8 ms·

Great! I was trying to remember the model name of this processor the other day in a conversation, and no amount of googling could bring it up, all my results w

Great! I was trying to remember the model name of this processor the other day in a conversation, and no amount of googling could bring it up, all my results were swamped by all the new AI-du-jour chips everyone is working on. The young person I was talking to could not believe there was a neural network chip in the 90s and looked at me like I was having some kind of memory issue. I remember the 80170 just not getting that much attention when it came out, at the time I was just out of high school, hammering away on Turbo C++ and working through the third volume of Parallel Distributed Processing and thinking to myself, "someone should make a chip that does this" and then eventually reading about the 80170 having come and gone. I vaguely recall a demo where they would quickly flash NIST digits at the chip and it could recognize them with high speed and accuracy. I think what maybe killed the effort was the extreme pace of Moore's law at the time, made a NN chip more efficient than software? No biggie in 8 months we'll just double the density and speed of the current x86 chip and beat it in software. There's probably also the extreme internal competition in Intel to kill anything that might even vaguely threaten x86. Anyway, good flashbacks. I should dig up my copy of PDP, I remember skimming over the second volume on biological models and I think at this stage in my life that might be the most interesting.

mtillman 3y ago
Whenever I run a training session on the history of ML people are shocked to learn when the maths were first developed for things that are reported as new on TechCrunch.
cyanydeez 3y ago
this of course demonstrates the limit of LLMs. without some kind of temporal/spatial knobs, they'll overwrite similar but distinctly different datasets.
ploxiln 3y ago
> extreme internal competition in Intel to kill anything that might even vaguely threaten x86 Actually, Intel has tried pretty hard to kill x86, multiple times, and failed. Most recently Itanium, but before that iAPX-432, and a couple others. New grand architectures to replace crusty old x86. But IMHO they were too academic. As you suggest, when an x86 is cheaper _and faster_ and compatible with existing software ... good luck replacing that.
- kjs3 3y ago
  Not sure 'academic' is the right word; IMHO it's a market issue. Users care about software, not architectures. This business is littered with the corpses of companies with 'better' in every way architectures, when the vast majority of folks that write checks are saying 'yeah, cool, but I just want it run Excel the same way my current one does but a little faster'.
  - ploxiln 3y ago
    "academic" as in "theoretically we can have object-oriented instructions in the CPU!" ... "theoretically the compiler can do all the instruction scheduling!" ... in practice oh please no
- Kon-Peki 3y ago
  iAPX-432 was supposed to be the last ISA Intel ever made, but it didn't quite work out that way [1]: > Gordon Moore believed Intel’s next ISA would last the lifetime of Intel, so he hired many clever computer science Ph.D.’s and sent them to a new facility in Portland to invent the next great ISA. The 8800, as Intel originally named it... > This ambitious project was alas several years late, forcing Intel to start an emergency replacement effort in Santa Clara to deliver a 16-bit microprocessor in 1979. Intel gave the new team 52 weeks to develop the new “8086” ISA and design and build the chip. Given the tight schedule, designing the ISA took only 10 person-weeks over three regular calendar weeks... > Intel’s original 8800 project was renamed iAPX-432 and finally announced in 1981... [1] https://dl.acm.org/doi/pdf/10.1145/3282307 https://dl.acm.org/doi/pdf/10.1145/3282307
  - a1369209993 3y ago
    > Gordon Moore believed []Intel's next ISA[] would last the lifetime of Intel > deliver a 16-bit microprocessor in 1979. [...] the new "8086" ISA > Intel's original 8800 project was renamed iAPX-432 and finally announced in 1981 I mean, he wasn't wrong...
    - Kon-Peki 3y ago
UncleOxidant 3y ago
> The young person I was talking to could not believe there was a neural network chip in the 90s This wasn't the only one, there were others. There was a startup here in Oregon called Adaptive Solutions that was working on a NN chip called CNAPS[1] [1] https://proceedings.neurips.cc/paper/1990/file/beed13602b9b0e6ecb5b568ff5058f07-Paper.pdf https://proceedings.neurips.cc/paper/1990/file/beed13602b9b0...
aidenn0 3y ago
> Great! I was trying to remember the model name of this processor the other day in a conversation, and no amount of googling could bring it up, all my results were swamped by all the new AI-du-jour chips everyone is working on. The young person I was talking to could not believe there was a neural network chip in the 90s and looked at me like I was having some kind of memory issue. And if you move from "chip" to "special purpose hardware" you can go back to 1958 with the Mark I Perceptron (Assuming "neural net" requires a hidden layer, it qualifies).

