Today, an interview with John Giannandrea (JG) was posted. JG recently retired from Apple where he was the SVP of ML and AI Strategy. Before Apple, he was the head of Search and AI at Google. The interview is on the Cobot podcast Deployed which is hosted by Brad Porter. Brad and I worked closely with JG at Tellme Networks at its start in 1999. It was a great company, founded on a good idea, that was years too early. I have deep appreciated JG insight and leadership. I followed JG from Tellme to Metaweb, which was acquired by Google to become the heart of the Google Knowledge Graph and transformed Google Search.

Paradigm Shift

AI is bringing a paradigm shift, but most people won’t recognize what the shift is until it’s already taken root. Paradigm shifts are usually obvious only in hindsight. Looking at pass paradigm shifts can be help understand how future shifts can impact society.

JG shared the first paradigm shift he saw was “the web”. When he first saw the Mosaic web browser he was amazed by the quality of the abstraction. He immediately knew it was going to change the world. A few days later he went into his manager’s office and quit saying “This internet thing is the future, and we aren’t doing that at General Magic”.

Brad and I had a similar reaction. Brad saw Mosaic while a student at MIT and joined Netscape (the company that developed Mosaic as a commercial product) when he graduated. I was impressed by a similar system a year earlier. Several of my colleagues at The Ohio State University built something almost identical call Waiter, but were unable to release it quickly. Mosaic was bit nicer than the Waiter browser and HTML/HTTP was gaining a following quickly. Rather than competing, an afternoon of hacking on the Waiter server converted it into a HTTP service that delivered HTML pages. By Aug 1993 OSU was listed on the Mosaic “What’s New Page” as providing a wide range of resources for the web including all of Usenet FAQs, IETF RFCs, Texinfo documents, and a variety of other information.

The idea of a “web of information had been around since Vannevar Bush’s “Memex” (1945), was part of Doug Engelbart’s Mother of All Demos (1968), and became a consumer product when Apple shipped HyperCard (1987). These systems had tight binding between the presentation of the information and the underlying system.

The Mosaic web browser brought three simple ideas together offering a new paradigm of how people could access and share information through computers:

• Information can be easily presented which contains links to related information that a reader could select – e.g. hypertext (HTML)
• The retrieval of information could be separated from the presentation of that information and pulled using a very simple protocol (HTTP)
• Content could be published by anyone, there is no centralized control (URLs)

Boiled down, Mosaic gave people an open system that was easy for anyone to contribute to, and was (at least initially) free for anyone to consume. The release of the browser that would run on any modern workstation and a web server that could easily be installed on any UNIX server allowed “the web” to spread quickly.

The web took off like a rocket. The world had changed, though at the time we didn’t know how radically.

Overhyped, Under Appreciated

Futurist Roy Amara famously observed, “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”

In the late 1990s there were countless companies trying to “get on the web” which produced the Dotcom Bubble. There was lots of hype, money poured in, lots of people came to Silicon Valley for “the new gold rush”, and then many of those companies disappeared: companies delivering web content, Internet hosting companies, and equipment manufacturers.

Almost no one envisioned how radically “the web” would re-shape the world we live in today. One exception was John Brunner’s 1975 book Shockwave Rider that predicted the rise of what he called “the net”, and many of the negative aspects of life when everything is online. He also briefly touched on the foolishness of perusing an unrooted AI.

Our daily lives look very different from the days before “the web”. Almost every aspect of our life is “online” today: smartphones, continuous connectivity, social media, items purchased with a single click and delivered in hours. This was not an instant transformation. As JG noted, it took 20 years before his dentist let patients manage their appointments online. People who were born after 2000 can hardly imagine life before the web.

AI will bring a bigger transformation than the web… but it will take time for us to see the real transformation. Right now people are expecting massive, immediate changes. Many of these changes won’t last. People will be laid off because AI will “replace them”, only to be hired back when it’s clear the AIs can’t get the job done. Large investments will be made and massive data centers will be built. Some of these will be the heart of the future, but others will rot away, unused, because investments were too far ahead of returns. Billions will be spent on building models which will ultimately be replaced by something significantly more powerful while being significantly more cost effective. As JG noted, current models are autoregressive and future models will not be as expensive to build as they are now.

