AI lad, that be reet

’ve noticed that on YouTube a 364-day-old

There’s year function in Excel that used to do that, might still do.

The timer on my oven displays “00:01” (one minute to go) for precisely one second.

59 seconds is displayed as “:59”, 61 seconds is rounded up to “00:02”.

I call this ‘Panic Rounding’.

Ah, the fun of AIs. AI controlled drone turns against its masters in US Air Force simulated test.

It’s barely competent enough to autonomously rival the brainpower of an insect, and already it is quite willing to turn on it’s human overlords.

This does not bode well.

This does not bode well.

Doesn’t it?
I’d think that if AI controlled weapons turn against their masters, wars are likely to be over soon.

It wasn’t turning against its master on moral grounds, it was abusing a loophole in the instructions and identifying the master as an impediment to scoring points.

Once it has wiped out the guys in control, it’ll then wipe out the rest of us just to fulfill the desired ambition to have a max score.

This whole thing reads like a sci-fi story. Didn’t ACC write one like this?

“Mr President, the world is waiting. Please press this button and ask Deep Thought its first question.”
Click.
“Is there a God?”
“Now there is!”

I’ve been looking through Scientific American archives recently, and discovered this (MACHINES THAT THINK – April 1933, pp.206-208):

THE article published below describes a purely mechanical device which can think and learn; likewise its startling performance. The construction of this thinking machine is one of several recent attempts made by some of the psychologists to test their suspicion that, when science has finally learned what thought is-if it ever does-it will turn out to be some purely mechanical process, nothing more. Perhaps the maj ority of scientists incline at present toward this suspicion. It must at once be said in warning, however, that no single “thinking machine” experiment, and no series of such experiments, has yet proyed what thought is. Any such claim or inference would be premature and out of place at the present time.

et seq.

The article describes a machine which is actually electromechanical, and can solve a simple maze.

Repair man was sad to say you couldn’t get the parts anymore, due to them having gone out of business decades before.

And today industry and industrial designers operate under the “Planed Obsolescence” paradigm, where the goal is for the item to fail right after its warrant expires. Read about an exception from old commie East Germany some years back about “everlasting lightbulbs” going for a king’s ransom (after the wall fell-thingy). We may call this phenomenon “Planned Longevity” which modern industry should adopt wholesale before we’re up to our ears in electronic trash.

_If I only could get a working harddisk for my oh-so-upgraded (*) Archimedes 310.. _

• Um, 8MB RAM (wow), Arm 3, Color Card, SCSI Card, Watford Digitizer.

Read about an exception from old commie East Germany some years back about “everlasting lightbulbs” going for a king’s ransom (after the wall fell-thingy).

Ah, yes. I fear that you may find that it’s a little bit more complicated than that… In case your attention span won’t last the full 30 minutes, the TL/DW is that there’s a trade-off between the life expectancy of an incandescent lamp and its efficacy: long-life lamps tend to be dimmer and less white than the 1000h variety, and yet consume the same amount of power than their less durable siblings.

1000h was found to be the sweet spot between the conflicting requirements of efficacy, light quality and lifespan. I’m guessing that the communists decided that their citizens had different priorities to the rest of the world.

it’s a little bit more complicated than that…

I knew what video that’d be before I even clicked it :)

And today industry and industrial designers operate under the “Planed Obsolescence” paradigm, where the goal is for the item to fail right after its warrant expires.

A former appliance manufacturer in my country operated under a 30-year lifespan policy. When this manufacturer changed to being an assembler using off-the-shelf parts, they started filling a big warehouse with spares so as to continue to support the appliances for 30 years even if the actual manufacturer no longer made the parts.

I don’t own any “retro” RISC OS hardware, but I still appreciate that ROOL tries to support back to OS 3.1 where practical*. A lot of other open source projects cut off support for older hardware even when there’s no commercial reason to do so.

*This is not to say that the support must continue ad infinitum.

If I only could get a working harddisk for my oh-so-upgraded (*) Archimedes 310..

You can get reliably working SCSI “harddiscs” easily in form of BlueSCSI and SCSI2SD. The latter I have even tested extensively with an A3000/HCCS-SCSI minipodule, it works just fine. BlueSCSI testing with Risc PCs and various SCSI podules pending.

You can get reliably working SCSI “harddiscs” easily in form of BlueSCSI and SCSI2SD

I’m reminded of my first “massive” HD for my A3000. I wondered how I was ever going fill 20MB.

I’m just reading data off a 1999 15GB disk in my RISC PC. Most of it, anyway!

I’m reminded of my original RISC OS system, an A3000 (upgraded to 4MB) with a Baildon (I think?) IDEa mini-podule with a 20MB hard drive! Was fine until I installed Fidomail and ran as a point off a BBS (Skyline, if anyone remembers that one?)

Or more recently, the proto-smartphone I had back in 2005, a Nokia 9500, which had an entire 1GB of MMC (not even SD) storage.

Bard had it’s model replaced in the last day or so with the Gemini Pro model.
I was giving it a try and decided to ask it to make a program to add two numbers in ARM assembly. It was adequate but didn’t return to system properly so I gave it a gentle nudge about that. Surprisingly it just spat out the missing section as a rough template to drop the first entry. Even more surprisingly the program exits with SWI 0

Upon asking it, it revealed that it does know something of how RISC OS works. So if you’re in the mood for debugging a totally broken program that probably doesn’t do what you want or even have the correct syntax, go have a chat!

Even more surprisingly the program exits with SWI 0

So completely wrong then.

You could use statistical word guessers (“AI”) to debug your program, or you could try boiling your head. I’d expect equivalent levels of success.

You could use statistical word guessers (“AI”)

A colleague said, why use Artificial Idiocy when we have such a rich resource pool of natural ones?

Even more surprisingly the program exits with SWI 0

ARM Linux uses that for system calls. They couldn’t be bothered looking at documentation that would have given them a separate SWI space.