ChatGPT Revolutionizes Data Interaction: Unleashing A New Era Of Human-Data Dynamics

The iPhone completely changed the way people interacted with mobile phones when the first generation of Apple’s AAPL revolutionary product launched in 2007. Similarly, ChatGPT has transformed—perhaps for good—the way people interact with data. Microsoft MSFT has emerged at the forefront of this paradigm shift, triggered by OpenAI’s debut of ChatGPT in November of last year. The Washington-based Internet giant has already begun to integrate AI into its products. Microsoft’s New Bing, which the company introduced in February, features an AI-powered chatbot to aid users’ web searches. Microsoft has also exhibited its capacity to quickly integrate AI into existing products, with the addition of Azure Cognitive Services to the Azure cloud platform. While these efforts demonstrate that even legacy titans like Microsoft expect AI to revolutionize our increasingly online age, in fundamental ways, they have not engendered a fundamental shift. Indeed, the more AI dominates headlines, the clearer it becomes that our exploration of this potentially world-changing technology has only just begun.

Today, people mostly interact with data indirectly, requiring specialized platforms to access backend databases. For example, we need the Expedia platform to interact with the airline database to book a flight, and the HubSpot HUBS Customer Relationship Management platform to store and manage backend databases for inbound marketing, sales, and customer service. Why do we need these platforms, the product of countless engineers’ and data scientists’ labor? Unlike these specialists, the non-technical Internet user lacks the knowledge and the resources necessary to interact with backend data. These casual users, for instance, might not know how to upload and download data stored in AWS S3 through APIs. They might be perfectly capable, on the other hand, of using platforms like Google Drive and Dropbox to store and manage data.

Large AI models like ChatGPT have powerful language processing capabilities that give them human-like abilities in judgment and reasoning. As AI models evolve, AI will render obsolete most of the specialized platforms currently in use, allowing people to interact directly with data. That is to say, people will only need to upload data; then, they can simply use AI to do anything with it, without using any other tools and applications as they do today. For example, you only need to upload photos from your phone, and then tell AI what you want to do with the photos (for instance, find all photos of you and your family watching the sunset at the beach. This cannot be done by current photo management applications). People can further ask AI to produce a short video montage of their photos tailored for a certain mood or occasion. HR can simply upload all employee contracts and information in any format, then ask AI to export spreadsheets with well-organized employee information (name, starting date, salary, etc).

AI will continue to cover new ground, but several technical hurdles still need to be crossed before AI-powered data interaction can realize its full potential. These hurdles are not trivial. Take data privacy, a critical problem in today’s AI space. Companies are hesitant to use ChatGPT because they lack a thorough understanding of how the AI will handle their sensitive data. This is of particular concern when dealing with highly confidential data, such as clinical or scientific research. Additionally, the cost of using large AI models is a significant factor for businesses and organizations as they develop their AI strategies. The expenses include hardware and software infrastructure, high electricity costs for model training, and storage and access fees for big amounts of data used in training and inference. For instance, it costs a few million dollars to train GPT-3 with 175 billion parameters.

Despite these obstacles, we stand today on the cusp of a new era in data interaction. AI is poised to change the game, and Microsoft is leading the change. Someday, interacting with data will be as easy as uploading your data and then talking to it.