When we were working at Tellme, JG would regularly reminded me significant changes took time…. Especially when dealing with AT&T… but that’s a story for another time.

Hill Climbing

In the interview, JG touched on something else that is important for real change. He calls it “hill climbing”. This is a disciplined, data-driven, refinement required to deliver an impactful product. This deserves more attention than I have right now. Eric Reis provides context of how hill climbing informs effective product development in his 2010 post Learning is Better Than Optimization.

Project Xanadu envisioned a system much like “the web” … arguably much better, but was excessively complex. It aimed for perfection, was unwilling to make incremental progress, so failed to deliver. I would encourage people to read Lessons from Xanadu and Richard P. Gabriel’s Worse is Better.

Humanistic Intent

The section of the episode that most resonated with me was JG’s observation:

I think that the best technology is, is created with a real genuine humanistic intention to do a good thing. Not as a business idea.

I mourn the Silicon Valley of the previous century, when the focus was mostly on innovation and building things with a positive impact. There is a good description of this in John Markoff’s book What the Dormouse Said and his recent article Silicon Valley’s forgotten formula for genius. Today the focus seems to be more on personal achievement and corporate profitability. Markoff article cites the book “Move Fast and Break Things” by Jonathan Taplin:

Taplin distinguished between two distinct cultures. The first was the technological utopianism of the Whole Earth Catalog era — engineers, designers, and entrepreneurs who believed the tools they built would make the world more humane, more democratic, more individually empowered. He dubbed the second, which he associated with the rise of Peter Thiel and the PayPal Mafia, “technological libertarianism.” Its founders were not, in the original sense, building tools for users; they were attempting to control markets. Technology was supplanted by money as the defining coin of the realm, and the question of what the technology was for receded to the extent of becoming a bullet point for a marketing department. 

This is what I think the dominate narrative is in Silicon Valley today. JG is more positive than I am. He seeing people who still embrace the old ways, and had hope, because those ideals are still present. I see the people holding on to the old ideals aa exceptions, they a small minority which is under constant pressure.

I have long said that Milton Friedman’s shareholder value theory started a cascade has slowly eroded the value of companies as described in the recent article Shareholder Capitalism is Anything but Good for Shareholders. I also wonder whether the modern MBA program is a contributing factor, and I’m not alone in that thought. A LinkedIn post led me to a New York Times opinion piece by Chris Marquis, a professor at the University of Cambridge (previously at Harvard and Cornell), titled I Teach at a Business School. It’s Time to Shut Them Down and an article in Forbes clarifying that it should be a shift in the system not closure.

Long View

My biggest delight in the interview was a story I hadn’t heard before… an encounter with a visiting cardinal from the Catholic Church. A clip from that section of podcast:

A paraphrase of the exchange:

Cardinal: How do you know that building these technologies is going to be good for humanity?

JG: Progress of technology since the Enlightenment… all the measures which show progress in the last 100 years…

Cardinal: Laughter – You are running an uncontrolled, 200-year experiment. You have a blind faith that technology is making our lives better, but how do you know?

JG: Oh, I get it. I am talking to the Catholic Church, which is 2,000 years old… they are taking the long view.

JG’s characteristic response about progress isn’t blind faith. Books like Factfulness by Hans Rosling and Enlightenment Now  by Steven Pinker have numbers that show many areas of life are improving. And yet, the Cardinal asked a critical question. I believe the answer is that we don’t know how the development of AI is going to impact the human experience. I think it’s safe to say it will be profound. We need to be extremely thoughtful. I appreciate initiatives like Delta which are asking the question: What does it mean to be human in an age of AI?

Additional Material

• Regurgitation or Intelligence? Can AI Generate Novel Thoughts? Another recent Interview with JG on the podcast General Magic and the Soul of Tech. Covered some of the same information but slightly different angle.
• Cooperation in the Bay Area… why Silicon Valley and not Route 126 (Boston to NY) dominated high tech.