My Interview with Forbes Magazine [Part 3]
In the last two Daily Prophecies I’ve been sharing parts of an interview I did at Forbes with Rich Karlgaard.
Keep scrolling to read the last excerpt…
Q: Why would a smart company like Google get it wrong?
The advances in machine learning that Google trumpets and preens about really just advance in the speed of processing. The Google guys woke up with a Moore’s Law bonanza from the chip industry and imagined that they had invented it. When their Go-playing computer can play more Go games in a minute than the whole human race has played through all of history, that’s not a great advance in intelligence. It’s the same intelligence just accelerated to terahertz speeds. Yet this creates an illusion of super-intelligence, that machine learning can somehow gain consciousness and usurp humans.
Q: Again, the nerd IQ arrogance that artificial life will trump real life.
In my book, there’s a chapter on the conference in Asilomar, that was financed by Elon Musk, and that I think was a kind of intellectual nervous breakdown. All the Google people were there to tell the world that the biggest threat to the survival of human beings was artificial intelligence, which they themselves were creating. What a great bonfire of vanities!
Q: To continue your Tom Wolfe allusion, where are you finding the people with the right stuff today?
Well, it was Alexandra Wolfe’s Valley of the Gods that focused on Peter Thiel and the Thiel Fellowship Program, which gives $100,000 in grants to kids who leave college to start an innovative new company. In Life After Google, I tell the story of Vitalik Buterin, who was born in Russia and moved to Canada and who was initially turned down for a Thiel Fellowship. Buterin was almost at the Cinderella age of 21, when they broke down and gave him his $100K. He created Ethereum, an amazing new computer platform based on a new blockchain that he invented and programmed with a new computer language called “Solidity” and financed with a new currency called “ether” that he generated based on a new unit of computational value called “gas.”
Buterin has violated all the principles of Silicon Valley Venture Capital—that you can only change one thing at a time. You can launch a new internal combustion engine in your carriage, as long as you’re careful not to remove the horse. If a company plans to make a software innovation coupled to a hardware innovation, it’s a very doubtful prospect. So Buterin’s quadrilateral achievement is a stunner. He’s one of the great figures in today’s world economy, leading a global movement of some 2,000 initial coin offerings on Ethereum, which has raised close to $5 billion for an array of new inventions, new distributed systems and new security architectures.
Q: Who else?
Another Thiel Fellow I’ve encountered is Austin Russell, founder of Luminar. Forbes just named him the chief manufacturing guy in its 30 Under 30 list. As a kid, Russell appraised all the current LiDAR systems for self-driving cars that were demoed during the cross-country DARPA test, on which the current self-driving car industry is founded. Google hired the key winners of that test, and they are the key figures in self-driving technology across the board, from Waymo to Aurora.
But Russell saw that all this technology couldn’t possibly create a self-driving car that could survive amidst the myriad conditions a car actually encounters on the world’s roads, which are full of surprises. Such things can’t be anticipated merely by statistical machine-learning models, because these are, fundamentally, high-entropy events. A car has to be able to see and respond to combinatorial events that can’t be anticipated. So you need to have LiDAR systems, vision systems, for the cars that are 50 times better than the ones that are currently being tried. After receiving the Thiel Fellowship, as well as money from the Thiel-backed 1517 Fund, Russell took Luminar into stealth mode. Starting from scratch, he created a 50-fold better system, which is based on new frequencies, innovative hybrid-chip technology and rastering lasers. It’s now been adopted by Toyota and by three other industry-leading auto companies. This is an amazing feat for Luminar, led by a young man Peter Thiel lured out of college.
Stephen Balaban at Lambda Labs. He based his work on the knowledge that the real source of machine learning is fast parallel computation. Balaban took game machines and made mobile data centers from them that function as well as Google centers do, but at a small fraction of the cost. He’s making tremendous advances and has all sorts of high-level customers. There’s a whole new Internet and security architecture emerging.
Q: A lot of the security stuff is coming out of Israel, isn’t it?
Yes. Buterin was a bitcoin exponent and a cofounder of Bitcoin Magazine. He went to Israel and encountered experiments that went beyond bitcoin in so-called “colored coins,” or tokens and “mastercoins,” that had smart-contract functions. He recognized that Israel was leading the world in cryptography. And that was really the source of Ethereum. It’s exciting to see all the advances that are coming from outside the normal channels. The venture capitalists are racing to catch up with these random guys from places like Guatemala and India, Estonia and Canada.
