I originally started this post by asking when did Google ever innovate as I would argue that from a product/solution perspective Google has never produced anything before anybody else or entered an under-serviced market with a truly game changing product.
Microsoft spent decades queering the word "innovation" until the conversation degenerated to this level.
Look up every time a Microsoft executive spouted the word "innovation" as one of their first-paragraph talking points. (Try not to break Google while doing so.) Every damn time "innovation" was used in the context of product innovation, as if that's the only (or main) kind of innovation worth paying attention to.
They did this because Microsoft was a spectacularly innovative company, but their preferred field of invention was "business methods (unethical)". They didn't really want people asking how they became worth hundreds of billions of dollars with very little product-side innovation under their belts. (The other explanation conveys a bitter whiff of anti-trust.)
There was at least a decade where Google was easily one of the most innovative companies ever founded.
The core innovation was figuring out that they could only monetize each search result to a fraction of a cent. Hence they had to deliver each search result for far less than a fraction of cent. Their innovations in algorithmic efficiency, parallel computing, computing at scale, data center management, hardware efficiency, and internal network efficiency are legendary, and with just cause.
On the revenue side, Google innovated running a sponsorship auction on every search result (I don't know this side very well myself, but Laszlo Bock talks about it in his recent book).
It's hard to say how this all rolls out internally, but Google remains formidable in the machine learning space. AlphaZero is one of the most jaw-dropping results of my lifetime, and I still remember 1972 with a fair degree of clarity. My family was so insular, I barely understood that hockey was part of the Canadian zeitgeist until the Canada–Russia series. Somehow that culture shock woke up my internal PVR, and I've been archiving my impressions of the greater world around me ever since.
Frankly, at this point, I'm not actually sure that Google should be in the business of innovating whole new product categories. When Yegge signed up, that was probably part of the mission that attracted him, but the world changes, and not every company grinds away in the same groove forever and ever.
Google's manta is "organize the world's knowledge" and the various product categories they developed were a means to an end. ML is far more instrumental to their long-term vision. Page understood from the beginning that ML is highly dependant on big data, so they innovated preferentially in product categories with an accrued data payoff (geography, the social graph, speech recognition, machine translation).
The problem for Google at this point is that they have become king of Passive Advertising Mountain. I'm using the word 'passive' here to mean that their advertising approach is largely based on influencing the lizard brain.
'Active' advertising (for which the word 'advertising' might not even be appropriate) is finding out where your users want to go in life, their deep goals and aspirations, and then helping to filter out everything that does contribute to this.
Filter bubbles are great, so long as the bubble is a mentorship cocoon with your own best interests at heart, rather than sculpted out of libidinous landfill.
99% of silicon valley is presently suffering from VC lock-in to libidinous-landfill business models.
The opposing business model hasn't been figured out, yet, so far as I can see.
One could envision a Guild system, rehashing MOOC technology, where a person signs a contract with a Guild, which acts as personal player agent, probably starting in early high school. Perhaps you assign a 10% cut of your entire future life earnings to your chosen Guild (your choice of Guild would put your present choice of U in the shade).
Then your Guild acts as an algorithmic filter on the internet (and all it contains) so as to bring to your attention what best helps you accomplish your life goals.
Of course, the VCs behind the curtain of the New Guild Order would go batshit insane if star earners started to drop out of the workforce at age 40 to devote five years to an Indian ashram. Sentiment detection algorithms would rapidly be developed and deployed to keep people centered and productive in their chosen material quest.
These inherent tensions make this business model more than a little hard to get off the ground.
If it worked, however, the social-media enhanced Gifted Guilders would soon possess or control half the world's total economy.
Even Google seems to know this, already, because they tend to hire people who are least captive to the distracting, demoralizing crap that presently funds their colossal business empire.
Huxley's Alpha-Plus elite are today's engineers with the best ad-blocking technology / personal discipline.
According to Boch, inside Google, every engineer has his or her Objectives and Key Results (OKR) prominently listed beside their name in the company director.
Their sustaining eyeballs at large have no such thing. I click on one video on YouTube where the thumbnail shows a little cleavage, and for the next week YouTube suggest is like a pesky Border Collie obsessed with frisbee in the park: huh, huh, huh, see more cleavage? huh, huh, huh ...
There's actually a little ... control beside each suggestion which allows the user to specify "not interesting". I've been slowly training YouTube that the sight of exposed breasts gives me the moral (or physical) heaves. I've so far had to click "not-interested" a few hundred times. Amazingly, YouTube now suggests fewer videos with cleavage thumbnails.
Google probably thinks I don't like cleavage (I suspect that's the level of sophistication it still has about its end users). From what I can tell, Google understanding that what I want is sharp boundaries between my different cognitive hats presently surpasses their most sophisticated ML algorithm. (If I want a second eye-full of some hot chick who floats past at random in my internet travels, Google image search serves my purpose admirably. Every so often, I let my tongue hang out for five minutes, and then back to business again. I'm at an age now where this is more about coming to terms with my own Jungian shadow than short-term gratification, but us old goats tend not to discuss this reality in the presence if callow, red-blooded lady killers. Your day will also arrive, modulo cocaine, alcohol, and potent painkillers.)
Innovation. Start with a problem that looks like it has no solution at all, and see where that takes you. (More will fail than succeed. May a deserving Johnny on the Spot finally win.)