Google IO: Google Is the New Microsoft

Two years ago I wrote a blog post titled “Is Google the New Microsoft?” hedging it with a question mark. I think after the announcements at Google IO there can be very little doubt that Google is the new Microsoft.

Here is a short check list. Dominant position in one market that generates huge profits: check. Desire to compete in every large market from payments to games to music: check. Extending influence from one market into another by integrating products with each other: check.  Abandoning open standars such as RSS and XMPP: check. Having another large company in the field publicly accuse you of locking them out: check (and in the what-comes-around-goes-around department: that company happens to be Microsoft).

Now as a shareholder in Google since the IPO I have been very happy with Google’s performance. Yesterday Google’s shares reached a new all time high giving the company a market cap of over $300 billion. And if any of these new initiatives succeed there is more room for growth. On the other hand as someone who cares about the Internet as an open network and invests in startups I can only say: fasten your seat belts — the ride is about to get bumpy!

Google Glass: Dangerous Adoption Strategy

I am quite optimistic about the long term potential of augmented reality. But I question whether now is the right time for a top down adoption strategy with a polished consumer product. It seems to me that we are at the hacker and early adopter stage instead.

By going with an immediate mass market strategy and embracing celebrities I think Google is taking a very big gamble. Let’s keep in mind that the iPhone was far from the first smart phone. A lot of software and application tinkering had happened on Blackberries and other predecessors most of which had niche use cases.

I often find myself saying to entrepreneurs that you “can’t push on a string.” It may be a tired cliche but it seems highly relevant for consumer products. If people aren’t ready for it, no amount of hype or spending will make the product stick. The people who are ready now are the hackers and tinkerers. So why not embrace them instead?

Google Continuing to Invest in Deep Learning

Google just acquired DNNresearch, a spinoff company from the University of Toronto led by Geoffrey Hinton. The DNN in the company name stands for deep neural networks. As I have been learning from Professor Hinton’s online course, these are networks with multiple “hidden” layers of neurons. A hidden layer sits between an input and output layer. Hidden layers are responsible for much of the power of modern neural networks and the breakthroughs that have occurred are how to train these hidden layers (something that our brain does magnificently well).

I can highly recommend Hinton’s class. Even if you just watch the first couple of lectures to learn more about the recent advances in neural networks. Hinton has had an unwavering commitment to researching neural networks even at a time when it wasn’t popular. With all the hoopla surrounding neural networks these days its easy to forget that there was a long period where neural network researchers labored in relative obscurity (fun aside: in looking up Hinton, I discovered that he is the great-great-grandson of George Boole).

Leistungsschutzrecht (Germany’s Google Tax)

Germany’s parliament today passed the highly controversial “Leistungsschutzrecht” which is an ancillary copyright that requires payment for excerpting from and linking to content in some instances. When I walked around in a slight daze at Chicago O’Hare this morning after an all too short red eye flight, my Twitter feed was full of German friends and others outraged about the bill (just look for #lsr). Jeff Jarvis — who spoke out strongly against the law at this year’s DLD conference — in particular was tweeting up a storm.

I agree that the LSR is the wrong approach because it is a blunt instrument that will have significant collateral damage. Excerpting and linking are the lifeblood of the web. But it would be naive to think that there isn’t a problem here that needs addressing. It is very different if I excerpt and link here on Continuations than when Google does it. Why? Because I account for some vanishingly small faction of Internet traffic whereas Google often has 80 percent or more of all search queries.

When Larry Page took over as CEO of Google I wrote that I hoped he would address the problem of Google’s role in the Internet marketplace by figuring out how to share some of Google’s economics. Because that’s what this is all about. It’s about trying to fix rent extraction by a near monopolist in search. And that rent extraction is a very real issue (how do you think Google is financing self driving cars and Glass?).

Imagine for a moment a situation in which there are many competitive search providers and consumers switch easily between them. Now consider a company such as Yelp that has accumulated a lot of information about restaurants. If one of the search engines starts to publish more and more of Yelp’s data in the search results instead of finding the listing and driving traffic to it, Yelp could simply choose to not let that search engine crawl and index Yelp. That becomes a credible threat as consumers might switch to a different search engine if they can no longer find high quality restaurant data. That credible threat would over time lead to a situation where companies such as Yelp receive a licensing fee from the search engines for inclusion of the content, whenever including the content directly creates more value rather than driving traffic.

