Apple's Reintroduction to IBM's Supercomputers Will Pay Off

Apple dabbled with and then got out of the supercomputer business a decade ago. These days, Apple is being reintroduced to the benefits of supercomputers, via Watson, thanks to its partnership with IBM. This is a good thing. A very good thing.

IBM Blue Gene supercomputer

The supercomputer industry has always been a small and exclusive club. Only those organizations, typically governments, with great wealth and great need have been able to fund supercomputers from companies like IBM and Cray. By its nature, that industry has always made it tough for the manufacturers to really strike it rich.

Back in the 2000-2005 timeframe, Apple was just bursting on the scene with its new BSD UNIX OS called Mac OS X. It was natural to exploit that new OS by appealing to scientists for whom UNIX is their lingua franca. After all, the iPhone had not yet come on the scene, and so this was a logical path to revenue and respectability. The Apple Xserve 1U rackmountable device with a PowerPC CPU made for a great laboratory cluster, and many organizations built even modest to fairly large supercomputers with the Xserve.

If Apple had pursued the supercomputer business with great aggressiveness, as IBM, Cray and SGI did at that time, it could well have become a money sink instead of a mass market consumer electronics bonanza like the iPhone. It was a good decision to make by Apple, even though computer scientists were begging Apple to get into the business.

I remember one scientist from the Los Alamos National Laboratory who old me at that time that he yearned for the day when Apple would make supercomputing as easy as it made UNIX for the everyday Mac user. He was tired of writing his code in the vi editor, running command line compilers, being constantly overworked and missing Christmas holidays with his kids. Government deadlines always seemed to get in the way of family life.

Supercomputers and ROI

The supercomputer industry has always found it difficult to convert the codes, tools and machines into a massive return on investment, like the Apple iPhone or even the iPad. Fortunately, IBM, with Watson, has been able to convert an excellent supercomputing platform into commercial use. Watson's exploitation for medicine is essentially the Holy Grail of supercomputing. See. for example, "IBM Intros Watson Health Cloud with Apple HealthKit Support."

While Apple, to my knowledge, is not in ownership of a supercomputer or even buying significant amounts of time on one, it can't hurt that Apple, especially under the watchful eye of Tim Cook, is being exposed on a frequent basis to the benefits of supercomputing to solve the most difficult of human problems.

Apple's expertise has always been making technology easier to use on a personal level. There is a certain class of activities, constantly changing, that invite creativity and new ideas for productivity on the desktop or the laptop. Or the tablet. But for activities that require massively parallel computing, beyond the realm of the everyday technical customer, there is much to be gained. Historically impossible problems can start to be attacked.

Supercomputers capable to 20 to 30 petaflops can do things that take our breath away and would be of great benefit to Apple in addition to medical research. Examples I can think of are simulations of various manufacturing methods, inventory optimization, a stochastic analysis (and prediction) of the consumer market, and the creation of and analysis of candidate artificial agents for both computer management and education.

For example, how many Apple Watches will be sold? A supercomputer, with the right codes of course, might be able to explore all the possible market and human scenarios and suggest a number with modest uncertainty.

Supercomputers and OS X

A Mac is is wonderful device, but it's fairly helpless when it comes to introspection into its own faults. Typically a human has to dig in, read a how-to and manually fix a problem. Apple OS engineers fix bugs on a priority basis, but don't always successfully flesh out hidden interactions or security implications. Especially when thousands of APIs are introduced annually.

In some cases the end user can diagnose a problem, and in some cases the problem is so deep that the Mac is a black box. We don't know why something has gone wrong, and we're helpless. We try workarounds.

If Apple is ever going to build a next generation OS, beyond OS X, it can't simply be more and more complex UNIX code. That's even more important when we put our lives in the hands of autonomous driving cars. Eventually, the code has to start taking care of and thinking for itself, not in terms of outward actions but, instead, internal maintenance. Prolonged exposure to supercomputing concepts and technologies will be profitable in this process.

Even though Apple is, nowadays, primarily the supplier of health data and letting IBM do the heavy lifting, the association of the two companies and an appreciation of what they can do together with Watson should end up being a significant source of inspiration to Apple executives and engineers.

Maybe, after a decade, Apple will come full circle and get back into supercomputing along a completely unexpected avenue. Not as a maker of hardware, but as a corporate, expert customer and IBM research partner. That would be good to see.