Google's AI Domination Strategy: A Threat to OpenAI in the AI Gold Rush?
In this week’s PLUS!, I make the case that it will be Google, not OpenAI/Microsoft that builds the rails upon which we, the people, ride our AI trains.
w/WiserPLUS!
A Wiser! Reader sent me an article from The Economist this week about OpenAI (thanks Pat). It asked the question, “Is OpenAI the Next BigTech Giant?” The article makes the case that OpenAI is putting together the building blocks for the AI generation. Blazing the trail while others follow. The story made a compelling point, much of it I agree with. After all, who knows if OpenAI will join The Four, Five or Six, however you count BigTech?
But, where I diverge from the article is that, IMHO, the key player isn’t OpenAI.
It’s not OpenAI that’s building the picks and shovels at scale, its Google. Just like they did with Internet Search, they’ll do the same with AI. Make Google ubiquitous.
In this week’s PLUS!, I make the case that it will be Google, not OpenAI/Microsoft that builds the rails upon which we, the people, ride our AI trains.
If you’re a PLUS! subscriber, I hope you enjoy the article. If you’re not, click this button and sign up for only €1/week and enjoy a 7 day free trial.
w/Artificial Intelligence
In The Beginning
Over the past 25 years, we, the people, have gotten used to one thing on the Internet, and that is that it’s free. Nobody pays for email, although you can if you want extra things. Nobody pays (directly) for Search. Neeva tried to introduce paid Search, and failed. Social media is free, although the man-child is talking about charging “a small amount” to use Twitter. Final nail, coffin and all that.
My point is that we, the people, like that the Internet is free. We are willing to give up data and privacy in order to get all this free stuff that, to all intents and purposes, are public utilities.
The same is going to happen with AI. It’s a path that’s been trodden before…in Search.
Google’s Dominance in Search
The most significant tech antitrust trial in 25 years kicked off last week in the United States against the monopolist in question, Google. The last time it was Microsoft and the case of Internet Explorer. That was in 1997.
This lawsuit hinges on the fact that since 2010, Google has paid companies, such as Apple and Samsung, to make Google Search the default option on mobile phones and computers. These deals cost Google more than $10 billion every year, according to the prosecutors who argue that without these agreements, Google would not have been able to maintain its 90% share of the internet Search market.
Google's main defence is simply that no other tech company has made a competitive product that can beat Google Search. Which is a fair point. Google has indexed two or three times more of the Internet than anyone else. Their ability to target advertising produces a 20-30% better ROI compared to Microsoft’s Bing, the number 2 Search engine.
Google Search is a money making flywheel. The more people use it, the better it gets, and so more people use it. Last year, Google made $160 billion from Search, accounting for 57% of its total revenue.
But if Google is so much better, why do they still spend $10 billion keeping everyone else out? This is the essence of the DoJ’s case against Google.
OK, what’s this got to do with AI?
Google have built their defensive moat by making Google “a thing.” It’s a verb in its own right, just like “hoover” became the noun for vacuum cleaners. I even say “I’ll google that” even though I use DuckDuckGo and Bing. My point is that Google got to dominate the space because they built Search to be a free utility. They’ll do the same with AI.
Google got to dominate the space because they built Search to be a free utility.
Google’s AI-ification of Productivity
This week, Google announced a ton of new AI features for its free to use apps and Workspace products, their alternative product to the pay-to-use software suite that used to be called Microsoft Office. Google Workspace comprises Google Docs, Google Sheets, Gmail etc and user can now connect Google Bard to Google Workspace and other Google apps like YouTube, Maps and Travel.
Google’s latest update integrates Bard into its most popular services through features called "Bard Extensions."
These extensions allow users to leverage the power of AI by giving Bard tasks that bring together data from different parts of their Google account. For example, if you're planning a holiday, you can ask Bard to find dates from emails in Gmail, search for flights and hotel options on Google Flights, and even create a daily itinerary based on information from YouTube.
One of the notable additions to Bard is the "Double Check the Response" tool. After asking a question, users can hit the "G" button, and Bard will verify if the answers are supported by information available on the web. It will also highlight any information that it may have generated incorrectly. This makes Bard the first major AI tool that fact-checks itself in real-time, ensuring more accurate and reliable responses.
By the way, Google has made it clear to users that their private data will not be used to train the AI, and these new features are opt-in only.
I’ve been using Bard extensions over the last few days and my first impressions are that it’s really cool, when it works. It’s fair to say that this is obviously a first version and functionality is limited. However, when it works, it’s pretty damn cool. I’ll give you an example.
I had to find a couple of pieces of information about a long running insurance claim that I knew was in some emails. I just didn’t know which ones. As I settled down with my coffee and plate of digestives for a laborious search through each email, one by one, I decided to test the new AI features. I asked Bard to get me the information, and it promptly did just that.
