How Network Effects Shapes AI: Updates on AI and Technology by Outreinfo
OpenAI’s introduction of ChatGPT last year drew both praise and concern from industry observers. It was hailed for its potential to disrupt various professions such as computer programming, teaching, financial trading and analysis, graphic design, and art. Universities scrambled to update their curricula, fearing that AI would render college essays obsolete. Among the immediate impacts predicted was the potential reinvention or even replacement of traditional internet search engines by ChatGPT. This raised questions about whether chatbots could pose a threat to Google, considering that search and related ads constitute the majority of Google’s revenue. How network effects shape AI further complicates the situation, influencing these developments and their implications.
While ChatGPT showcased remarkable machine learning capabilities, it was not a viable standalone service. OpenAI recognized the need for a partner to leverage its technological prowess, leading to a swift collaboration announcement with Microsoft. This union between the AI startup and the legacy tech company presents a credible challenge to Google’s dominance, intensifying the competition in the “AI arms race.” It also provides va’uable insights into the factors that will determine the success or failure of companies in deploying this technology.
To comprehend OpenAI’s decision to align itself with Bing and evaluate whether Google can still maintain its supremacy, it is crucial to understand the unique characteristics of this technology compared to past developments like telephones or market platforms such as Uber or Airbnb. In each of those cases, network effects, where the value of a product increases with its user base, played a pivotal role in shaping the growth and success of the respective companies. Generative AI services like ChatGPT are subject to a similar but distinct type of network effects. Managers and entrepreneurs must grasp the workings of these new AI network effects to formulate effective strategies for AI integration.
AI operates differently from traditional products and services due to its reliance on accurate predictions and suggestions. Unlike turning supplies into outputs, AI requires large data sets that need to be continually updated through customer interactions. To stay competitive, an AI operator must gather, analyze, and utilize data to offer predictions and seek feedback for improving suggestions. The value of the system is directly tied to the data it receives from users, leading to the emergence of data network effects.
Data network effects, also known as data-driven learning, are distinct from direct network effects seen in products like telephones or indirect network effects observed in platforms like Etsy or Airbnb. In data network effects, the technology’s performance and value increase as the user base expands. However, the value is not derived from the number of peers or the presence of buyers and sellers. Instead, it arises from the nature of the technology itself: AI improves through reinforcement learning, where predictions are followed by feedback.
As the AI system becomes more intelligent, it generates better predictions, making it more useful and attracting new users while retaining existing ones. With more users, the system receives additional responses, further enhancing prediction accuracy and creating a positive feedback loop. In this way, data network effects drive the growth and improvement of AI technology.
OpenAI’s partnership with Microsoft addresses the limitations of ChatGPT by leveraging the feedback loop of Bing users. By testing predictions and rating answers, the collaboration allows OpenAI to update and enhance ChatGPT. The next phase could involve Microsoft utilizing its extensive user data from Excel, PowerPoint, Word, and LinkedIn to further improve the AI’s ability to recreate various documents, benefiting office workers. This union enables OpenAI to overcome the challenges and refine its AI technology.
To maximize the benefits of AI, two key considerations emerge. Firstly, continuous feedback is paramount. The intelligence of an algorithm thrives on an ongoing stream of user reactions, current choices, and ratings of past suggestions. OpenAI recognized the necessity of connecting sophisticated models to dynamic data sources, serving as a reminder for AI entrepreneurs. Secondly, executives must institutionalize the meticulous collection of information. Valuable data can be found in various places, both within and outside the organization. By tracking consumer interactions, operational records, and external factors like weather, companies can greatly enhance AI predictions. Even infrequent data points beyond their control, such as keywords used by recruiters or consumer browsing patterns, hold potential value.
Finally, everyone should recognize the importance of the data they share, whether knowingly or inadvertently. Sharing facts and feedback is crucial for improving predictions, but it also allows others to capture the value of that data. Executives need to carefully consider whose AI stands to benefit from the data they share or grant access to, and in some cases, limit sharing. For example, when Uber drivers use the Waze app for navigation, they inadvertently provide valuable data to Google, the owner of Waze, regarding ride-hailing trip frequency and duration. This data could be invaluable for Google’s autonomous taxi operations. Similarly, when brands like Adidas sell on Amazon, they enable Amazon to estimate demand and price sensitivity across brands and categories, potentially benefiting Amazon’s private label offerings or its competitors. Executives can counter these risks by bypassing platform intermediaries, negotiating data access agreements, or maintaining direct customer contact. Collaboration through data exchanges, similar to how banks share creditworthiness data, can also be an effective solution. Understanding the dynamics of AI network effects reveals that early adopters can reap significant rewards, while followers, no matter how quick, may struggle to catch up. Access to AI algorithms and a continuous flow of data leads to cumulative advantages that are difficult to overcome. For executives, entrepreneurs, policymakers, and everyone else, the full potential of AI is yet to be realized, bringing both promising opportunities and potential challenges.
