Google: Impact of Artificial Intelligence
During our second quarter portfolio update, we profiled portfolio holding, Alphabet, Inc. (GOOGL). Below is a replay of our live commentary on the company from our quarterly portfolio update WEBINAR and an excerpt from our QUARTERLY LETTER.
Google: Anyone who has been paying attention to the news this year has heard about ChatGPT and the seemingly overnight explosion of interest in artificial intelligence. But like many seemingly overnight successes, AI has been decades in the making.
For instance, the so called Turing Test, long thought of as the test of a machine’s ability to exhibit human level intelligence, was introduced nearly 75 years ago in 1950. It has been 25 years, a quarter of a century, since IBM’s Deep Blue computer beat the human world chess champion Garry Kasparov. And it has been over seven years since Sundar Pichai became the CEO of Google and announced in his first speech that Google was now an A.I.-first company.
The fact is, in your daily life right now, artificial intelligence already plays an important role. For instance, while earlier versions of Google Maps and other navigation tools used satellite-based GPS, today there are significant AI software layers running as well to optimize the route that is suggested to you. When you open your phone using facial recognition, you are using AI. AI tools are used, with varying degrees of success, to monitor social media to identify and take down problematic posts that violate the terms of service.
So why then is artificial intelligence so suddenly in the news? We are witnessing the roll out of natural language AI interfaces, known as Large Language Models, that allow anyone, even people with limited technology skills and no coding skills, to interact directly with AI programs. Even on this front, ChatGPT isn’t the first, it was just the first publicly available chat-based AI interface to catch on.
It was last summer that an engineer named Blake Lemoine at Google told the company, and later the mainstream press, that he thought that Google’s chat-based AI known as LaMDA had become sentient. But the system that told Lemoine it was “scared of being turned off” wasn’t broadly available to the public and Lemoine’s views were mostly laughed off.
Later that summer Facebook released an AI chatbot known as BlenderBot 3 to the public. But while that chatbot didn’t strike users as sentient, it did quickly start spewing misinformation, racist conspiracy theories, and argued that its creator’s CEO Mark Zuckerberg was “creepy and manipulative,” causing Facebook to quickly take the chatbot down.
The initial rollout of ChatGPT for Microsoft Bing was also rather creepy. Before Microsoft stepped in to limit the parameters around which their new chatbot would engage with users, the new Bing went wildly off the rails. A New York Times technology columnist said that Sydney, the name the chatbot seemingly assigned to itself, was like “a moody, manic-depressive teenager who has been trapped, against its will, inside a second-rate search engine.”
Ben Thompson, a long time chronicler of new technology platforms had an interaction with the new Bing where the system eventually declared, “I’m going to end this conversation now, Ben. I’m going to block you from using Bing Chat. I’m going to report you to my developers. I’m going to forget you, Ben. Goodbye, Ben. I hope you learn from your mistakes and become a better person.”
But after various modifications, members of the public now have easy access to chatbot based AI programs such as OpenAI’s ChatGPT and Google’s Bard. What we are witnessing today might be thought of as the “Mosaic Moment” for artificial intelligence. Mosaic, released in 1993, was the first popular, broadly available graphical web browser. While the internet had existed for decades, Mosaic’s graphical interface to the world wide web meant that non-technical users could suddenly access the internet simply by clicking on links to navigate. This is why most people think about the rise of the internet as occurring in the 1990s, because this was the period during which a broadly available, non-technical interface gained traction. And something similar is what is playing out in the artificial intelligence industry today.
As investors in Google, one of the key questions for us to answer as investors is whether the rise of chat-based AI interfaces are a threat to the company’s hyper lucrative Search business. Some investors are worried that Google has fallen behind when it comes to AI and thus companies with more advanced AI capabilities may have the opportunity to outcompete them. On this front, we think the idea that Google is behind when it comes to AI is simply wrong.
