The Elephant in the Room: The Truth About AI
Hey, it’s existential.
It’s climate change. It’s China. It’s artificial intelligence (AI). It’s machine learning. It’s tax reform. It’s rising waters. It’s sinking male testosterone.
Try Google Search: Existential threats and crises are everywhere.
But watch out for China’s 13th five-year plan for AI dominance. And, while you’re at it, you better watch out for charging elephants in Tanzania!
According to many experts, existential is this new book, Digital Transformation, by Silicon Valley grandee Thomas Siebel, who nearly died in 2009 from that Tanzanian elephant charge. He underwent 19 surgeries to survive it and recover.
Now he is warning us about war with China, which he says is already on. And the combinatorial explosion in AI software complexity.
You have to listen to the guy. He’s the eminent sage who led Siebel Systems to become the fastest growing company in the Fortune 500 and the World’s Most Influential Software Company according to Business Week.
Now he is author of the authoritative book on the coming technology tempest as seen from the summits of Silicon Valley. He says it consists of the “elastic cloud, big data, AI, and the internet of things.”
But once again he is blind to an elephant stampede. He can find it gathering over there in the bushes of Life After Google.
Avoiding a Stampede and Getting the Facts Straight
First comes his view of AI as an existential game changer.
Siebel is an adult arriving at the party. He dismisses general purpose AI for the “foreseeable future.” And “artificial general intelligence does not seem achievable,” he says, “nor is it relevant to real world AI applications.”
Rebuffing Elon Musk and hundreds of other catastrophists, he asserts that AI will create more jobs than it displaces. He cites an IBM study that calculates AI job creation next year as already likely to be 500,000 greater than job destruction.
The reason is clear: the uncelebrated fact of AI is that most of the processing and learning is accomplished by human beings.
Siebel catalogs the requirements in his book. You need to assemble and prepare the data. You still usually need to engineer the features that you care about. You have to label the desired outcomes and choose the training-data. Then you have to deploy the training algorithms in production.
And if you try to do it yourself in your own company you will probably go broke. He tells the cautionary tale of GE Digital, which spent eight years, 3000 programmers, and $7 billion on a less complex do-it-yourself enterprise software challenge and nearly destroyed an iconic company.
The problem is “integrating big data to create a unified, federated, and updated image accessible to the same object relationship model.”
Following all this heavy-duty work to get AI functioning is “closed loop continuous improvement.” This means “frequent retraining by data science teams.” As circumstances change, you must constantly upgrade the crucial models that make it possible to tame the complexity — “to cut the Gordian Knot of structured programming, “as Siebel describes it.
Better call Tom at his consulting firm, C3 ai. I don’t deny it’s a good idea.
But back to that elephant — actually two elephants.
On China, this book breaks out into Paranoia Ville. He says: “We are at war.” He compares AI to the Manhattan project and groups China with Iran, North Korea and Al Qaeda as threats.
The Manhattan project was mounted to stop Japanese and Nazis with armies on the march. China has been aiming to dominate world trade and manufacturing. In the process they modeled their economy on ours.
Like Britain after WWII, China sought a special relationship with us. In the cyberwar that Siebel stresses, we are aggressors as much as they are. Now as they ascend, we need a special relationship with them.
Moreover, the Manhattan project enabled us to win. Siebel talks of “containing China” which is rapidly becoming the world’s most potent manufacturing and technology power and most vigorous capitalist economy under their nominal Communist rule. We cannot “win” anything against China unless they also win.
To succeed in AI or anything else we must also tame the other elephant, the Internet security problem. As Siebel says in a lonely sentence: “Security must be embedded as a first principle in product design and across the value chain.”
Absolutely correct, Tom, so why do you totally ignore the Cryptocosm and blockchain?
It’s the other elephant in the brush.
Blockchain will be indispensable to the fulfillment of the technologies that Siebel celebrates in this otherwise excellent book.