This blog is a summary of a presentation made by Rushi Luhar (Jeavio CTO) at TiECON East in April 2023.
You can download the presentation here. A video version of the presentation will be posted shortly.
Summary
Generative AI will transform software development by allowing developers to focus on core business logic, boost productivity, and improve efficiency of software delivery.
The shift to AI enabled software architecture will transform user interfaces, application logic and workflows across the software lifecycle. As we look to the future, AI-driven tools and LLMs will continue to advance and reshape the industry. However, understanding the risks and pitfalls of using these technologies is crucial for success.
The Impact of AI on Software Delivery
Generative AI has already had a significant impact on software development. By focusing on core business logic, developers can work more efficiently and experience greater fulfillment. AI-powered tools like GitHub Copilot, Uizard, Writer, and Notion are supercharging productivity throughout the software lifecycle, changing software architecture, and driving the creation of flexible applications.
Developers use AI for various tasks, from generating boilerplate code to finding bugs and creating test cases. AI-assisted tools are not only limited to software development but also design, marketing, content creation, and product management.
AI Will Change Software Architecture
AI also transforms software architecture, making user interfaces more flexible, application logic more understanding-based, and testing more probabilistic.
Applications built on top of Generative Models will use radically different architectures than we currently use.
- User Interfaces will go from Rigid UIs and APIs to Flexible Conversational Interfaces
- Application Logic will go from rules based workflows to leveraging LLM capabilties in understanding user intent that will drive the behavior of the underlying model.
- Data will go from being stored in databases to being used to train a family of ML models.
- Testing must account for the change from deterministic applications to those that behave probabilistically.
The future of AI in software development will involve giving LLMs “real-world” capabilities and evolving infrastructure and platform support from major tech companies and emerging providers. LLMs will become more open and accessible, enabling customization and deployment on various devices.
Risks and Pitfalls
However, there are risks and pitfalls associated with building on top of generative AI models. Limited understanding of how LLMs work, reputational risks, proving correctness, legal complexities, security threats, and intellectual property issues all pose challenges. Developers must remain aware of these risks and learn how to navigate them effectively.
Changing Nature of Knowledge Work
Working as a developer in 2025 will involve a blurred line between code, data, and infrastructure. AI-assisted coding, testing, and documentation will be the norm, and prompt engineering and AI wrangling will become essential skills. Small, distributed teams will optimize cohesion and productivity, engaging in rapid experimentation.
The changing nature of knowledge work will move from creation to curation as AI-generated content becomes more prevalent. Understanding the capabilities and drawbacks of generative models will be crucial for leveraging their productivity benefits.
Preparing Jeavio for an AI Driven Future
We see adopting AI as an opportunity to improve productivity, efficiency, and effectiveness of our services and accelerate Product Delivery for our Portfolio.
At Jeavio, we are preparing for an AI-driven future by building AI capabilities, deploying AI-enabled productivity tools, and working with portfolio companies to leverage AI opportunities.
References & Further Reading
- Eight Things to Know About Large Language Models
- Generative AI is here: How tools like ChatGPT will change business (McKinsey)
- SPQA: The AI-based Architecture That ‘ll Replace Most Existing Software – Daniel Miessler
- How Generative AI is changing the way developers work – GitHub Blog
- ChatGPT and Software Development – InfoWorld
- The Productivity Impact of AI Coding Tools – The Pragmatic Engineer
- What is Auto-GPT and Why does it matter – TechCrunch
- Explaining Reinforcement Learning with Human Feedback with Star Trek – Rushi Luhar
- From Creator to Curator: The Impact of Generative AI on Knowledge Work – Rushi Luhar
- All generated images in the deck created using Stable Diffusion via DreamStudio