The Death of Classical Computer Science

The interview presents a compelling and thought-provoking perspective on the future of computer science, driven by the rapid advancement of large language models. Matt Welsh, with his background as a former Harvard professor and AI researcher, offers credible insights into the potential of AI to replace traditional programming. The argument is well-structured, starting with the premise that language models are becoming capable of general-purpose reasoning and computation, and then extrapolating to a future where AI directly solves problems without human-written code. This is a plausible scenario given the current trajectory of AI development, and Welsh supports his claims with references to existing tools like Copilot and the potential for AI to use computers like humans.

However, the discussion is largely speculative and lacks empirical evidence. While Welsh’s predictions are grounded in his expertise, they are not backed by data or peer-reviewed research. The conversation remains at a high level, without delving into the technical challenges or limitations of current AI systems. For instance, the issue of AI reliability and the potential for errors is only briefly mentioned, and the discussion does not address the computational costs or environmental impact of training large models.

The interview also touches on the societal implications, such as job displacement and the democratization of computing, but these are not explored in depth. The speakers acknowledge concerns but do not provide a detailed analysis of how these issues might be mitigated.

In terms of sources, the conversation references tools and concepts like Copilot, Llama models, and function calling, but does not cite specific research papers or studies. The description includes links to the speakers’ profiles and recommended books, which are useful for further reading but are not directly cited in the discussion.

The title accurately reflects the content, and the discussion is engaging and accessible, making it suitable for a broad audience interested in the future of technology. However, for a more rigorous scientific analysis, the claims would need to be substantiated with data and a more critical examination of the potential pitfalls.

Overall, the interview offers valuable insights into the potential direction of AI and computer science, but it should be viewed as an opinion piece rather than a scientifically rigorous analysis.

The title accurately reflects the provocative thesis that classical computer science is being replaced by AI-driven problem-solving, which is the central theme of the discussion.

7/10

The discussion is based on the expert opinion of Matt Welsh, a former Harvard professor and AI researcher, and Julian Wood, a developer advocate. They present a forward-looking perspective on the impact of AI on computer science, but the claims are not backed by empirical data or peer-reviewed research. The conversation is speculative and opinion-driven, though it draws on the speakers' extensive experience in the field.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:30Introduction: Matt Welsh states that classical computer science is coming to an end.
• 02:00Welsh explains that language models are becoming capable of general-purpose reasoning and computation.
• 04:00Discussion on the evolution of interfaces, from text to voice, and the potential for AI to use computers like humans.
• 06:00Welsh predicts that within five years, the industry will shift dramatically, with AI agents performing tasks traditionally done by programmers.
• 08:00Discussion on the commoditization of AI models and the role of open-source models like Llama.
• 10:00Welsh discusses the need for AI to be integrated into everyday computing, moving beyond the current siloed text-box interactions.
• 12:00The conversation touches on the challenges of training expensive models and the potential for AI to become more accessible.
• 14:00Welsh compares the current state of AI to the pre-iPod MP3 players, suggesting that a company like Apple might bring it all together.
• 16:00Discussion on the future of data and open-source models, including the importance of privacy and control.

• GOTO Unscripted — The interview is part of the GOTO Unscripted series.
• Article on GOTO Unscripted — The article associated with this interview.
• Matt Welsh's website — Matt Welsh's personal website.
• Matt Welsh on Twitter — Matt Welsh's Twitter profile.
• Matt Welsh on GitHub — Matt Welsh's GitHub profile.
• Matt Welsh on LinkedIn — Matt Welsh's LinkedIn profile.
• Ultravox AI — Matt Welsh's AI company.
• Julian Wood on Bluesky — Julian Wood's Bluesky profile.
• Julian Wood on Twitter — Julian Wood's Twitter profile.
• Julian Wood on GitHub — Julian Wood's GitHub profile.
• Julian Wood's website — Julian Wood's personal website.
• Julian Wood on LinkedIn — Julian Wood's LinkedIn profile.
• AI Assisted Programming — Recommended book by Michael Feathers.
• Running Linux — Recommended book by Matthias Kalle Dalheimer and Matt Welsh.
• AI for People and Business — Recommended book by Alex Castrounis.
• Reinforcement Learning — Recommended book by Phil Winder.
• Data Science (The MIT Press Essential Knowledge series) — Recommended book by Kelleher and Tierney.

• GOTO Unscripted — The interview is part of the GOTO Unscripted series, which features expert discussions on software development.
• Matt Welsh's website — Matt Welsh's personal website, providing background on his work in AI.
• Ultravox AI — Matt Welsh's AI company, which aligns with his views on AI's future.

• No discordant sources found — The interview does not cite any sources that contradict its claims.

• www.linkedin.com
• www.youtube.com