"Later he became a developer of one of the first machine learning algorithms, techniques for character and image recognition."

"The AKAT-1 was the world's first transistor differential analyzer, constructed by Jacek Karpiński at the Polish Academy of Science's Institute of Automatics in 1959. It was designed to solve systems of differential equations and modeling processes."

https://en.wikipedia.org/wiki/Jacek_Karpi%C5%84ski

So good-looking too:

(Photo by Topory, on wikimedia)

It's an analogue computer - anyone here got any experience with those? There was a time when they were quite the thing.

I did one of my college projects on an EIA analog computer. I loved it. I solved van der Pohl's boundary layer problem on it. Although I don't have the connection diagram for the analog computer program, I still have the paper tape of the digital solution that I wrote in Basic+2 for the PDP11/70. Thirty-three years later it still emits the characteristic machine oil smell of paper tape media.

It depends on the definition. The analog "computers" I worked with weren't designed to solve a mathematician's equations. They were what you might call embedded analog computers.

In one category were some 1970's-vintage electronic photo-typesetters I used to service in the 80s and early 90s. These machines used what was essentially closed-circuit television to render an electronic document file into an optical image on film. This involved CRT-based scanning of characters from a font grid, simultaneous with reproduction of the characters onto film via another CRT. The machine's front end was a TTL-based mincomputer but all the scanning was analog and vector-based (ie; no pixels or raster in the usual sense). Most of the analog computation involved merely op-amps to scale the X and Y sawtooth sweep waveforms. But there were also analog multipliers for both axes, adjusted via trim-pots. IIRC these addressed issues of linearity and geometry correction for the magnetic deflection on the larger CRT.

As is common with analog circuitry, the results were only as good as the power supplies you were using. There was no tolerance for noise or out-of-spec voltage on the supply -- unlike a digital computer, which deals in quantized values and thus has a degree of inherent noise rejection.

The other category of gear wasn't something I serviced -- it was something I performed on! Not in public, mind you, but in the electronic music lab at Brock University. Here are photos of two Moog synthesizers -- the Mini-Moog , a portable unit with limited configuration options, and the Model 15, which used patch cords and was thus entirely flexible. In both cases the basic modules to be interconnected included VCO's, VCA's and VCF's (voltage-controlled oscillators, amplifiers and filters).

-- Jeff

(Here's the final image belonging to the previous post. It seems the forum software presently enforces a limit of three images per post.)