In the second part of the conversation on the Emerj podcast, Tsavo Knott, CEO of Pieces, joins Daniel Faggella, Emerj CEO and Head of Research, to discuss the rapid progression of generative AI (GenAI) capabilities and their profound impact on developer markets and the future of software development. This discussion highlights both the limitations and the boundless potential of GenAI tools, offering valuable insights for business leaders and developers alike.

Tsavo also shares how systems that augment human work are transforming the roles of individual workers, making them less about performing specific tasks and more about becoming cross-functional assets within their organizations. He highlights how Pieces is at the forefront of this transformation, providing tools that help developers navigate and thrive in this evolving landscape.

The Evolution of Software Development

Tsavo begins by expressing his excitement about the evolution of machine learning and its implications for the products that Pieces is building. He emphasizes how the company aims to augment developer work streams, enhancing productivity and enabling the creation of unique experiences at a faster pace. All of this is driven by an increased demand for GenAI software solutions in various industries, a shortage of skilled developers, and the ever-rising complexity of software projects.

A key concept discussed is the "context window," which refers to the amount of background knowledge a model has at processing time. Tsavo explains that traditionally, context windows were very small, containing about 2,000 tokens, which is equivalent to a handful of files. In contrast, modern models like GPT-4 can handle around 32,000 tokens or roughly 300 pages of text. He summarizes that "in theory, more context is better."

Limitations of the Current GenAI Tooling

Discussing the constraints of the current machine learning and generative AI tools, Tsavo highlights the dual role of GenAI as both a search and authorship mechanism. He provides examples of authorship, including writing new code, authoring marketing content, and generating financial projections. While GenAI systems enable faster creation of digital assets, Tsavo warns that the quality of the output can be average due to inherent biases in the underlying data.

He cautions that while GenAI tools can produce millions of lines of code quickly, the quality might not be high. He underscores the importance of improving models in areas such as context windows, tokens per second, and output to keep up with future demands.

Growing Use Cases and the Potential of GenAI

Despite the challenges with Gen AI, Tsavo anticipates massive growth in server farms dedicated to powering AI systems and foresees significant advancements in on-device models and supporting consumer hardware. Tsavo identifies several fundamentals that will limit the ability to ship products with desired features and functionality, including supply, GPU, compute, and energy.

However, he believes that humans will continue to find creative ways to apply these models across various industries. He points out that there is an abundance of untapped data, including hundreds of years of information from Fortune 100 companies. Tsavo emphasizes that we are only scratching the surface of the power requirements needed to leverage this data.

Shift in the Role of Developers

Tsavo also anticipates a shift in the role of individual workers, with developers becoming more cross-functional across their organizations. He believes that this cross-functional behavior will naturally change the structure of organizations, flattening layers and making them more efficient. This, in turn, will enable the US to upskill its development workforce to the highest level, or "10x," helping it keep up with countries that have larger workforces.

Final Thoughts

The second part of the Emerj podcast with Tsavo Knott provided a comprehensive overview of the challenges and opportunities presented by generative IA in software development. As the industry continues to evolve, the insights shared underscore the importance of adaptability, the potential for GenAI to augment human creativity, and the strategic considerations for leveraging these advancements. At Pieces, we are committed to providing the tools and technologies that empower developers to navigate this changing landscape successfully.

Want to dive deeper into these topics? Tune into the full podcast episode for more expert insights from Tsavo Knott on the rise of the exciting phase of generative AI startups and other GenAI tools.