Will OpenAI End Google’s Search Monopoly?

Revolution and Evolution of Search

Many of us born before the Internet era remember life before search. To get an answer to a simple question we either asked someone with that knowledge, went to the library, or, used the Yellow Pages. The Internet changed everything. In just six years from 1995 to 2001 we went from bulletin board systems (BBSs), to websites that could be accessed with browsers, to link aggregators and online directories, to search. We saw how Yahoo and free volunteer-curated link directories like the Open Directory Project (ODP) became the dominant online “Yellow Pages” for the entire Internet. It was truly amazing to go into a directory you like and find a bunch of links to relevant websites with detailed descriptions. It was paradise. Then came the era of search engines. Yahoo also added a search window at the top of the page. But the page was heavy, slow, and inaccurate.

Then Altavista took the market by the storm – a simple one-line search interface with quick and relevant results. Many search engines competed for accuracy and coverage. If you could not find a link you were looking for, you searched on Yahoo, then Altavista, then on Bing. And then, seemingly overnight, we all switched to Google. It delivered relevant results and was snappy.

The Rise of the Google Search Empire

Google was founded in 1998 by Larry Page and Sergey Brin while they were Ph.D. students at Stanford University in California. They developed a new type of search engine called PageRank, which used the power of the link structure of the World Wide Web to improve the quality of search results. This was a significant advancement over existing search engines like AltaVista and Yahoo, which relied on keyword matching to provide search results.

Google quickly gained popularity and attracted a significant amount of funding, allowing it to expand its operations and improve its search technology. By 2000, Google had become the top search engine on the internet, surpassing its competitors like AltaVista and Yahoo.

In 2003, Google introduced a number of new features, such as the ability to search for images and news, and to personalize search results based on an individual’s search history. These features helped Google to solidify its position as the top search engine, and by this time, it had assumed a monopolistic position in the market.

Google continued to innovate and introduce new products and services, such as Google Maps, which was introduced in 2005. This online mapping service was a significant improvement over existing mapping services, and it quickly gained popularity among users.

Google Maps used a combination of satellite imagery, aerial photography, and street-level imagery to provide detailed and accurate maps of locations around the world. It also provided features like turn-by-turn navigation and real-time traffic updates, making it a valuable tool for drivers and commuters.

Google’s dominance in the online mapping market was further solidified in 2007 when it acquired the popular mapping service, Waze. This acquisition gave Google access to Waze’s large user base and its crowd-sourced traffic data, which allowed it to improve the accuracy and usefulness of its maps even further.

Today, Google Maps is the most widely used online mapping service, and it holds a near-monopolistic position in the market. It continues to innovate and introduce new features, such as indoor maps and augmented reality, to stay ahead of its competitors.

From Search to Knowledge Generation – The Era of Large Language Models

Google, as it exists today, has not changed much over the past 20 years. Yes, it is faster, with more features and coverage. It also created a monopolistic ecosystem that only Microsoft and Apple can compete with and only in some areas. But the basic functions and user experience have remained the same.

Then, in 2022, ChatGPT changed the paradigm again. Now, instead of searching for a link with a brief description, you can ask the platform a very complex question or even give it a task, and it returns a complete answer to a question or completes the task. No more links as output – what an experience! And, thanks to the generous support from Microsoft, which, surprisingly, made the system fast and snappy, you get the answer in near-real time. The interface, ease of use, and the entire experience using ChatGPT is phenomenal – OpenAI just defeated Google in its own game of being simple, fast, and useful. Even experts in generative systems who used GPT-3 for over a year, were awed by the performance of ChatGPT. My company is using transformers in both biology and chemistry platforms for target discovery and generative chemistry and these tools are used industry-wide. However, we use these models for narrow tasks, and in combination with other models. ChatGPT is just next level when it comes to text. It got so addictive that I co-published an academic article with ChatGPT as a co-author and used it to debate complex philosophical problems. And since Microsoft is now massively supporting OpenAI, this is likely only the beginning – it is a new revolution and the real disruption came from the ease of use.

The End of Google Search Dominance?

Let’s admit it. Most of us love Google products. I do miss Google when traveling in China. The world is not going to switch to Chat overnight and Google will have time to respond.

