Generative AI, particularly in the form of OpenAI’s ChatGPT, is set to revolutionize the software development process, according to Mark Porter, CTO of MongoDB. While there is speculation that generative AI may replace developers, Porter believes that it will actually enhance their productivity. Generative AI can assist with code writing, testing, bug detection, and faster documentation lookup, allowing developers to write code with better quality, speed, and completeness. Porter also mentions the potential of other AI models that have been around for 15 to 20 years, stating that they can transform the way developers write code.
During MongoDB.local, the company’s developer conference in New York, Porter shared his insights with ZDNET. He emphasized that AI is an acceleration of the developer ecosystem, predicting that more apps will be written as a result. MongoDB itself plays a crucial role in AI development, serving as the foundation for hundreds of companies building AI. The conference showcased new AI capabilities for the MongoDB database, attracting companies like Confluent, Hashicorp, IBM, and Amazon AWS, who explained their use of MongoDB in conjunction with their software technologies.
Porter highlighted the incorporation of vector values as a native data type in MongoDB. This allows developers to store context vectors, which are approximate answers to queries produced by large language models, in the database. These vectors can then be retrieved later using relevance searches to obtain precise answers. When a user poses a question to ChatGPT or another large language model, the vector of that question is stored in the database. By searching for vectors near it, relevant articles can be retrieved. These articles are then used to prompt the language model to answer the question with information exclusively from those articles. Porter also mentioned the ability of language models to summarize lengthy articles.
Porter stressed that AI and databases have distinct roles and should not be mixed. While language models are useful for certain tasks, they should not be integrated into online transaction processing systems. Instead, they should be used where they are most effective, while databases handle core data and metadata. MongoDB’s incorporation of vector functionality eliminates the need for data copying and managing separate systems, streamlining the development process.
In conclusion, Porter believes that AI will not replace developers. They will continue to be essential in understanding customer needs and making decisions about what to develop. Despite the potential of AI, developers will remain at the heart of the software development process.