Disclosure: NVIDIA is a client of the author.
I’ve been thinking about an NVIDIA offering called Omniverse. It’s designed to work with the company’s graphics cards and use game elements to create content rapidly, but it can also create visual simulations.
NVIDIA has been a major simulation advocate for autonomous cars, and its tools could be used to simulate other things. (NVIDIA’s new headquarters existed virtually years before it was built.) I bring this up because, as we see Reddit folks mess with the hedge funds, it strikes me we don’t use simulations enough to validate decisions before they’re made. That’s especially true of companies.
So let’s talk about how models and simulations could improve productivity and reduce waste and lead to better, smarter outcomes.
The critical need for tools
While we are surrounded with simulation tools and various companies in the defense, finance, marketing, and disaster mitigation industries use simulations and models extensively, we don’t use them for personal or corporate decisions. That’s like having a crystal ball that can tell you the future and not using it because the learning curve is too high. (This reminds me of the old joke where a kid is pushing his bicycle to school and a friend riding by aske him why he’s on foot? His response: he’s late and doesn’t have time to get on the bike). It’s funny until you realize critical decisions are being made by companies and government without first simulating outcomes. I’ll bet the reason is that they don’t feel they have the time or money.
The fascinating thing about simulations is they can often model changes and deliver results in real-time. More importantly, as AI capabilities advance, simulation systems can learn from past use cases to reduce the time to set them up and increase their predictive accuracy. You do have to be careful about introducing bias, but it is less damaging to be wrong than to have a significant project fail.
This issue comes down to our unwillingness to appear to be wrong and a habit of taking a position before we’ve researched it. As analysts, we are trained to defend positions, and to do research before taking that position. It makes this job different than most others, but is something everyone should do.
Let’s take buying a car. An analyst will study reviews — particularly customer reviews of a car and the dealer — they have a hierarchy of what they want in a car, and then they test-drive those that look compelling. They’ll also know how to get the best price and the tradeoffs connected with post-sale support. Others see an ad, test drive the car, and end up with something less than an ideal deal. (I bought two cars that way when young and regretted both.)
I’ve seen firms make catastrophic purchases by companies without doing adequate research, fail to learn from past mistakes, or ignore the need to bring onboard resources that can assure the purchase is a good idea. That’s why simulations and modeling are important.
Years ago, a guy came into my office — I was in marketing at the time — and asked me to build a marketing plan for a product we’d spent $20 million building. I asked him to describe who would buy this thing, because it made no sense to me. After doing a $20 study, we discovered there was no market for the product. Had that been done first, $20 million could have been saved.
Wrapping up
Many of the problems we see in Washington or in executive offices involves people making decisions as they were done decades ago. But we now have the ability with artificial intelligence to create simulations at a small fraction of the cost associated with a bad decision, vastly reducing risk. And while you might look bad if your proposed decision fails a simulation, if you made a bad decision and cost your company millions, there’s a pretty good chance you’ll have killed your career.
One last example: when I was in competitive analysis, we had an instructor who drew an x-y chart on the board. Vertical represented speed; horizontal represented direction. He argued that if you found the right direction first, regardless of speed, you were more likely to be successful; if you didn’t, the more speed you applied, the worse things would get because you’d be accelerating in the wrong direction. Creating tools that better choose right directions, and making those tools easier to use and more accessible, is the best way to assure positive timely outcomes.
