“Personally, I don’t want the perception in a few years to be, ‘Those old school web ranking types just got steamrolled and somehow never saw it comin’…’”
Google engineer Eric Lehman, from an internal email in 2018, titled “AI is a serious risk to our business”
I should, of course, have put a question mark at the end of the title of this, but I very much do not want to fall foul of my own law. And, of course, talking about the end of the line for Google as a company is like talking about “the end of the line for IBM” in 2000, or “the end of the line for Microsoft” in 2008. Whatever happens, Google has so much impetus behind it, so much revenue, that a quick collapse is about as likely as my beloved Derby County winning League One, Championship and Premier League in three consecutive years. It isn’t happening, much as I might dream.
This is one of the reasons I quipped that Google could see the $2.3billion that Axel Springer and other European media groups want for its alleged monopolisation of digital advertising as “just the cost of doing business.” It’s the equivalent of someone having to pay a £250 fine for speeding: annoying, but not the end of the world, and not actually that likely to keep you down to under 70mph in the future.
Google’s problems, though, do run deep. Other than, as my friend Cory Doctorow has noted, the 1.5 good products it invented itself (“a search engine and a Hotmail clone”), the most successful Google products are acquisitions. Android? Acquired. YouTube? Acquired? Adtech? Acquired. Even Chrome, which dominates web browsing in a way which many people (including me) find more than a little scary, was based on Apple’s WebKit rendering engine – which was, in turn, based on the open source KHTML.
The fact is, Google is incredibly bad at successfully bringing products to market, to such a degree that no one trusts them to do it and stick with it for long. It continually enters markets with fanfare, only to exit not long after.
Take social networking. You probably remember Google+ (2011–2019). You may even remember Orkut (2004–2014). Perhaps you know about Google Buzz (2010–2011). But do you remember Jaiku, an early Twitter competitor which Google bought – and buried? The resources of Google could have been used to accelerate Jaiku’s development and – perhaps – win the battle against Twitter and the nascent Facebook. Instead, the company took two years rebuilding Jaiku on top of Google’s App Engine, with no new features or marketing spend to support the product. Two years later, they killed it.
What Google is pretty good at is producing research. Its 2017 paper on transformers directly led to many of the large language model breakthroughs which OpenAI used to create ChatGPT. Failing to spot the potential for its research isn’t unknown in technology history, but really great companies don’t allow others to turn themselves into competitors worth $80 billion on the back of it.
And particularly not when those other companies create technology which directly threatens core businesses, in this case, Google’s “one good product” – search. The bad news for Google is that even in the middle of last year, research showed people using ChatGPT for search tasks performed just as well as using a traditional search engine, with one exception — fact checking tasks. That, of course, is a big exception, but ordinary people use search engines for a lot more than just checking facts.
What’s also notable about the same research is that ChatGPT levelled the playing field between different educational levels, giving better access to information for those who have lower educational achievement. That strikes at the heart of Google’s mission statement, which promotes its goal of “organis[ing] the world’s information and making it universally accessible and useful” (my italics). Search, as good as it is, has always required the user to adapt to it. Conversational interaction models, which ChatGPT follows (the clue is in the name), change that profoundly.
In The Innovator’s Dilemma, Clayton Christiansen talks about the difficulties that successful companies have in sustaining innovation. Established businesses, he notes, are excellent at optimising their existing products and processes to serve current customers (this is called “sustaining innovation”). However, they often struggle when faced with a “disruptive innovation” – a new technology or business model that creates a whole new market and customer segment.
One of the potential solutions to this which Christiansen looks at is structural: Creating smaller, independent units or spin-offs tasked with exploring the disruptive technology can allow them to operate outside the constraints of the main company. This, of course, is probably what Google intended to do when it changed its structure to create Alphabet, a group of companies of which Google itself is just one part.
The biggest problem with this putative solution is that if you do it well, innovation doesn’t necessarily flow to where it is most needed. Google’s search products needed to seize on the research made in 2017 and integrate it. It didn’t, and – worse still – no one saw this as a potential disruption of the core business. The blinkers were too firmly on.
Perhaps that’s changing. Notably, last year that Google moved all its AI efforts into a single group, Google DeepMind. The “Google” in its name is significant: previously DeepMind was a separate business within Alphabet (and, in true Google style, it was acquired rather than built in-house). Now, on the surface, it looks likely to focus more on continuing Google’s mission, which means disrupting the traditional ten blue links.
Can it succeed? I’m not optimistic (publishers, take note). What we have here is a company which is good at research, but not at exploiting it; whose history is of one good product and a good Hotmail clone; that has a terrible record of announcing, releasing, and killing products, often multiple efforts in different categories all of which fail; and which has failed to keep its core product – search – up to date.
Perhaps the real question isn’t whether Google has reached the end of the line, but how exactly it made it this far?
