Hackaday readers might know [Victor Scheinman] as the pioneer who built some of the first practical robot arms. But what was a kid like that doing in high school? Thanks to a film about the 1958 New York City Science Fair, we know he was building a voice-activated typewriter. Don’t believe it? Watch it yourself below, thanks to [David Hoffman].
Ok, we know. Voice typing is no big deal today, and, frankly, [Victor’s] attempt isn’t going to amaze anyone today. But think about it. It was 1958! All those boat anchor ham radios behind him aren’t antiques. That’s what radios looked like in 1958. Plus, the kid is 16 years old. We’d say he did pretty darn good!
Not that he didn’t continue to do well. His project earned him a spot at MIT, worked for Boeing, NASA, and Stanford. You can read an interview from 2010 that sums up many of his experiences. In 2016, [Victor] passed away, but his legacy lives on not only in this newsreel footage but in robot arms today that still bear a striking resemblance to his MIT Arm.
The film is not very detailed about exactly how the typewriter worked. But it only recognized letters, and maybe not all 26 letters at that. We have a feeling that analog filters picked up the differences in a few letters (the COND BANK block on the chalkboard) and triggered a solenoid to strike the key. There are two amplifiers feeding that block from one microphone, so maybe there was a bit of a phase delay to pick up two frequencies. Or perhaps one was just an analog threshold trigger to figure out you were actually talking.
But we aren’t sure. In a eulogy written by [Victor’s] friend [Harvey Cohen], we can read:
For this purpose he adapted electromagnets to actuate the keys of a manual typewriter. The user spoke into a microphone, and the signal was classified (as A, B, C, … etc) by analogue circuity. This was in 1959… Understandably — this project won for Vic a Science Prize — and entry into MIT…In discussions I once had with Vic he reported the (analogue) algorithms he devised were in fact essentially the same as were then being developed in the early AI Labs.
So maybe there was more to it than we thought. But, regardless, it was quite an accomplishment.
A 2016 attempt at doing something similar has the benefit of nearly 60 years of progress behind it. If you don’t mind sharing your speech with Google or Amazon, such a project now is downright simple.
17 thoughts on “Retrotechtacular: Voice Controlled Typewriter Science Project In 1958”
Is that the same year Lee Felsenstein had a scince fair project? I can’t remember what, but there’s a photo, I think in Fire in the Valley.
That was before he was at Berkeley, and Community Memory.
Around a decade before Star Trek had the episode where Gary 7 zapped a typewriter with his sonic screwdriver knockoff and made it type by voice.
You mean “Around a decade before, Star Trek had the episode where Gary 7 zapped a typewriter with his sonic screwdriver knockoff and made it type by voice.”?
Because, without the comma it sounds as if it happened in the 1950s. ;)
Come on were’ eating grandma!
Come on, we’re eating, grandma!
Punctuation can save lifes. 😂
No, he was right the first time — this typewriter (1958) was about a decade before Star Trek (1966-’69). With the comma, he would have been saying Star Trek aired in the 1940s.
Thanks, I’ve already corrected that below. It was my bad, yes. I made a mistake here. A funny one, too.
No idea why that mistake happened, but I have an idea. 🙂
I think it was like this: Since the sentence started with “Around a decade before” rather than “It was around a decade before”, my mind unconsciously got this idea of the missing comma at exactly this point (position). Which in turn also changed the meaning. I hope that’s understandable, despite my poor English.
Long story short: My subconscious must have assumed that it a) wasn’t a complete sentence, thus, but rather a snippet of one. Hence, punctuation must have been missing at some point. And so it was looking for it.
Or b), that the stylistic writing must imply a side sentence that must follow soon, which would happen at exactly that spot (like I said before), but which also would change the relationship on a temporal basis. Or something like that. ^^
Anyway, it’s no big deal. But it’s interesting to see what thought patterns are involved in the process.
I think it also had to do with that I unconsciously assumed it happened in the 1970s.
