BitBender
Carl T. BergstromThe New York Times just discovered parallel computing.
Matt Blaze@ct_bergstrom That is impressively wrong.
Rob@mattblaze @ct_bergstrom a fractal of incorrect
Going T. Maine
Ignazio Palmisano@simrob @mattblaze @ct_bergstrom it's wrong in multiple ways at the same time, so in their world it's AI
Finitum@ignaziop1977 @simrob @mattblaze @ct_bergstrom is the opposite of artificial intelligence real dumb?
bewilderbeast23
Sophie Schmieg@mattblaze @ct_bergstrom I legit thought it was some misunderstanding of quantum computing at first, but then the title says "AI computing" and I got nothing.
(And to be clear, quantum computing also does not work like that)
@sophieschmieg @ct_bergstrom Yeah, I can't even figure out what they got wrong. It's that wrong.
Robin Adams
Callisto
David Mankins@mattblaze @sophieschmieg @ct_bergstrom I’m guessing someone told them at a high level how GPUs do their thing and they mapped “GPU” to “AI”.
wrosecrans@lain_7 @mattblaze @sophieschmieg @ct_bergstrom
For anybody not interested in graphics, Nvidia brands itself as an "AI Chip" company these days. GPU's are parallel. AI runs on GPU's. GPU's are "AI Chips" these days. So "AI compute" is compute running on parallel hardware.
This makes perfect sense as long as you never heard of "computation" before six months ago, and everything you know about computers is from vapid press releases hyping AI scams.
derptron@wrosecrans @lain_7 @mattblaze @sophieschmieg @ct_bergstrom A lot of those chips are integrated GPU chips. It's both a regular CPU and a GPU. They can do a lot, but yeah AI is usually what they are used for.
The Jensen Orin is an example of this.
One of the ideas for that thing is to use AI to analyze brain waves and control your limbs if you've got some injury that severed something in your spine. You'd wear something like an Orin among other things.
Joby 
(he/him)@mattblaze @sophieschmieg @ct_bergstrom My theory is that somebody was trying to explain to them why massively parallel computing is better for neural nets, and they heard it as we'd invented the whole concept just for AI, and just extrapolated all wrong from there.
Chilly 
🛡️ 
💀 𝓕airchild 💀@mattblaze @sophieschmieg @ct_bergstrom I _think_ they're comparing a single-threaded CPU to a modern GPU??? Which is nonsensical. We've had multiprocessing for decades.
As Pauli might have said, this is not only not right, it is not even wrong.
@tankgrrl @sophieschmieg @ct_bergstrom Right. The part I struggle with is how you can know enough to describe parallel processing reasonably well, but not know enough to know that AI and parallel processing are completely different concepts.
Björn ❄️ StarkimarmIt's a case of "so wrong, even the opposite is not correct". F for failed the topic of the assignment.
bws
➴➴➴Æ🜔Ɲ.Ƈꭚ⍴𝔥єɼ👩🏻💻@mattblaze @ct_bergstrom I think it actually may qualify as "not even wrong".
Norman Wilson@mattblaze @ct_bergstrom Multiple concurrent wrongs.
Mike (no not that one)@mattblaze @ct_bergstrom It's wrong. At scale!
Julio J. 🀲@ct_bergstrom did an AI make that infographic?
Jonas Geiler@ct_bergstrom Why do you need AI computing for that, wtf
Icil
Weltzeitgeist@ct_bergstrom Is this reality?
Jim Kreft@ct_bergstrom it would be fun to see how many times this diagram has been created over the years for the latest way to parallelize things.
Earthshine --> moved@ct_bergstrom Good God, SMP is "AI" now too?
bomberfish@ct_bergstrom oh my god 
echopapaich habe Schmerzen, ziemliche.
Vor 40 Jahren habe ich 30 Jahre älteren Menschen erklärt wie ein Computer funktioniert. Und heute erkläre ich 30 Jahre jüngeren Menschen wie ein Computer funktioniert.
Oder irgendwie so
Digital naives, grmpfl
Armin Hanisch@echopapa Journalisten… 🙄
WWSchoof 📯
plague round@echopapa also war deine beste Zeit, Gleichaltrigen Computer zu erklären, vor 20 Jahren?
Die Zeit hat nie geendet, seit ich mit dem ganzen Digital-Krempel angefangen habe. Das Unwissen ist bei den Gleichaltrigen ist weiterhin erschreckend.
Florian Schmidt@echopapa
Und das hat damit zu tun, dass diejenigen, die Computer verstehen nicht etwa hochbezahlt in die Schulen geschickt wurden, um sie allen zu erklären.
