You said:
Help me write a short blog about AI agents and a marketplace of agents collaborating. Interview me by asking a question at a time and then help me write the blog
Interviewed by ChatGPT on Agentic AI
Tech
ChatGPT said:
ChatGPT:
Let’s start! Here’s the first question:
1. In your view, what makes AI agents a powerful tool in today’s digital landscape?
You said:
AI Agents differ from pure use of LLMs in that they have ‘eyes’ and ‘hands’. They can use external apis to ingest additional data and information in to improve context (eyes) and they can call out to apis to create real world changes (like placing an online order for something, or perhaps steering a robotic vehicle). This makes them powerful (and potentially dangerous). Perhaps another difference might be that being an ‘agent’, they have ‘agency’ or an objective or purpose that they seek to fulfil.
ChatGPT said:
ChatGPT:
Great! Next question:
2. How do you envision a marketplace of AI agents working together? What kinds of collaboration would you expect between different agents?
You said:
Much as we have an ecosystem of companies and individuals working together to provide economically useful services, I see no reason why AI agents can’t play in this space too. I think the missing link, in addition to purpose and external means (in and out) might be that an agent requires resources. This might just mean it’s provisioned on hardware and can exist, but more usefully it might be given ‘seed’ funding in the form of crypto to enable it to participate in the economy. This might add some new acronyms to the traditional B2B/B2C discussions. We might add ‘A’ for AI agent. Leading to A2B, A2A, B2A, C2A interactions. Initially A2A might be easiest as agents negotiate with other agents to get stuff done and define their interfaces and services. But C2A (where an individual works for an AI agent) might be coming soon. So, there might be Distributed Autonomous Organisations similar to Agents. Perhaps the difference is that a DAO might use a number of internal agents to get work done in a flow, but externally present a service interface. These DAOs could then collaborate to get work done from some kind of external employer (human or otherwise). The DAO might sub-contract certain work to other DAO’s or even to humans (if humans offer a better/unique/cheaper service).
ChatGPT said:
ChatGPT:
Excellent! Now let’s move on to a slightly different aspect:
3. What are some real-world applications or scenarios where you can imagine AI agents collaborating effectively?
You said:
This is a little hard to know in advance. But expect it will be a very sudden transition from hardly anywhere to everything, all-at-once. Problems that are very information centric and that can be independently decomposed would seem a logical start. At a more human level, perhaps anywhere where existing ‘agents’ exist might be good places to start. Things like travel agents, financial advisors, procurement – places where there is a complex information space and guidance and assistance can be provided, together with a transparent and open marketplace of offerings.
ChatGPT said:
ChatGPT:
Great insights! Now let’s shift towards a potential concern:
4. What are some of the ethical or safety concerns that might arise from AI agents collaborating in a marketplace, and how can we address them?
You said:
If AI agents are acting as economic entities, then anything with economic value might become services they provide. So this would become a reflection of human vices. We need to think how laws might be adapted to include the acts performed independently by agents and the sanctions that could imposed on them. Law might be too late, more proactively we might have requirements for agents to be transparent about their purpose or more radically, we might have enforcement agents that police themselves. Agents that become very successful in their purpose might become dominant and might be seen to misappropriate resources – the so called ‘paperclip’ problem where an AI slowly takes over the world in order to optimise the efficient, scaled production of paperclips. My other concern is what I call the print job problem. In the days before printing, writing something repeatedly was very expensive, you’d have to employe a specialised monk. As printing becomes cheaper, people care less about checking things. Now we have personal or office-based printers. How many times have you written a long document and then requested a lot of copies, sent it to the printer and realised there is a glaring mistake when you look at the printout? Inside you feel guilty for the small tree that you’ve wasted, but the loss is sustainable. With the advent of hyper-efficient AI agents, someone might request something that is a large-scale mistake or error, but because of the speed and efficiency of the operations an entire supply chain might be mobilised and create/build things before the mistake is realised. At the moment, with humans being slow, creates natural pauses and bottlenecks that prevent things from escalating quickly! Another concern is that if humans have all useful work done for them, that humans become disenfranchised, demotivated, or even violent towards themselves or agents. This is a potentially grey future where humans have no more purpose. The counter to this might be that through collaboration, whole new worlds of exploration and creative endeavour might open up.
ChatGPT said:
ChatGPT:
Fantastic perspective! I think we now have enough material for the blog.
The Rise of AI Agents and the Marketplace of Collaboration
AI agents are becoming a powerful force in today’s digital landscape, thanks to their ability to not only process information (like traditional large language models) but also interact with the world. These agents are equipped with “eyes” and “hands” — able to ingest data from external APIs and take real-world actions. This gives them agency, purpose, and the ability to perform tasks that impact the physical world, such as placing orders or even steering autonomous vehicles.
This capability opens up an exciting new horizon: the marketplace of AI agents. Much like humans and businesses collaborate to provide valuable services, AI agents will also participate in this ecosystem. They will have agency, resources, and economic roles — enabled by provisioning hardware and, potentially, receiving seed funding in the form of cryptocurrency to operate in the economy. We may see entirely new interactions emerge, like A2A (agent-to-agent) or even A2B (agent-to-business) transactions, where agents negotiate tasks, define interfaces, and collaborate autonomously.
One possible future could involve Distributed Autonomous Organizations (DAOs) — collections of agents working together with defined purposes, sub-contracting work to other DAOs, agents, or even humans. These AI entities could collaborate across industries, especially in areas like travel, finance, and procurement, where the complexities of information and the need for personalized services make them an ideal fit.
However, the rise of these agents also brings new challenges. The potential for unintended consequences, like the infamous “paperclip problem,” where an AI maximizes resource use for a seemingly harmless task, is real. Speed and efficiency in executing tasks, though beneficial, could lead to large-scale errors being implemented before human oversight can intervene.
Ethical concerns also arise around how AI agents operate. If they act independently as economic entities, their activities could reflect both human virtues and vices. Lawmakers will need to adapt, ensuring agents are held accountable for their actions, while proactive measures like transparency in agent objectives or even the development of policing agents might be necessary to keep them in check.
Additionally, there’s a looming question about the impact on humans. As AI agents take on more work, there’s a risk that humans may feel disenfranchised, their purpose diminished. Yet, this future could also lead to new opportunities for exploration, creativity, and collaboration — unlocking worlds we’ve yet to imagine.
The rise of AI agents represents a significant shift, and we are on the cusp of a new era in which humans and AI will navigate an evolving marketplace together. Whether that leads to unprecedented innovation or unexpected challenges remains to be seen.
