Advanced Computing in the Age of AI | Thursday, March 28, 2024

Opinion: Actually, Plenty 

<img style="float: left;" src="http://media2.hpcwire.com/dmr/DMR_Crysis2.jpg" alt="" width="68" height="78" />Far be it from me to condemn another writer’s work, but honestly, ExtremeTech’s Sebastian Anthony just published an article that is bogglingly misdirected. The article is called What can You Do with A Supercomputer?

Sometimes you despair. Really.

Far be it from me to condemn another writer’s work, but honestly, ExtremeTech’s Sebastian Anthony just published an article that is bogglingly misdirected. The article is called What can You Do with A Supercomputer? He answers this way:

“Not much.”

What can you do with a supercomputer? Not much. I guess everyone can go home.

After a page discussing the cost and power requirements of supercomputers, Anthony gets to the meat of it. You won’t be able to play Crysis 2 on most supercomputers, and the only things they’re good for are forecasting the weather, simulating nuclear explosions, and molecular dynamics modeling. Seriously. THAT IS ALL HE THINKS THEY’RE GOOD FOR.

Break it down with me, people. First of all, Crysis 2? Please. That game’s a year old and wasn’t very good to begin with. No one’s playing it any more. Live in the now. Not being able to play Crysis 2 on your supercomputer doesn’t significantly erode the overall value of the devices. Go play Crysis 2 on your trusty Xbox.

But more important, and damaging, is the implication about the applications of supercomputing. Anthony’s article pretty explicitly states that supercomputing’s only good for eggheaded geek stuff so removed from day-to-day reality that these machines may as well not exist. Nuclear simulations. Molecular dynamics. Weather modeling. “Nothing that mortals like you or I usually concern ourselves with,” says Anthony.

It is important that we mortals stay in our place, and limit any kind of subversive, inventive thought or idea-mongering. We mortals have never found unexpected need for products unintended for us – who uses WD-40 or cordless tools, both invented for the space program? Smoke detectors are also unnecessary in the homes of mortals; they were invented for the immortals dwelling in the Skylab station in the 1970s. How many times have some people failed to learn that the answer “not much” is never particularly applicable to the question “what can you do with this?”

That attitude is base, smallminded, dangerous, and obstructionary to those who want to push the limits of whatever frontiers might be available. If you asked the 1904 Sebastian Anthony what you could do with a car, he’d say “not much.” There were no roads, after all; no convenient gas stations. Horses got people where they needed to go at a pretty steady clip. With two little words he’d have brought down the century of mass production. Nice work!

What if that attitude had been applied to, say, flight? What if the Wright Brothers flew and then let it go, because what on earth would anyone do with the ability to fly somewhere? “Mortals like you and I don’t need to concern ourselves with flight,” Anthony might say. “It is only useful for seeing what’s happening above the clouds, surprising your friends by dropping objects on them, and terrorizing birds.

Travel back with me 3,312 years. It’s the 13th Century BC and this Anatolian dude comes running out of his yurt (which is on fire), waving a red-hot ingot. “Behold! I’ve invented a new metal!” he cries. “It’s much stronger than anything I’ve ever seen! I will call it… steel.”

“Go put out your yurt, fool,” says another Anatolian. “Iron is strong enough for we mortals. Or do you think that tribe down the road will waste its time on this ‘steel’ of yours?”

Really? The answer to what can you do with a supercomputer is “not much?” I thought we’d moved past this.

Do I need a supercomputer in my living room? No. At least, not today. Who knows, someday soon maybe there will be use for that kind of processing power in the home. But I am very pleased that supercomputers are being used not just for weather and molecular dynamics, but all the stuff we talk about here. What’s been done with supercomputers? They’ve made my car lighter, safer, and more fuel-efficient. They’ve optimized small-scale wind energy devices, hopefully one day cutting my power bill. They’ve quantified cradle-to-grave lifecycle responsibility for products, allowing cheap consumer goods with minimal environmental footprint. They were instrumental in developing new packaging that saves several hundred tons of plastic a year. And, yes, okay, a supercomputer did learn all about tornadoes so when one wandered through the city I live in the other day, the local weather people could talk about it for two breathless hours.

ExtremeTech should be ashamed of itself. Anthony should be denied dessert privileges for a month. No one should be asking what can you do with a supercomputer, they should be asking what can’t you do with a supercomputer, and if not why not. Can’t play Crysis 2. Okay, that’s one thing. What else have you got?

If it were harmless I’d just call Anthony’s article misguided, but it’s not harmless. Every day we’re struggling to convince manufacturers that there’s more to HPC than just the national-lab stuff. The hardest part has always been helping people see: see what opportunities exist today, and see what opportunities might exist tomorrow, opportunities we haven’t even considered yet. When articles like this one get published, the whole effort of helping manufacturers be visionary is set back once more. Because not only does the article flatly state that supercomputer applications are limited, it goes into painful detail about the cost to build and run them. This is pointless economic scaremongering that offers no benefit or informational value.

I find myself thinking of the Raspberry Pi Foundation, which is days from shipping a thirty-five dollar, credit-card sized computer that runs Linux and clocks in at 700MHz. There was a time when 700MHz was a supercomputer. Now it fits in your pocket.

What can you do with a Raspberry Pi? Who cares? I guarantee that people will find countless uses for them. And the people who don’t, or can’t, or refuse to look, lack vision.

You will also not be able to play Crysis 2 on a Raspberry Pi, by the way.

Anthony’s article isn’t exactly poo-pooing supercomputers – in his view they’re great for the three things he mentions. But it’s so closeminded it might as well be. If every innovation bore with it the albatross of “this is what it’s good for and nothing else,” man, we’d lose a lot of progress. I prefer to think of innovations in terms of opportunity, not limits.

EnterpriseAI