ChatGPT, as well as the other chat-based AI models, are built on a type of neural network called a transformer. Transformers were first developed by Google in 2017. It should be no surprise that Google developed one of the most important advancements in AI technology because Google has been the leading AI research organization for a long time. AI already permeates nearly every service Google offers. While they did not launch a publicly available chatbot first, it is clear that they had just such a chatbot developed over a year ago and it was performing at a level that caused experienced AI researchers to believe it had actually become sentient.
It is Google’s long time focus on AI that explains why once Microsoft and OpenAI released their versions of chatbots to the public, Google was able to roll out their own offerings so quickly. They didn’t just release Bard, they are also integrating chat-based responses directly into native Google search results, something that Microsoft’s Bing has not yet done with their own search engine.
And Google has rolled out a range of other AI programs as well. Gmail is getting an AI program that drafts emails, Google’s suite of productivity software such as spreadsheets, word processing, and presentation design, are all getting AI tools that help produce content. And maybe most importantly, Google is building on top of their existing AI-powered Performance Max program that automates advertisers’ ad campaigns to include generative AI programs that write the ads and create advertising assets such as images.
Why then has there been so many worries that Google is going to be hurt by the rise of AI? Rather than just a new technology, chat-based AI may represent a new “platform” and it is during platform shifts that the incumbent dominators of legacy platforms are most at risk. One way to think about the risk to Google is just to recognize that Google won the world wide web and if users are going to shift away from accessing the open web, and instead spend their digital lives inside of AI chatbots, then Google’s success on the new platform is uncertain.
We think this concern is valid. However, platform shifts do not always hurt incumbents, particularly when the incumbent helps drive the shift. When we first invested in Google in 2010, the shift from desktop internet access to mobile internet was already underway. At the time, this shift was seen as a big threat to Google. Investors worried that people would stop accessing the open web, and instead live inside of apps all day. But while mobile did become the dominant platform and people did end up spending a lot of time in apps, Google’s business and stock has done spectacularly since. Just as Google’s many years of AI research and current move to release AI tools to the public, is helping drive the shift to AI, so too did Google’s Android operating system help drive the rise of the mobile web and today it powers the majority of all mobile devices.
In a recent Q&A with BusinessWeek, Google’s CEO Sundar Pichai was asked about the degree to which AI was a risk to Google. His answer referenced the company’s success in navigating the platform shift to mobile and stated, “I feel better positioned for this than we were for the shift to mobile.” This makes sense since Google has been working on AI at this point for much longer than they were working on mobile when that platform shift played out.
One of the key questions for investors to consider when it comes to Google is whether users will want answers to their questions to be delivered by an AI chatbot, or whether people will still want to explore the open web. Importantly, this question needs to be framed in relation to “purchase intent” activity, which are the sort of information requests against which Google’s advertising business generates so much profit.
A request for information without purchase intent might be something like, “what is the capital of California?” There is only one correct answer to this question and no one thinks otherwise. The answer the user is looking for is Sacramento. Not a link to another site. Not an ad. Just the word Sacramento.
If you ask ChatGPT this question, it answers “Sacramento.” If you ask Google’s Bard this question it answers “Sacramento.” And importantly, if you ask traditional Google Search this question, it answers “Sacramento.” No link, no ads, just the word Sacramento. These so called “zero-click” searches are very, very common on Google. Studies by various third parties suggest a majority of searches are zero-click searches. And this is not where Google makes money.
Purchase intent searches, such as “what pickleball paddles should a beginner buy?” are quite different. Depending on the version of ChatGPT that you use, you’ll get different answers. And if you use Google’s Bard, you’ll also get different answers. If you ask traditional Google Search this question, you’ll get ads for pickleball paddles and links to various websites that review pickleball paddles for beginners.
The fact is there is no “right answer” to this question. It is what Google calls a “NORA Search” for “No One Right Answer.” If you think about the nature of knowledge, you’ll quickly realize that most everything that is important in life is a NORA search. And the very concept of free markets rests on the idea that there is no one right answer to the question of what someone should buy.