Nein, sie wurden hochbezahlt in Unternehmen geschickt, um hinter einer Armee aus Sales, Marketing, Support und Anwälten die Computer für Dumme zu programmieren, und diese möglichst dumm zu halten.
Sorry, geht gleich wieder. Ich gehe jetzt wieder Leuten Computer erklären...
@ct_bergstrom
Und das hat damit zu tun, dass diejenigen, die Computer verstehen nicht etwa hochbezahlt in die Schulen geschickt wurden, um sie allen zu erklären.
Nein, sie wurden hochbezahlt in Unternehmen geschickt, um hinter einer Armee aus Sales, Marketing, Support und Anwälten die Computer für Dumme zu programmieren, und diese möglichst dumm zu halten.
Sorry, geht gleich wieder. Ich gehe jetzt wieder Leuten Computer erklären...
@ct_bergstrom
Angela Scholder@ct_bergstrom Wasn't that invented in the 60's by Lynn Conway when she worked for IBM?
Parallel and out of sequence that was.
Before she was kicked out for being herself.
g_teley@ct_bergstrom
My God! The Ada language with its tasks and rendezvous system already provided for reliable and efficient parallel processing 45 years ago!
Everything needs to be AI nowadays, I guess.
My God! The Ada language with its tasks and rendezvous system already provided for reliable and efficient parallel processing 45 years ago!
Everything needs to be AI nowadays, I guess.
Elias Mårtenson@g_teley @ct_bergstrom I thought about the Connection Machine back in 1986: https://en.wikipedia.org/wiki/Connection_Machine
@loke @ct_bergstrom
Yes! I remember that one. Specifically because of those 64k 1-bit processors and a whopping 512Mb of internal memory.
At that time, I had a 64Kb BBC-B 8-bit computer. 🙈
Yes! I remember that one. Specifically because of those 64k 1-bit processors and a whopping 512Mb of internal memory.
At that time, I had a 64Kb BBC-B 8-bit computer. 🙈
SK53@ct_bergstrom we were talking about "farm parallelism" 40 years ago, but my boss at the time had designed an array processor (SIMD) for a CT scanner about a decade before that. But as a computational technique I imagine it existed well before digital computers.
At least as far back as 1922:
https://thatsmaths.com/2016/01/07/richardsons-fantastic-forecast-factory/
Weather prediction using a gigantic staff of human computers, each calculating the equations describing the weather system for their own region of the globe.
@lain_7 @ct_bergstrom Great example. I had the "Harvard Computers" and compilers of mathematical tables in mind when I made the observation.
Ms Ngo@ct_bergstrom When you see them write about something you know & get it confidently & spectacularly wrong, remember they also write with similar confidence about things you don't know very well.
Elias Probst@causticmsngo
"The Gell-Mann amnesia effect is a cognitive bias describing the tendency of individuals to critically assess media reports in a domain they are knowledgeable about, yet continue to trust reporting in other areas despite recognizing similar potential inaccuracies."
Kura 
:kura_explode:@ct_bergstrom no. It's ai computing!
Daniel@ct_bergstrom Unbelievable what this AI technology can do.. It even disables features
@sunscheinwerfer @ct_bergstrom Is this real?
AdamIt sure is real
Fedi has also compiled a helpful list of remedies, see for example:
TiredRaven@sunscheinwerfer @ct_bergstrom I guess Alexa's users should decide to stop using Alexa ¯\_(ツ)_/¯
@sunscheinwerfer @ct_bergstrom and if I had that particular piece of spy hardware around the house, this'd be the point where I look for an electrical devices recycling point.
meejah@ct_bergstrom don't you mean "AI computing"??
(Did an "A" "I" draw this diagram??)
Fedo ¶
The Problem With Science Communication
@ct_bergstrom Parallel computing never existed before AI! /s
@ct_bergstrom I can’t help but think this is the outcome of a game-of-telephone applied to a description of how GPUs work and are used in the calculations underlying ML.
Exandra@ct_bergstrom @xgranade wow, that's… like the opposite of what happens.
Traditional computing; multiple processors efficiently compute distinct and useful tasks in parallel, quickly delivering accurate results or effects
AI computing:
(Sound of lake being slurped into fiery crack in ground)
Stefan Ritter@ct_bergstrom Woah, that's how my brain works.
Jim Jones@ct_bergstrom so disappointing.
I had an undergrad software engineering course that was all about optimizing serially processed programs we were given by the professor to run on chips that made use of parallel processing. This was in the mid 90s.
There was a lot of loop unrolling and loop jamming and other parallel processing techniques that had been known for a long time. Our grade was based on how fast our final programs would run.
No AI, just a lot of analytics and tinkering.