CS Professor Sounds Alarm on AI and Programmers

Summary

Dr. Matt Welsh gave a talk to Harvard students predicting that AI language models like GPT will replace programming. He argues AI can generate code for a fraction of the cost of employing developers. The future software team may just be product managers, AI coders, and human reviewers. However, we aren't there yet. AI still has poor accuracy, lacks trustworthiness, and can't handle complex tasks. Learning to code is still a vital skill. The industry is changing, not disappearing. There will be new tech jobs we can't envision yet. Embrace AI tools now, but keep pursuing coding to open future opportunities.

Takeaways

• 👀 AI can now generate computational results from natural language prompts
• 😲 Current AI costs could be 10,000x cheaper than a software engineer
• 🤔 Future software teams may consist of PMs, AI coders, and human code reviewers
• 🎯 Always use the latest AI tools to boost your programming productivity
• 💪 Keep learning to code - programming will still be a vital future skill
• 🕑 We're still early - AI isn't trusted yet for full production use in many orgs
• 🌟 AI's multi-turn accuracy remains low for complex tasks like PPT editing
• 📈 AI will enable many new tech jobs that don't exist yet
• 💡 Treat coding as a 'baseline skill' that opens new opportunities
• 👍 Programming skills will help you adapt as industries change

Q & A

What breakthrough capability do current AI models have?

They can now take in natural language and produce computational results, without needing traditional programming.

How much cheaper could AI coders be compared to software engineers?

Potentially 10,000 times cheaper, according to the video's cost analysis.

What roles might be on a future software team?

Product managers, AI coders generating code automatically, and human code reviewers.

Why should programmers start using the latest AI coding tools?

To boost productivity and not fall behind others using these tools.

Why should people still pursue coding careers despite AI advances?

Programming will likely remain a vital skill as industries change, opening new tech job opportunities.

What limits wider use of AI in production now?

Issues with accuracy, ethics and lack of trust in AI to function correctly without human oversight.

Where does AI still struggle with accuracy?

In multi-turn conversations and complex tasks like editing a PowerPoint presentation.

How will programming skills help in the future?

As a baseline skill that enables adapting to new tech jobs and opportunities that arise.

Why treat coding as a broad enabling skill?

It opens up many future possibilities beyond current programming jobs and roles.

How can programming help you as industries change?

It provides a technical skillset to help transition to the new roles and opportunities that emerge.