Why AI Agents Don't Work (yet) - with Kanjun Qiu of Imbue

TL;DR

Kin, CTO of Mbw, discusses their mission to build AI agents that are reliable, inspectable, and controllable, and how they are working towards rekindling the dream of the personal computer.

Transcript

hey everyone welcome to the laden space podcast this is alesio partner and CTO and residents at deel partners and I'm joined by my co-host swix founder of small AI hey and today in the studio we have kin from mbw welcome thank you so uh we you and I have I guess cross paths a number of times uh and you're form you're formerly named General generall... Read More

Key Insights

• 🏛️ Mbw aims to rekindle the dream of the personal computer by building reliable and inspectable agents.
• 🤩 They believe in optimizing models for reasoning, as it is a key blocker for agents that can handle complex and long-running tasks.
• 👤 Mbw is focused on creating non-leaky abstractions and building interfaces that empower users to interact with agents effectively.
• ❓ Memory and context are considered important challenges to overcome for future advancements in AI.
• 🔨 Mbw is exploring the possibility of releasing models and tools that can be helpful to other AI builders, particularly in the evaluation and benchmarking areas.

Questions & Answers

Q: How did Kin's experience at Dropbox and as a founder shape their journey to Mbw?

Kin's experience at Dropbox and as a founder allowed them to gain insights into product development, building teams, and identifying gaps in the market. It also sparked their interest in AI and its potential.

Q: What inspired Kin to focus on human agency and potential?

Kin believes that humans have the potential to achieve much more than they currently do, and they are passionate about unlocking human agency. They were inspired by their experiences with friends and individuals who felt limited by traditional perceptions of what they could achieve.

Q: How does Mbw approach the challenge of building reliable and trustworthy AI agents?

Mbw is focused on optimizing their models for reasoning and building interfaces that allow users to inspect and modify the model's actions. They also explore ideas such as multi-role critiquing and aim to make their agents as reliable and controllable as possible.

Q: What makes Mbw's approach unique compared to other AI research labs?

Mbw's strong focus on reasoning, reliability, and inspectability sets them apart. They also emphasize the use of their own agents internally and strive to build tools and interfaces that are non-leaky and user-friendly.

Summary & Key Takeaways

• Kin shares their background in computer science, their experience at MIT and MIT media lab, as well as their journey as a founder in the tech industry.

• Kin talks about their passion for human agency and potential, and how they believe AI can enable humans to achieve more and be more than they are today.

• Kin discusses their previous venture, Sorceress, and how their experience with AI in recruiting taught them the importance of trustworthy and reliable agents.

• Kin explains how Mbw is focused on building reasoning-based AI agents, optimizing for reliability and working on interfaces to enhance user interaction and trust.