mtillman 3y ago

Whenever I run a training session on the history of ML people are shocked to learn when the maths were first developed for things that are reported as new on TechCrunch.

cyanydeez 3y ago

this of course demonstrates the limit of LLMs. without some kind of temporal/spatial knobs, they'll overwrite similar but distinctly different datasets.

ploxiln 3y ago

> extreme internal competition in Intel to kill anything that might even vaguely threaten x86 Actually, Intel has tried pretty hard to kill x86, multiple times, and failed. Most recently Itanium, but before that iAPX-432, and a couple others. New grand architectures to replace crusty old x86. But IMHO they were too academic. As you suggest, when an x86 is cheaper _and faster_ and compatible with existing software ... good luck replacing that.

kjs3 3y ago
Not sure 'academic' is the right word; IMHO it's a market issue. Users care about software, not architectures. This business is littered with the corpses of companies with 'better' in every way architectures, when the vast majority of folks that write checks are saying 'yeah, cool, but I just want it run Excel the same way my current one does but a little faster'.
- ploxiln 3y ago
  "academic" as in "theoretically we can have object-oriented instructions in the CPU!" ... "theoretically the compiler can do all the instruction scheduling!" ... in practice oh please no
Kon-Peki 3y ago
iAPX-432 was supposed to be the last ISA Intel ever made, but it didn't quite work out that way [1]: > Gordon Moore believed Intel’s next ISA would last the lifetime of Intel, so he hired many clever computer science Ph.D.’s and sent them to a new facility in Portland to invent the next great ISA. The 8800, as Intel originally named it... > This ambitious project was alas several years late, forcing Intel to start an emergency replacement effort in Santa Clara to deliver a 16-bit microprocessor in 1979. Intel gave the new team 52 weeks to develop the new “8086” ISA and design and build the chip. Given the tight schedule, designing the ISA took only 10 person-weeks over three regular calendar weeks... > Intel’s original 8800 project was renamed iAPX-432 and finally announced in 1981... [1] https://dl.acm.org/doi/pdf/10.1145/3282307 https://dl.acm.org/doi/pdf/10.1145/3282307
- a1369209993 3y ago
  > Gordon Moore believed []Intel's next ISA[] would last the lifetime of Intel > deliver a 16-bit microprocessor in 1979. [...] the new "8086" ISA > Intel's original 8800 project was renamed iAPX-432 and finally announced in 1981 I mean, he wasn't wrong...
  - Kon-Peki 3y ago

UncleOxidant 3y ago

> The young person I was talking to could not believe there was a neural network chip in the 90s This wasn't the only one, there were others. There was a startup here in Oregon called Adaptive Solutions that was working on a NN chip called CNAPS[1] [1] https://proceedings.neurips.cc/paper/1990/file/beed13602b9b0e6ecb5b568ff5058f07-Paper.pdf https://proceedings.neurips.cc/paper/1990/file/beed13602b9b0...

aidenn0 3y ago

> Great! I was trying to remember the model name of this processor the other day in a conversation, and no amount of googling could bring it up, all my results were swamped by all the new AI-du-jour chips everyone is working on. The young person I was talking to could not believe there was a neural network chip in the 90s and looked at me like I was having some kind of memory issue. And if you move from "chip" to "special purpose hardware" you can go back to 1958 with the Mark I Perceptron (Assuming "neural net" requires a hidden layer, it qualifies).