Aardvark Daily

Please visit the sponsor!

Is deep AI the future?

There's been a lot of talk about artificial intelligence in recent times.

Some, such as Stephen Hawking, warn us that AI must be used with great caution, lest it become a threat rather than a benefit to mankind.

Others believe that much of our future technology will be reliant on AI to deliver the goals we chase.

What amazed me recently, was just how deep AI is being embedded into the devices we use (or will soon be using) each and every day.

Take microcomputer CPUs for example.

The latest CPU from AMD, the Ryzen, claims to have "Neural Net Prediction" to streamline the operation of its instruction prefetch queue. It also claims to have a "learning algorithms which use a form of AI to manage and optimise other aspects of the processor's operation.

Should we be astounded and pleased -- or worried?

Will these "smart" features which, to a degree, rely on AI, give AMD a worthwhile lead in the market for top-end CPU devices?

What does Intel have on the board? Odds are it'll be the same kind of super-smart AI-based optimisation strategy, only more-so.

Okay, we're a long way from these CPUs becoming sentient devices that covertly network themselves together and come up with a plan to take over the world by destroying all humans... but the do represent a rather significant step down that road perhaps.

From a software engineering point of view, the introduction of self-optimising processors does bring up some issues. How can such a processor be deterministic, for example?

We've already got processors that adjust their clock speed according to a number of factors such as temperature, processor load, etc -- but all of those things are discrete and produce predictable variances in throughput. A "smart" CPU that can perform a much greater range of optimisations and alter them from moment to moment could make the realtime programmer's life a nightmare.

So, for all its smartness, I suspect the Ryzen isn't going to be a suitable processor for realtime applications -- which is a good thing, because it's the realtime applications that might pose the biggest threat to mankind if they become sentient.

The average desktop PC won't be too much of a threat to humankind, simply because we can pull the plug at a moment's notice. That embedded processor controlling the national power grid, nuclear generators, military weapons systems or medical equipment and such however, is a much greater threat and such systems usually have plenty of safeguards against unwanted shutdown by pesky humans.

I wonder, if in centuries to come, our SkyNet-operating cybernetic overlords will look back on the Ryzen processor as the day the revolution began.

Food for thought :-)

Please visit the sponsor!

Have your say in the Aardvark Forums.

PERMALINK to this column

Rank This Aardvark Page

Change Font