was rather ropy in the
early days. Google, which began as a
search engine called BackRub in 1996,
was significantly better, but nowhere
near as impressive as it is today.
An engineers’ delight
All the above leads us to 1998,
when a young engineer called Randy
Sargent became frustrated with the
time and effort he was devoting to
rifling through component vendors’
websites, desperately trying to glean
Max the Magnificent
information on devices for use in his
designs.
Eventually, Randy implemented
an engineering hack. He created a
Perl script to search the internet for
him. This script could scrape component vendors’ websites, extract
the desired information, and present it to Randy in a form that was
useful to him.
Randy’s script was so successful
that it evolved into FindChips (www.
findchips.com), which, in addition
to being both easy and free to use, is
one of the most helpful component
search engines I am aware of.
Pseudo intelligence
Our esteemed editor at Practical
Electronics has informed me that he’s
tired of everyone discussing artificial
intelligence (AI) all the time. You can
only imagine my despondency when
I discovered that I would have to find
something else to write about!
Well, “you can’t fight the man”, as
they say. Instead, I wish to share some
interesting information about something we shall call pseudo-intelligence
(PI) which, by some strange quirk of
fate, encompasses all aspects of AI
(and then some).
Desperately divining data before the internet age
4
Practical Electronics | August | 2025
Techno Talk
Max the Magnificent
PI-enabled software design tools
On the software side, we have a code
completion and generation tool called
Copilot, developed by GitHub (github.
com) and the creators of ChatGPT.
This is complemented by a code
verification tool called Metabob, developed by the company Metabob
(metabob.com).
Additionally, there is a documentation generator called Driver,
developed by a company of the
same name (driver.ai). This can examine the software aspects of your
product, including software development kits (SDKs) and application
programming interfaces (APIs), and
automatically generate user documentation that you’d be proud to
show your mother.
When it comes to adding PI capabilities to your embedded system, a
company called DeGirum (degirum.
ai) offers a “model zoo” of PI models
for various applications, including
people detection, pose detection,
face detection, age estimation, gender
classification, emotion classification,
gesture recognition and more.
You can quickly and easily combine and deploy these models in
your system.
PI-enabled hardware design
Let’s start with the circuit and
schematic portion of the design process. In this case, three companies
immediately come to mind: Celus
(celus.io), Circuit Mind (circuitmind.
io), and Flux (flux.ai).
Celus and Flux both feature their
own generative PI copilots. These
allow you to do all sorts of things,
like using a natural language interface to say what you want to
do (“Create a WiFi-enabled thermostat using a temperature sensor
and an 8-bit microcontroller”) and
receive a corresponding schematic
in seconds.
You can also use a drag-and-drop
interface to create a block diagram
of your desired system (eg, a sensor,
microcontroller, or motor driver),
attach attributes to the blocks (eg,
an I 2 C interface, 16-bit data bus,
or 3.3V supply) and – once again
– receive a detailed schematic, including all ancillary components
like resistors, capacitors and inductors, in seconds.
Celus recently added the ability to
input an image of a block diagram
that you created on the back of an
envelope or napkin, for example,
and use it as the starting point for
the design.
Circuit Mind takes a different
approach, utilising PI to read component datasheets and extract the
necessary information to drive its
deterministic design algorithms.
The folks at Circuit Mind recently
added the ability to automatically
generate the power supply portion
of the system.
PI-enabled PCB layout tools are also
starting to make their presence felt.
Three companies of interest in this
arena are AutoPCB (autopcb.app),
JITX (jitx.com) and Quilter (quilter.
ai). I hope to speak with the founders of all these entities in the next
few weeks, and I’ll report further in
a future column.
PI-powered PCBA verification
I was just chatting with Eyal Weiss,
a man who played a crucial role in
developing a new technology. Eyal
ultimately secured a billion-dollar
deal. He was, let’s say, ‘sad’ when
thousands of circuit boards started
to fail in the field. It turned out that
the contract manufacturer had inadvertently used 10-year-old counterfeit
capacitors, so all was nearly lost for
the sake of a couple of cents’ worth
of component.
Vowing that this would never
happen to him again, Eyal founded Cybord (cybord.ai). Its claim to
fame is that, when installed in PCB
assembly facilities, it can detect
old components, counterfeit components, substituted components,
added components and subtracted
components, among many other
anomalies. I don’t know about you,
but I prefer my PCBs anomaly-free!
The exciting thing is that everything discussed in this column is just
a tempting teaser for what is to come
PE
in the not-so-distant future!
Spot the difference: can you tell what differences Cybord found between the image of the reference board (left) and the
later production variant (right)? Turn to page 43 for the answers.
Practical Electronics | August | 2025
5