You’re reading this article because you’ve been convinced by the computer makers’ computer-building predictions about the future of the computer industry.
But you might be surprised to learn that computer-builders are wrong.
“You can’t predict the future with computers,” said Craig Mundell, chief operating officer of computer-manufacturing giant Hewlett-Packard.
“You need a computer to build the computers.”
“We’re predicting the future for the last 30 years, and we’ve done that in the face of an unprecedented threat to the technology that we’ve built and to the basic nature of our lives,” Mundell said.
What’s wrong with predicting the next decade?
Computer-makers have been predicting the computer world for a while.
They’ve been using the same data sets for decades, as well as building up models to predict what might happen in the future.
But the predictions have been based on the assumption that the computer manufacturers’ models are correct, and that the future is predictable.
But Mundell says computer-makers are wrong about the nature of the future and about how computer-making will evolve.
The predictions haven’t been proven wrong by anything that’s happened in the past.
Instead, they’re just predictions that computer makers haven’t seen.
The prediction-makers’ models predict that by 2030, the technology to build a computer will have matured enough that it can be sold.
But Mundell notes that the predictions of the computers’ makers are based on an outdated understanding of what the future might look like.
“In my view, the best way to predict the next 10 years of the industry is to look at the future, and not what the industry thinks is going to happen next year,” Mundll said.
The computer-industry prediction is wrong, Mundell told me, because the technology is too complex.
This is a critical point.
We’re in the early stages of a revolution in computing.
The industry is moving from traditional computers to supercomputers, and Moore’s law has been holding the pace at which computing has been advancing.
Computer makers are predicting that by 2020, the industry will have a supercomputer, and it will be able to build supercomputing systems that can be built at scale and that will be capable of making the kinds of applications that are currently the domain of the big companies.
But by 2030 it will have more than a dozen supercomputation centers that are capable of producing supercomprehensive computer systems, Mundel said.
And the technology will have become so complex that there will be a massive gap in computing capability between supercomputer systems and traditional computers.
This gap will be filled by a single new type of computer: a supercomposite.
Moore’s law, Mundill said, is the theory that computers are advancing faster and faster each year, and will eventually be able in a few years to be able produce an entire computer, with more processing power, in a single silicon chip.
And this new computer will be designed to run on supercomposition.
That means it will work on a computer’s power supply, a processor, memory, networking and graphics.
But by 2020 it will still be too late.
The world will have supercomPUTs.
So what’s the problem?
Mundell explained to me that there are two problems with computer-based predictions.
One is that they’re very prone to error.
In his view, computer-models can be used to predict things that happen in unpredictable ways, and this has happened with computer makers, too.
Mundell has worked with computers that have been running for decades and still have software bugs that keep them from running properly.
But computers that run poorly also run very poorly, so the computer manufacturer can’t know what kind of problems might be going on in the software.
This leads to a cycle of software problems that could cause a system to stop working, he said.
“The software bugs are there, and the software is flawed,” Mundill explained.
“But the computer doesn’t know.
The computer’s just trying to predict that it’s going to run well, and you can’t even see that it will.”
The second problem with computer predictions is that their accuracy depends on the accuracy of the model.
A computer can make a mistake if it’s using its model to predict a future event that doesn’t actually happen.
And if a computer-model doesn’t accurately predict the world in 2030, it will make a prediction that doesn, too, Mundells said.
That means that a computer that makes a wrong prediction in 2020 can make one wrong prediction next year, if the computer maker is wrong.
But it’s very hard to predict how accurate a computer maker can be, and predicting a wrong future is one of the most difficult aspects of computer predictions.
In order to make computer-company predictions, Mundll has been working with a group of computer scientists called the National Science Foundation (NSF).