Network effects dictate the success of technologies from the phone to shopping platforms like Etsy, and AI tools like ChatGPT are no exception. What is different, however, is how network effects work. Data network effects new form. As with the more familiar direct and indirect network effects, the value of the technology increases as it acquires users. Here, however, the value does not come from the number of peers (as in the telephone) or the presence of many buyers and sellers (as in platforms such as Etsy), but from the feedback that helps it to create more good predictions. More users means more answers, which increases the accuracy of the prediction, creating a virtuous cycle. Companies should consider three lessons: 1) feedback is important, 2) routinize meticulous information gathering, and 3) consider the data you share, intentionally or not.
Late last year, when OpenAI introduced ChatGPT, industry observers responded with both praise and concern. We hear about how technology is being eliminated computer programmers, teachers, financial traders and analysts, graphic designersand artists. Afraid to do with AI kill the college essaythe universities rushed to change the curriculum. Perhaps the most immediate effect, as others say, is ChatGPT can be invented or replaced the traditional internet search engine. Search and related ads bring in most of Google’s revenue. Make chatbots kill Google?
ChatGPT is a unique demonstration of machine learning technology, but it can hardly be done as a standalone service. To adapt its technological expertise, OpenAI needs a partner. So we were not surprised when the company quickly announced a deal with Microsoft. The union of the AI startup and the legacy tech company could finally provide a credible threat to Google’s dominance, raising the stakes in “AI arms race.” It also offers a lesson in the forces that will dictate which companies will thrive and which will fail to deploy this technology.
To understand what forced OpenAI to ally itself with Bing (and why Google may still win), we consider how this technology differs from previous developments, such as telephone or market platforms like Uber or Airbnb. In each of these examples, network effects – where the value of a product increases as it acquires users – play a large role in shaping how products grow, and which companies succeeded. Generative AI services like ChatGPT are subject to similar, but different network effects. In order to choose strategies that work with AI, managers and entrepreneurs must understand how this new type of AI network effects.
Network Effects Work Differently for AI
The value of AI lies in accurate predictions and suggestions. But unlike traditional products and services, which rely on making supplies (such as electricity or human capital) outputs (such as light or tax advice), AI requires large sets of data to be keep fresh through interactions with customers. To remain competitive, an AI operator must corral data, analyze it, offer predictions, and then solicit feedback to improve the proposals. The value of the system depends on – and increases with – the data that comes from the users.
The technology’s performance — its ability to accurately predict and predict — depends on an economic principle called data network effects (chosen by others DATA–driven learning). It is different from the familiar directly network effects, such as those that make a phone more valuable as subscribers grow, because there are more people you can call. They are also different from indirect or second order network effects, which describe how more buyers invite more sellers to a platform and vice versa — shopping on Etsy or booking on Airbnb becomes more attractive when more sellers are present.
The network effects of the data are new form: As the more familiar the effects, the more users, the more valuable the technology. But here, the value does not come from the number of peers (like the telephone) or the presence of many buyers and sellers (like platforms like Etsy). Instead, the effects stem from the nature of the technology: AI advances through reinforcement learning, predictions followed by feedback. As its intelligence grows, the system makes better predictions, improving its usefulness, attracting new users and keeping existing ones. More users means more answers, which increases the accuracy of the prediction, creating a virtuous cycle.
Take, for example, Google Maps. It uses AI to recommend the fastest route to your destination. This ability depends on anticipating traffic patterns on alternative routes, which it does by drawing on data coming from multiple users. (Yes, data users are also suppliers.) The more people use the app, the more historical and concurrent data it collects. With heaps of data, Google can compare multiple predictions with actual results: Did you arrive at the time the app predicted? To perfect the predictions, the app also needs your impressions: How good are the instructions? As objective facts and subjective reviews accumulate, network effects begin. These effects improve predictions and increase the value of the app for users — and for Google.
Once we understand how network effects drive AI, we can imagine the new strategies the technology requires.
OpenAI and Microsoft
Let’s start with the marriage of OpenAI and Microsoft. When we tested ChatGPT in beta, we were impressed by the creative, human-like answers, but knew that it was restricted: It relied on a lot of data collected in 2021 (so don’t ask about new events or even the weather) . Worse, it lacks a strong feedback loop: You can’t ring the alarm bell when there are suggestions hallucinatory (The company allows a “thumbs down” response). Yet by linking up with Microsoft, OpenAI has found a way to test those predictions. What Bing users ask – and how they rate the answers – is crucial to updating and improving ChatGPT. The next step, we imagine, is for Microsoft to feed the algorithm the massive cloud of user data it maintains. While digesting countless amounts of Excel sheets, PowerPoint presentations, Word documents, and LinkedIn resumes, ChatGPT can get better at doing it, to the delight (or horror) of office dwellers.
There are at least three broad lessons here.
First, feedback is important. The benefits of AI will intensify with the constant flow of user reactions. To stay intelligent, an algorithm needs a data stream of current user choices and ratings of past suggestions. Without feedback, even the best engineering algorithms won’t stay smart for long. As OpenAI realizes, even the most sophisticated models need to be linked to constantly flowing data sources. AI entrepreneurs should keep this in mind.
Second, executives must painstaking information gathering is routine to maximize the benefits of these effects. They need to study common financial and operational records. Useful bits of data can be found everywhere, inside and outside the corporation. This can range from interactions with buyers, suppliers, and co-workers. A retailer, for example, can track what consumers view, what they put in their cart, and what they end up paying for. Collectively, these minute details can improve the predictions of an AI system. Even infrequent pieces of data, including those not controlled by the company, may be worth collecting. Weather data helps Google Maps predict traffic. Tracking the keywords that recruiters use to find resumes helps LinkedIn offer winning tips for job seekers.
Finally, everyone needs to be mindful of the data they share, intentionally or not. Facts and feedback are essential for making better predictions. But the value of your data will be caught another person. Executives must consider which AI will benefit from the data they share (or allow access to). Sometimes, they need to limit sharing. For example, when Uber drivers navigate using the Waze app, they help Google, the owner, estimate the frequency and duration of ridehailing trips. As Google considers operating autonomous taxis, such data will be invaluable. When a brand like Adidas sells on Amazon, it allows the retail behemoth to estimate the demand for brands (compared to Nike) and categories (shoes) and the price sensitivity of buyers. The results can be fed to a competitor – or benefit from the private label offered by Amazon. To counter that, executives can avoid platform intermediaries or third parties. They can negotiate access to data. They may try to maintain direct contact with customers. Sometimes, the best solution may be for data owners to band together and share a data exchangeas banks do when establishing ways to share creditworthiness data.
When you consider the network effects of AI, you can better understand the future of technology. You can also see how these effects, like other network effects, are even richer. The dynamics behind AI mean that early movers can be rewarded handsomely, and followers, however quickly, can be left aside. It also means that when one has access to an AI algorithm and a flow of data, the advantages accumulate over time and are not easily overcome. For executives, entrepreneurs, policymakers, and others, the best (and worst) of AI is yet to come.