5

What was the hardware that makes up HAL9000 in the spaceship Discovery One, from Arthur C Clarke's 2001 series?

Wikipedia says that HAL became operational on 12 January 1997 and the books say 1992 and 1991. This would have been the age of the desktop PentiumMMX, and the 486 respectively.

The question was inspired by the last paragraph in this answer on Space Exploration.

I suspect this question cannot be answered, we can only conjecture.

  • 2
    It is not mentioned in the movie or the novel. However it is mentioned that HAL's 'brain' was grown in such a way that it was analogous to the development of a human brain. – bobbyalex Jul 6 '16 at 4:39
  • 1
    And considering the original story was written in 1968, there probably is no good or canon answer. – Criggie Jul 6 '16 at 5:21
  • 1
    How many "FLOPS" did Arthur C Clarke surmise that HAL would've had to been capable of? He was way into computer science; I'm sure there's a direct quote from him somewhere that can sort of answer this. – Mazura Jul 6 '16 at 5:44
  • As per current events, "Juno is controlled by a 200 megahertz RAD750 computer with 128 megabytes of memory and 256 megabytes of flash storage. That's the equivalent of a low cost personal computer from 1999. There's not much room in that rather limited system for very much sophistication." This being the autonomous space probe that has entered Jupiter Orbit on July 5 2016. – Criggie Jul 6 '16 at 8:00
  • To parrot the answers in the linked question: my iPhone's alarm clock is "autonomous" - except that, I told it to do that. As I elude to in my answer, it is the point at which the system (the software) becomes 'self-aware'... that is the crux of true A.I. – Mazura Jul 6 '16 at 8:10
5

Here's the bare minimum specs: Deep Blue (the first computer to beat a reigning grand master at chess)

  • computer class: mainframe

  • architecture: parallel, based on SP2 technology

  • 32 Power2 SC CPUs (135MHz, 32-bit registers, RISC, 15 million transistors)

  • 512 Chess processors

  • One trillion operations per second (500,000,000 for a high-end PC in 1999)

  • 500 computation threads, 1.2 m2 footprint


(Deep Blue – Wikipedia)

Have today's desktop PCs surpassed IBM's Deep Blue of 1997? (YES) – Super User.SE

Processing performance [of Deep Blue] was 11.38 GFLOPS

Keeping in mind that a modern video card, e.g., the

AMD Radeon R800 can achieve 3.04 TFLOPS

... it's the software that's still science fiction, not the hardware.

  • 1
    Info graphics are fun: Deep Blue > Blue Gene > Watson. Realistically, HAL would have to be 'smarter' than Watson. – Mazura Jul 6 '16 at 7:14
  • Note my analysis and references, under-appreciated on a sister site. – JDługosz Jul 6 '16 at 19:17
  • Compare your answer to the modern CPU noted here. 32 flops per cycle per core equals 32×4×3.8G makes about 480 Gflops (half that for double precision). But how is floating point arithmetic speed relevant to this problem? The figure of merit might be about connectivity and mips. – JDługosz Jul 6 '16 at 19:27
  • @Mazura Your ">" are going the wrong way? Watson is a stronger computer than Blue Gene or Deep Blue. – Shane Jul 6 '16 at 21:50
  • @Shane - I don't know how to make a Unicode yield sign. They're just arrows denoting the progression, like in the info graph. – Mazura Jul 6 '16 at 22:26

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .