Computer-Use Agents and the Future of the Agentic Internet

The episode explores the evolution of the MLops community, which is now integrating large language models and agents, leading to the formation of the Agents AI community in collaboration with the Agents AI Foundation. This rebranding reflects a significant shift in focus towards agentic technologies.

Demetrius Springman announces the upcoming Agent Con event, a key gathering for the Agents AI community, with a North American edition set for October 22nd in San Jose. The event will feature a 36-hour live stream to accommodate both in-person and virtual attendees, although potential technical issues are anticipated.

The Agents AI Foundation has emerged as a neutral governing body for the MCP, previously owned by Anthropic, and currently includes over 250 member companies. The ambassador program has attracted over a thousand applications, and local chapters have been established in more than 80 cities to foster community engagement.

The rise of agentics is reshaping the technology landscape, enabling agents to automate tasks like checking emails and booking flights. However, there are concerns about granting agents purchasing authority without oversight, and public hesitance towards AI technologies persists despite their potential advantages.

The episode highlights the challenges of B2B applications of computer use compared to B2C scenarios, while predicting a growing demand for agentic interfaces as consumers begin to recognize their benefits. The speaker expresses cautious optimism about AI's future, noting that positive personal experiences with AI tools can lead to broader acceptance among less tech-savvy users.

The discussion also addresses the changing role of computer-use agents in e-commerce, suggesting that chat interfaces are becoming the new browsing method for online shopping. E-commerce platforms will need to adapt to attract agents rather than just human customers, as these agents are expected to learn consumer preferences and operate autonomously.

The relationship between companies and customers is evolving with the integration of agents, although current agents lack the necessary discernment for effective upselling or promotional strategies. Future e-commerce may feature agentic interfaces that enable negotiations between agents, potentially enhancing the shopping experience through the collaboration of purchasing and negotiation agents.

There is uncertainty regarding how major companies like Amazon will respond to the rise of agent-driven shopping, with the possibility that agents could handle purchases directly, circumventing traditional marketplaces like Shopify. Consumers may favor agents that understand their needs, which could lessen their involvement in shopping, but significant uncertainties remain about whether these agents will prioritize consumer interests over profit motives.

The rapid advancement of AI technology could lead to substantial changes in consumer-agent interactions before the end of the year. Concerns are raised about OpenAI potentially becoming a marketplace that prioritizes its own products, which could undermine consumer trust. Despite these worries, consumers are generally willing to share personal information with agents if they believe those agents are acting in their best interests, although the future of agent-based shopping is expected to be complex as various stakeholders explore different models.