Bingo, within a second, Google Bard had opened up my Gmail account, read all my emails, found the information I wanted, summarised it and presented it back to me in a formatted answer. For free (that’s the key point to remember which I’ll come back to.)
While this new version of Bard is still in its infancy, it does take us one step closer to the integration of AI into our daily lives (see story about Amazon’s latest AI updates to Alexa.)
OpenAI Is The One To Beat
OpenAI is the undisputed market leader in generative AI, at the moment. On performance, ChatGPT beats it closest rival, Claude, by a length. In terms of traffic, ChatGPT has 6 or 7 times more traffic than the second most used AI app, Character.AI.
OpenAI are building on that momentum by rolling out feature after feature and ChatGPT has become the “word” that defines the whole sector. There’s no question that OpenAI are at the forefront of the AI gold rush and they have the momentum and cash. Just not as much cash as Google. Despite reports that OpenAI is now making a $1 billion in annual revenue, that’s c270 x less than Google.
And whilst it’s also true that OpenAI has Microsoft as its largest backer, the unique structure of OpenAI may become its biggest constraint. In 2019, OpenAI created a "capped-profit company" within their non-profit structure, offering investors the opportunity to make 100 times their initial investment. However, rather than distributing equity, OpenAI distributes profit-participation units, which provide claims on future profits without ownership rights. The company has also expressed its intention to reinvest all profits until the board determines that their goal of achieving AGI (artificial general intelligence, which is different to generative AI) has been reached. Putting money into OpenAI is more like a "donation" than a conventional investment.
To remain appealing to investors, OpenAI has adjusted its profit cap and shifted to a model based on the annual rate of return. While the specific maximum rate is undisclosed, the profit units themselves can be sold on the market like standard equities. The company has already provided opportunities for early employees to sell their units.
They’ve raised around $14 billion in total to date, with a significant portion coming from Microsoft, which also provides OpenAI with the computing power it requires through its Azure cloud division. In return, Microsoft will receive a significant share of OpenAI's profits and gain licensing rights to their technology, which they can offer to their corporate customers, including many of the world's largest companies.
These deep-pocketed backers are crucial for OpenAI, as the company requires substantial capital to acquire the necessary data and computing power to continue developing increasingly intelligent AI models.
Considering the financial resources and profitable cash flow that Google possesses, it may be challenging for OpenAI, as a single-purpose startup, to compete on the same level. While OpenAI has made significant strides in generative AI and has gained momentum in the industry, Google's dominance and larger financial capabilities present a formidable challenge. Nevertheless, OpenAI's innovative approach and ongoing advancements in AI continue to attract attention and support from investors.
Which all means that it’s going to be hard for OpenAI, as a single purpose startup with one hand tied behind their back, to compete with Google. No matter how much you like Sam Altman, and he’s pretty likeable, the reality is that Google has much deeper pockets and a highly profitable cash flow that dwarfs theirs. Plus Sundar Pichai is also a pretty likeable guy.
Here’s The Thing:
Google has been the pioneer in generative AI. In 2017, they invented the “T” in GPT.
T is for transformer, the approach to AI development that fundamentally changed the way AI was being developed. But then OpenAI stole the march last November with ChatGPT. That led to a flood of alternative AI chatbots; Claude, Bard, Llama2, PI. Venture capitalists poured over $40 billion into AI firms in the first half of 2023, nearly a quarter of all venture dollars this year. Google needed to respond quickly.
Bard was their answer to ChatGPT and, frankly, it’s clunky, awkward and I hardly used it before this week. But that’s apparently about to change with Gemini, a soon to be released alternative that is reportedly better than GPT4. We’ll have to see about that!
Then comes Google’s solution for AI Search, which they call their Search Generative Engine (SGE), an awful name but with some impressive capability. At the moment it is restricted to users in the US, Japan and India and is in beta mode, but its purpose is to combine AI and Search results to improve the Search experience but without completely cannibalising the advertising revenues that comes from presenting the links that pay the most to be top of your Search results, even if there are better answers out there.
However, these new features from Google have struck a cord with me. In the same way that Zuckerberg leveraged a billion users on Instagram to launch Threads as an alternative to Twitter, Pichai has leveraged the 4.3 billion users that use Google’s portfolio of (utility) services to launch his suite of AI tools. And like Search, they’re all free to use. Google do have deep pockets remember!
They’re willing to spend $10 billion a year keeping Search the market default. How much are they willing to spend to make their version of AI the default?
It’s an unfair advantage, but you can’t blame them for exploiting it.
Further Reading
Google (Bard)
The Economist (Bard)
Gizmodo (Bard)
The Independant (Bard)
BeeBom (Alexa)