I was thinking about where R&D comes from. And then I realized that what I really cared about was the R of R&D. Research, real research which has no immediate product connection, is something only big companies can do. Start ups do their R at their last job or in the university. Then once they’ve “started” they do the D—sometimes it even works.
Medium size companies do D with a longer time horizon—but it is all product directed and ROI calculated.
As a company gets larger it can allocate one half percent of its revenue, maybe even a full percent to basic research, investigations into things that look interesting but have no short term, or maybe even medium term payoff.
If you had a company that had sales of $50 billion that means you could spend $500 million on pure research—a year. If you had a $5 billion company you could invest $50 million into long range stuff and not feel a significant hit on the bottom line—most companies spill or lose 3% or more of their revenue. Be a little more efficient, careful, and you’ve got a hellofa research budget.
Philips, IBM, Intel, and Microsoft do that. They’ve got the revenue to support really long range stuff. They explore things that may have an impact on them in ten years—how’s that for long range planning? And they discover stuff, and we all benefit.
That led me to the GPU compute argument—every researcher can now have a super computer at his or her desk. The impact of that statement is profound and literally life changing. Researchers at universities and national labs plan their research on the basis of compute time. A researcher may want to investigate the electron attraction of protein-protein interactions in the photosynthetic pathway, and knows that it will take ten years of interactive computational time on a super computer (assuming he or she can get access to a machine.) That influences the level of investigation—where will I be in ten years the researcher asks—retired, at a different lab? And so they scale back their ambitions and look for a simpler case—so humanity suffers.
But, if now the same calculations can be done in two years on a local machine (that costs less than $10,000) the researcher will have answers that will undoubtedly be catalytic to a new set of questions. THIS—is the singularity—not smart robots that will enslave us—but tools that will deliver answers as fast as we can think of the questions.
And it’s happening in all sorts of disciplines—medicine, computer visualization including entertainment, socio-political, finance, and simulations too dangerous or expensive to do in real life.
GPU compute lowers the investment cost for pure R. It can’t replace surveys and some experiments, and those elements of research will always be time consuming. But once the data is collected the time to process it, and play with it approaches zero and answers come out faster. R on steroids.
We are at the threshold of an explosion in new scientific discoveries. Products will be faster, stronger, safe, and less expensive. Extremely accurately designed target drugs will be created—perhaps on an individual genome basis in real time in the doctor’s office—no “there may be side effects if you suffer from boredom…” And entertainment will approach, if not reach the holodeck—the holy grail of computer sim-viz.
You are a witness to this revolution—this is your second one. You were here when we had the internet revolution, the biggest thing since electronics, and prior to that steam—you’ve seen the history charts. Well welcome to the singularity, it’s going to be great, it’s going to be insidious, and most of the population won’t even know it’s happening, they’ll just live longer, happier, healthier lives and their kids will grow up taking that as the norm—no kidding, people only lived to be 100 in the early 2000s—wow!