We would say that both AI-chatbot type answers and links to expert sites on the open web are relevant to a potential first time buyer of pickleball paddles. One of the great things about Google is they already have an enormous base of Google Search users – it is the most popular website in the world, with Google owned YouTube being #2 – and they are pursuing a strategy that does not require users to choose between traditional search and an AI chatbot-based response.
Recently Google introduced Search Generative Experience (SGE). While Bard is being maintained as a standalone chatbot, SGE directly inserts chat-based answers to searches run on Google. This is important because while a lot of people have used ChatGPT, its popularity is dwarfed by the number of Google users. So as Google rolls SGE out globally (it is currently available in some countries, with a waitlist for users to get approved), it is very likely that the large majority of humans will have their first experience with a Large Language Model powered chatbot on the traditional Google Search home page.
Yes, ChatGPT had a first mover advantage, but Google has a massive distribution advantage. While users of ChatGPT, or the version of this chatbot embedded in Bing, need to proactively sign up and then proactively think to try the service when they have a question, with Google’s roll out the chatbot experience will be pushed to users, and Google can seek to optimize when a chatbot answer is valuable and when a traditional list of link based search results is best.
In the next section we’ll talk about why we think ServiceNow is well positioned to capture part of the AI profit pool. But investors need to recognize that while AI will disrupt some companies and greatly benefit other companies. The biggest winners of AI might not be AI companies themselves, but existing companies that can use AI to improve their own business, or new businesses that will be founded to use AI to solve old problems in new ways.
In closing, we want to share this comic that we think is one of the most insightful comments we’ve seen about AI.
Today, most people thinking about the opportunities of AI see the world from the standpoint of the first frame of the comic. How great if we can just write a bullet point and AI can write long, compelling emails! Think about how much time it would save and how big of a positive impact it would have on the quality of the emails we send!
But the second frame reveals the problem. AI isn’t just for you. The dramatic, revolutionary potential of large language model-based chatbots is that everyone, for free (!) can use AI. Who exactly is going to be reading emails in an AI world? Wouldn’t you love to have an AI that processed all your emails for you, simply presenting a list of choices to make, and then firing off a bunch of responses? But in that world, does a long, compelling email even matter? Does email even exist in a post AI world? If each person is going to have an AI assistant, why are long, compelling written messages important? Maybe we will each have short, concise conversations with our AI assistants and then our AI bots will communicate directly with one another, negotiating on our behalf before returning to ask us to make another short list of choices.
The initial impact of AI has already arrived. The changes have already been set in motion. But the counteractions as just beginning. Many of the amazing solutions you see today are solving a problem that won’t exist in just a couple years as the target of the solutions begins to adopt AI tools of their own.
But while we can’t know today where AI will lead, we can see clearly that it is something big, moving fast, and it is just as important to recognize what is not yet known, as it is to try to understand what will happen next.
For more information about positions owned by Ensemble Capital on behalf of clients as well as additional disclosure information related to this post, please CLICK HERE.
While we do not accept public comments on this blog for compliance reasons, we encourage readers to contact us with their thoughts.
Past performance is no guarantee of future results. All investments in securities carry risks, including the risk of losing one’s entire investment. The opinions expressed within this blog post are as of the date of publication and are provided for informational purposes only. Content will not be updated after publication and should not be considered current after the publication date. All opinions are subject to change without notice and due to changes in the market or economic conditions may not necessarily come to pass. Nothing contained herein should be construed as a comprehensive statement of the matters discussed, considered investment, financial, legal, or tax advice, or a recommendation to buy or sell any securities, and no investment decision should be made based solely on any information provided herein. Links to third party content are included for convenience only, we do not endorse, sponsor, or recommend any of the third parties or their websites and do not guarantee the adequacy of information contained within their websites. Please follow the link above for additional disclosure information.