This phenomenon isn’t so different from the first group of outliers we saw at Intel in the 1960s. That’s when Silicon Valley emerged from the semiconductor industry, which began with Intel Corp., National Semiconductor and a few others in the Valley. Those guys came from all around the world and gathered in Silicon Valley. Gordon Moore was from San Francisco, but Bob Noyce was from Iowa! Most of the profits in the early years came from Dov Frohman, who was from Israel, and Andy Grove, who was from Hungary, and Jean Hoerni, who was from Switzerland. So I see a kind of rebirth of the Silicon Valley dream. It defies the current Silicon Valley re-centralization of leviathan companies that chiefly buy up their own rivals and buy up their own stocks with zero-interest money pumped out by central banks.
Q: Nvidia’s Jensen Huang is a classic outlier. Although Nvidia’s been around for two decades, few saw its potential at the center of AI and machine learning.
Nvidia’s an example I use in my book. Nvidia is what makes machine learning possible, not the incremental improvements in machine-learning algorithms that have been developed by the “software-eats-everything” crowd. Nvidia exploits the natural parallelism of image processing. Images are millions of pixels all coming in parallel; graphics processors compute those millions of pixels and, thus, are intrinsically a parallel-processing technology. They were developed for the gaming industry, and now they’re taking over the world.
That’s a hardware breakthrough, not a software breakthrough. And Nvidia led it.
Q: Nvidia is now big itself. It won’t be sneaking up on anyone. Its stock doubled in 2017.
Yes, and I fear that Nvidia is now trying to control the whole process. People are buying game modules and putting them together to form little data centers that can do machine learning more efficiently than big data centers, and the Nvidia marketers don’t like it. They say that everyone should buy the company’s hifalutin machine-learning chips, not use their G-force gaming processors. They want you to pay top dollar if you’re going to do parallel processing using Nvidia to do machine learning. This is a classic mistake, which I hope Jensen and Bill Dally and all the other brilliant people at Nvidia reverse in a hurry.
Q: Well, Nvidia’s current stance will just encourage knockoff competition; that’s inevitable.
Not only knockoff competition, it’ll encourage others to leap past. There’s a Thiel Fellow named Thomas Sohmers, who left school as a 16-year-old and started Rex Computing. He’s a chip genius and wants to completely change chip architecture. He gave a two-hour speech at Stanford that you can watch on YouTube. He, Austin Russell and Stephen Balaban, were all roommates in a house in Atherton. They’re spreading out now, but they still support one another’s projects through the 1517 Fund.
Q: Does this give you hope for Silicon Valley’s future?
It’s fascinating that, despite all of what I regard as delusional ideas, Silicon Valley remains the spearhead of world technology. Balaban is head of Lambda Labs, which is creating these new mobile data centers. He was one of the contributors to facial recognition and is an amazingly inventive guy. He learned Mandarin and then went to Beijing to try to create a Y Combinator technology incubator there. But he was pretty much confounded by the environment in China and came back to Silicon Valley with a new admiration for the amazing spirit of creativity and openness that still prevails here.
Q: Despite the increasing concentration of the Internet in a few giant siloes run by a few big companies that buy up their competitors and buy up their own shares?
That won’t last. I’ll refer you to Gordon Bell’s law. Gordon Bell was the technical chief of Digital Equipment and is now at Microsoft. He says that every ten years the rate of progress predicted by Moore’s Law produces a 100-fold rise in computer cost-effectiveness, which then requires a completely new computer architecture. Well, I first started writing about cloud computing in 2006. As a matter of fact, I wrote the first big article on cloud computing in Wired rather than in Forbes.
Q: We’re now beyond the ten-year point of Bell’s Law and cloud computing.
And a new architecture is arising. And this new architecture solves the Internet’s increasing concentration problem, as well as remedies the security vulnerability that comes from that concentration problem. And with blockchain, it may even prevent the repeated monetary crises that come from the daily $5.1 trillion in currency trading. A new architecture is emerging around the world, but it’s still largely centered in the Bay area, where Hewlett-Packard started the whole ball rolling from its garage in the 1930s.
So Bell’s Law says we’re ready for a new architecture, and, lo and behold, here it is. It’s based on cryptography, the same cryptography that Claude Shannon and Alan Turing developed during the Second World War. And that cryptography is now generating a new computer architecture based on blockchains, mathematical hashes and an array of associated inventions.
It’s the Great Unbundling. We’ll dissolve all the GAFA fab-five conglomerates. We’ll disperse the clouds of concentrated computing and commerce. We’re moving beyond digital and silicon to analog and carbon nanotubes and hybrid chips with sensors and 5G antennas everywhere. Even money is being disaggregated and reinvented. The clouds are dispersing into the skies — sky computing rendered on your laptop and smartphone, spread across blockchains, transparent and transformative.
Editor, Gilder’s Daily Prophecy