In fact, it was Google scientists who made seminal breakthroughs in transformer neural networks that paved the way for GPT-3. In 2017, at the Conference on Neural Information Processing System (NIPS, later re-named NeurIPS) Google scientists presented a seminal paper titled “Attention is all you need”. By January 2023 this paper was cited over 62,000 times making it one of the most cited papers in AI. And one of the greatest scientists in generative AI, Ian Goodfellow, the inventor of Generative Adversarial Networks (GANs), who left Google to work for OpenAI, and then went to Apple, is now back at Google’s AI subsidiary, DeepMind. Both Google Brain’s in-house AI team and DeepMind published multiple papers on transformers, and Google even launched a bot called LaMDA.

However, Google’s experiments were not well-received and Google’s own employees were the source of these problems. LaMDA was called out as racist by one engineer and then another engineer referred to it as sentient.

These challenges, lack of vision, and focus on productization of the large language models at Google, allowed OpenAI to take the market by the storm and the new deals with Microsoft will definitely serve as a warning signal to Google. Like Yahoo in 2000, Google got too big and too bureaucratic to work at the pace of a fast-paced transformative startup. Demis Hassabis seems to be focused on getting the Nobel prize, a guess based on the recent scientific advisory board announcement, and building Isomorphic labs, and DeepMind can not simply drop everything else and focus on developing user-friendly large language models. But we should expect Google to react and release some competing tools in 2023.

Prepare for the Battle of the Titans!

With the rising popularity of ChatGPT and Microsoft announcements on integration of OpenAI tools across its entire ecosystem, we should expect many teams specializing in large language models to switch from pure research to rapid application development. Google is likely to add GPT functionality into its search engine, docs, and assistant. We should also expect new GPT-enabled tools from Apple and Amazon. Likewise, we should expect Amazon to enable the entire AWS ecosystem with GPT functionality via its cloud. This race is certainly great for NVIDIA, which not only supplies the bullets to the war of transformer neural networks but is also developing cloud-based transformer solutions.

Regardless of who lands on top and gains the most users, 2023 will be a wonderful year for generative AI in every industry and Google is likely to win due to the deeper penetration of technology into the global population.

Huge Opportunity for Publishing Houses

Pretty much every AI company that pitched to a venture capitalist got “how much proprietary data do you have?” as the first question. And while ChatGPT demonstrated the power of the algorithm trained on publicly-available data like Wikipedia and repositories of books, it often fails to produce accurate responses to queries involving specific domain knowledge. This presents a huge opportunity for publishing houses like the Holtzbrinck Publishing group, which in turns owns Nature, as well as Elsevier and others. These publishing houses own the copyright on millions of full-text scientific articles and books. With the advent of transformers, these proprietary texts became significantly more valuable. We are finally getting into the era of data economics, especially in healthcare, and we should expect these publishing houses to develop tools to trace the origin of the generated content and invent new licensing models for their proprietary data. I would not be surprised to see the big tech firms acquiring these publishing houses.

The Wild Wild East – China to Further Separate, Evolve, and Grow in AI

With over 1.4 Billion people, China generates an enormous amount of data. And Google does not have access to this data in the same way it can access it in the West. Tencent, Baidu, Alibaba, and other massive technology players are now working on large language models and developing their own tools. Considering the abundance of unique training data in China and the availability of public data in the West, we should expect these players to develop very sophisticated knowledge generation tools and compete globally.

Epilogué: The Impact of Search to Generation Transition

The advancements in generative AI are staggering and will impact every industry on the planet. It is an exciting time to be alive as we will get to witness the “before” and “after” similar to the impact of mobile phones, the internet, and social networks.

As I write this article, I am confident that it will be used by large language models for training and later for generation. I also hope that Forbes.com figures out how to monetize this content, as generative models trained on its content will bring substantial economic benefits.

As competition in generative systems takes over search, this article will not be lost like many forgotten books, articles, and links on the internet that have been deprioritized or removed by the search era. Everything we do that gets recorded will be used for training and generation, and it is important for us to strive to be the best versions of ourselves to contribute high-quality content for the future of humanity. It is also a privilege to be a published author and Forbes contributor before the Great Generative Revolution, as it allows us to showcase original content created without the help of GPT. The value of each article is now exponentially higher as it will not only be read by a few thousand humans, but it will also be immortalized as part of many generative platforms around the world. It truly is an exciting time to be alive!