But weirdly, while reading, I consciousnessly already knew at same time that this was before the 1970s. Weird. 🤷♂️
Just re-read the article carefully at home. So it happened in 1958. Okay, that would be indeed a decade before Starship Enterprise was on TV screen.
So never mind. 🙂
I’m still glad for having made the punctuation joke here, though. It deserves way more attention. Not all people write as nicely as you do. 👍
He never “zapped it”… he put a piece of paper in it turned it on and started dictating to it. It surprised his new secretary Roberta (Teri Garr) when it started typing everything she said. He never “zapped” it with his sonic screwdriver knockoff. This is not coming from Beta 5 computer “snobbery” ( if you know the episode you’ll get it ).
“The film is not very detailed about exactly how the typewriter worked. But it only recognized letters, and maybe not all 26 letters at that. ”
Makes me wonder if he already had figured out that glass fibre trick..
“Siri, tell me about your grandmother.”
The more interesting question is, “How did it work?” It would have been analog, of course.
My guess is that he used two or more audio filters. The vowel sounds are tones with various harmonics, so perhaps it compared the amplitude of the harmonics to decide which vowel it was.
Consonants are harder, because they are short impulses that would be difficult to analyze quickly.
I remember an electronic toy that used a simple micro and two filters (low-pass and high pass) to recognize a limited vocabulary. It was basically looking at the sequence of filter outputs. So “stop” would be “sss” (both), “T” (high only), “ooo” (low only), “P” (high-only). Relay logic might be good enough to decode the sequences to recognize spoken words.
The comment about “AI techniques” is almost certainly just referring to classification (i.e. looking at all the data he had on each sound and figuring out which pieces distinguished one word from another). He would have done it manually, whereas the early AI researchers were working on algorithms to do it automatically.
How does a young person in that era get access to such gadgetry? It must have been nice to have those kinds of resources.
By being kind and showing fascination?
If you’re modest and kind, but also persistent, there’s a chance that a young person gets support by its environment.
If people with the right equipment notice your true interest in their hobby/job, they take heart and share their equipment. Or they assist you in getting access to a similar one. Like, by introducing you witg the people who may help you further.
In the worst case, a person of a different, but related field will notice your fascination/interest and try to help you. This may take longer, but will eventually lead to the same goal.
That’s how I got a lot of stuff in the late 20th century.
People were flattered that a young person had interest in their hobby.
Perhaps it made them feeling younger, too.
Back in the 1950s, the situation was even better, perhaps.
There was surplus hardware from after the war.
With a little bit of assistance, could stuff could be built from retired hardware.
But that’s way before my time. I merely heard wondrous stories about these times. 🙂
Really nice National NC 300 receiver behind him. I’ve got one of those 😎
I noticed, but wasn’t sure if it was a 300 or 303.
The kid was well off.
I don’t think I ever found a picture of me with my 1958 science fair project “Felsnik” (the planetarium at the Franlkin Institute Science Museum was the gift of Samuel S. Fels and bore his name, so I could get away with sharing my name with his) which won the Third Division Award in the Delaware Valley Science Fair that year.
It was a joint project by me and my friend Murray Kaplan – a satellite built in a clear acrylic sphere that broadcast beeps to a nearby radio at a rate controlled by light falling on a selenium photocell. Murray got the electronics working by accident – I had designed a transistor multivibrator at the beep rate and he got it working somehow with sufficient amount of oscillation that the radio could pick it up.
Group projects were not permitted by the rules then and Murray told me he would be the silent partner. I got no prize other than a paper certificate and I waited for the publicity to roll in – some kid in Ohio got written up in our Scholastic Magazine with a satellite that didn’t even work! Oh, I was bitter, but it kept me entering science fairs all through college with stuff that didn’t work.
The picture is of me with Autocom – a switched intercom using analog techniques along with relays and motor-driven switches – not reliable and I could have used some proper engineering advice but I would never ask for help. As a science fair judge now I realize that not working was less important than analyzing the failures, writing them up, and making suggestions for improvement. But at 13 years old those points were too subtle for me.
How mach are they worth now?
Please be kind and respectful to help make the comments section excellent. (Comment Policy)