But with the market structure in search as it is, Google has been holding on to all of the value creation from including ever more content in its search results. All the while adding insult to injury by starting to build or buy competitive content services. That is not sustainable and Google needs to change its practices or we all risk seeing more legislation like the LSR that in trying to fix a market structure problem in search is likely to do more overall harm than good.

Google Glass

Despite my sarcastic tweet from last night, I am actually intrigued by Google Glass (and I even believe that Google Plus will matter as Google is all in). First, as Fred pointed out yesterday, Google’s machine learning capabilities are a key enabler for making Glass useful. You need voice commands to change Glass is doing or displaying and Google’s speech recognition is exceptionally good. Second, only a technology such as Glass will make augmented reality a viable experience. Holding up a phone is simply not compelling. Third, the success of GoPro shows clearly that people have a desire to record their experiences. When I was just skiing in Alta, my guess is as many as 1 in 20 skiers was wearing some kind of camera. Fourth, it’s not hard to come up with real world applications. For example, I recently drove a BMW with a head up display for navigation that was excellent and Glass could provide that experience (it will be interesting to see what the safety implications are). Of course the most likely early use cases will be for police and military but eventually I can see many specific professional applications (eg construction, medecine, etc). So good for Google to be pushing the envelope here and even showing off that they know how to make Twitter work for them in promoting Glass!

One Graph (Search) To Rule Them All?

Yesterday Facebook announced its long awaited search offering which they are calling graph search. It is the logical way to make the graph data that Facebook has been accumulating on its own and through its open graph accessible to users. Graph search is launching in beta with the obvious types of searches: people based on interests, places, movies, books, music, etc based on people. This is of course what Google has been fearing and why Google has been doubling down on Google+ at every turn (and I happen to agree with Brad Feld that Google’s patience has the potential to pay off).

At this point then it couldn’t be clearer that Facebook and Google are on an epic collision course. What is disappointing to me is that both are pursuing a vision which seems counter to the spirit of the internet as a collection of small pieces loosely joined. As an enduser I would much prefer a rich ecosystem of competing smaller services that work on specific problems, such as foursquare and yelp for places, netflix and vudu for movies, rdio and spotify for music, etc. The desire for a network of networks has also informed how we have invested.

I continue to believe that smaller services focused on specific areas can ultimately deliver a richer experience for endusers and also avoid a concentration of power. But the individual players are up against the powerful effects of supermodularity of information based production functions. On the margin I can make a better recommendation for say books if I also have information on movies. The big question then is whether the resulting complementarities doom us to face a few highly integrated players or whether it is possible for independent services to be sufficiently differentiated as to offset this advantage.

Some Points re Sparrow

There has been a flurry of posts about Google’s acquisition of Sparrow.  As I was reading the posts, I felt that a couple of important points were missed (or at least didn’t jump out at me).  Let’s start with the discussion of whether users have a right to be upset.  No matter what view you ultimately take here it seems worth remembering that the dollars spent on the software are not the only costs that were incurred by users.

First there is the cost of time.  Let’s say that over the course of a few days after you pay for Sparrow you spend a total of an hour setting up your email accounts and getting up to speed on the product.  If you value your time at $50 per hour, then your total cost of Sparrow has just gone up 5x on Mac and 15x on iOS. And I think many of the early adopters of Sparrow value their time at more than that.  Even if you don’t put a strict dollar value on it people generally hate feeling like they wasted time.

Second, the emotional cost: there is a strong emotional tie to key tools.  There are epic pieces on the Internet about text editors.  Just search for “love vim”!  Email is for most people what editors are for developers.  So to fall in love with a new tool only to then find out that it will be “unsupported” which essentially means outdated soon and bugs not fixed (no piece of software the size of Sparrow is without bugs) can be a bit crushing.  I am sure people’s reaction would be quite different if this was some disk optimizer utility that you use once a year.

There was also at least one post arguing that the acquisition somehow calls in questions whether you can succeed with paid software.  I don’t think it questions that at all.  It only proves that whatever Google offered was more attractive than the alternatives the team considered.  The Sparrow team chose to go after a particularly hard problem and so we should read especially little into their choice.  In case you are not convinced that a paid email client is hard, please consider the following:

First, you are competing with free offerings which are heavily subsidized by big vendors with huge resources and are seen as integral by those vendors to their platform strategies.  So the bar for being better is actually reasonably high (I know a lot of people hate their email but that doesn’t mean improving it is easy).  Add to that the cost of time argument from about and you will see that getting people to try a new email client is hard.

