_ It is time to stop learning the computer’s programming language, and get computers to learn our language. _Prepare think different - or should we say alien 👾 - in this CTO podcast with Stephen Wolfram (Founder & CEO of Wolfram Research). Find out why we need to use computational thinking to create code in a way humans can both write ✍️ and read 📖. Listen to find out:
Listen Here
0:00:00 - Why Computational Language Trumps Programming Languages 0:12:30 - Livestreaming software design meetings 0:13:07 - Live Streaming and Infrastructure Technologies 0:19:19- Software Development in a Chat-Based World (LUIs) 0:31:49 - The Intersection of Technology and Judgment 0:45:53 - The CTO Role in an AI-aided World 0:59:13 - LLM Technology and Language Building 1:06:29 - Computational Invention and Exploration
(0:00:00) - The Power of Computational Thinking
I'm joined today by Stephen Wolfram, the founder of Wolfram Alpha, a physicist, mathematician and IT expert. We discuss Stephen's life-long passion of understanding the world computationally and how that has led to the development of Wolfram Language, a language intended for both humans to think in and for computers to read. We explore the practical uses of Wolfram Language and the impact it has had on enterprise systems and consumer transactions. (0:12:30) - Live Streaming Software Design Reviews
(0:13:07) - Live Streaming and Infrastructure Technologies
Stephen Wolfram and I discuss the challenges of streaming live events, the emergence of Grafana and Kubernetes, and the benefits of running own infrastructure. We explore the option of running the system in AWS or other virtualization platforms, and the difficulties in containerizing the system. We look at the diurnal rhythm of traffic and discuss the advantages of running our own operation. Finally, we discuss Wolfram Alpha and the ChatGPT's Wolfram Plug-in, and the new Chat Notebook feature.
(0:19:19) - Software Development in a Chat-Based World
Stephen Wolfram and I explore how to do software development in a world where chat is not reproducible. We examine how to extract useful information from chat conversations, how to version individual steps, and how writing specs in natural language may not be precise enough. We also discuss the unique advantages of Wolfram Language, which was designed to be read as it is written by humans, and how the LLM (Large Language Model) can summarize what is happening with the code beyond what a human could see. This is a novel concept in modern workflow and we consider the implications of this.
(0:31:49) - The Intersection of Technology and Judgment
We examine the implications of the emergence of language models for the future of AI and the impact they may have on the industry. We discuss how the advancements in language models may not be as revolutionary as initially thought, and that the real value for language models may come from the tooling and use cases that develop from them. We also explore the concept of 'technology meeting judgement', and how this has been seen before in the history of the technology industry.
(0:45:53) - The AI-aided CTO role
Stephen Wolfram and I explore the potential of language models and their impact on AI. We discuss how language models can help automate certain tasks and offer guidance to humans. We examine how these models can be used to teach humans to solve well-defined problems and what skills are necessary for a CTO to manage the combination of mission, automation, and people. Finally, we discuss the importance of strategy in the success of any effort.
(0:59:13) - Leveraging all types of Langauge Models (Not just LLMs)
Stephen Wolfram and I discuss how language models can be used to bridge the gap between a company's individual data and public data. We explore the potential of language models, and how they can be used to answer natural language queries and automate certain tasks. We also look at the concept of making companies computable by using language models to read documentation and build an ERP system from existing data.
(1:06:29) - Computational Invention and Exploration
We explore how language models can be used for prototyping and how they allow us to explore the alien minds of generative AIs. We also discuss how technology is largely about taking what already exists in the universe and finding ways to apply it for human purposes. We consider the question of whether there will ever be an end to invention and Stephen introduces the concept of computational irreducibility, which explains that even when you know the rules of a system you can't necessarily know what it will do without just following the rules and seeing what happens. We examine how machine learning is a way of finding a program that will satisfy a technological use case and discuss the challenge of defining use cases.