Second, email itself is hard.  Anyone who has ever dealt with the intricacies of the standards involved knows that there are tons of weird edge cases and odd implementations out there.  Getting an email product to work well at all times and with all accounts is hard even if you are leveraging existing open source components.

Third, going after a general purpose application for a broad audience is hard.  How do you build differentiated value?  How do people find that they want to use Sparrow as opposed to what they already have?  All of this get easier when you make a more niche application where figuring out what drives value and who might be willing to pay for that value is easier to identify.

Fourth, marketing diffuse benefits such as elegance and simplicity is difficult.  Here I think the Sparrow could have maybe tried to be more aggressive.  Some Sparrow users I know claimed that Sparrow saved them an hour or more of time per week.  If that’s indeed the case then more hard charging marketing copy and a much higher price point might have both been justified (this is cost of time again from the very first point).

So all in all I think that drawing conclusions about the viability of the paid model in general from the Sparrow acquisition is vastly overstating the case.  And as my last point suggests, it may even be possible to make this work for email — it’s just a particularly hard problem.

Chrome for Android

Google has been showing off a lot of neat goodies at Google I/O.  But the most interesting announcement as far as I can tell of the day is the availability of Chrome for Android. The existing Android browser is relatively wonky and not exactly fast.  It also is surprisingly behind the iPhone browser in the adoption of some of the HTML5 javascript APIs.  Chrome could really change this equation and as a by product make HTML5 apps a reality far more quickly.  My best guesstimate is that we are still 1-2 years away but that time will go by fast.

Revisiting the Groupon-Google Deal That Wasn’t

In December 2010 I wrote a post saying that Google shouldn’t buy Groupon but that it would be a good idea for Groupon to sell at the then rumored price of $5.3 billion with a $0.7 billion earn out.  As of Friday’s close, Groupon’s market cap is $6.3 billion or 5% above what the full deal consideration would have been.  So should they have sold back then?

It is still far from clear, especially now that important details around the deal have emerged from an article in the WSJ.  The key point from the article is that a deal might have taken as long as 18 months to close due to scrutiny by regulators.  Google was willing to offer a $800 million break up fee to account for that but it is still a monstrous overhang and is an issue for Google in acquiring companies of scale.

Groupon did succeed in going public and is an independent company now.  So one way to look at the decision is trading off execution risk against regulatory risk.  While execution risk is huge it is to some degree more controllable than regulatory risk.  It does seem though that Groupon may have relied on some overly optimistic projections on how much better it could do using data going forward (as Business Insider found the researcher who came up with that estimate had just joined Groupon from Google).

We will soon be able to read more about how the deal almost happened and the decision making around it when the book “Groupon’s Biggest Deal Ever” comes out tomorrow.  Should be a great case study in decision making under extreme uncertainty.

Supermodularity And Service Bundling

This will be a bit of a wonky and short post with a longer and less technical one to follow some time soon.  Google has just announced a coming update to their privacy policy which will essentially make it possible for Google to integrate all the information it has about a user across its many different services.  This comes at the same time as the revelation that Larry Page apparently explicitly stated the goal of building “a single unified, ‘beautiful’ product across everything.”

While one can come up with many possible verbal explanations for why Google might want to go this direction, there is some powerful math that lies at the heart of it: supermodularity.  Here is the definition:

A function

is supermodular if

for all x, y  R^k, where x  y denotes the componentwise maximum and x  y the componentwise minimum of x and y.

If a production function is supermodular then x and y are strongly complementary.  If you want to read the bible on this consult Don Topkis “Supermodularity and Complementarity.”

A firm such as Google for which the production function relies almost exclusively on information (yes, there are servers and people as well) will exhibit super modularity almost by definition.  Why?  Because if X and Y are different information vectors, then as long as they carry some joint signal, the inequality will be met as you can always choose to discard additional information (meaning you always have access to the component wise minimum).  In plain English: if you have access to both the search history (X) and the social graph (Y) of a user, you can always “do better” than two separate services that only have access to one of these respectively.