AI Agents: The Convergence

The world's dedicated space for AI Agents and the AI Agents Global Challenge

Hello,

This is issue #3 of the AI Agentplex! The world’s dedicated newsletter for AI Agents covering key developments in AI research, industry and start ups as well as updates on the AI Agents Global Challenge (www.aiagentschallenge.com)

This issue we will cover:

• Updates on the AI Agents Global Challenge

• Industry Headlines

• The Convergence

• Hackathons, Events, & Summer Internship

---

AI Agents Global Challenge updates

We officially launched the US$ 1million Challenge on 11 March and have already received hundreds of sign ups to the competition. Remember, you can keep refining your applications up to the submission deadline of 1 September so we encourage you to apply early and get access to our team throughout the application window.

Based on the feedback we received from the community and prospective applicants, we will soon be announcing important further details and guidelines to the Challenge. Please make sure you are subscribed to this newsletter for updates.

In the meantime, please get in touch via our Discord to join our community or email us for a 1:1 discussion

---

Industry headlines and talks

• Automating the Enterprise with Foundation Models (ECLAIR): Researchers at Stanford University propose a system to automate enterprise workflows with minimal human supervision. They conducted initial experiments showing that multimodal FMs can address the limitations of traditional RPA and they also identify human-AI collaboration, validation, and self-improvement as open challenges, and suggest ways they can be solved with data management techniques.

• Automating Complex Business Workflows with Cohere:  Cohere’s models Command R and R+ have been trained to interact with a variety of tools; automatically take action in an enterprise systems, analyze internal documents and databases, conduct web searches, or write Python code.

• Survey of AI Agent Architectures for Reasoning, Planning, and Tool Calling: Researchers at IBM and Microsoft examine the recent advancements in AI agent implementations, with a focus on their ability to achieve complex goals that require enhanced reasoning, planning, and tool execution capabilities.

• Generalist Computer Agents With Self-Improvement: OS-Copilot, a framework to build generalist agents capable of interfacing with comprehensive elements in an operating system, including the web, code terminals, files, multimedia, and various third-party applications.

---

The Convergence

We believe most efforts and resources on AI will converge towards the field of AI Agents.

All the buzz on AI and how it will change the world needs grounding with realistic examples. Getting answers to questions in a smarter look and feel like ChatGPT (vs Google’s 10-blue-links) is cool, but this is not how our world will significantly benefit from AI!

While our claim that most efforts and resources on AI will converge to the field of AI Agents lacks deep analysis and is simply an opinion or a “hunch”, one thing we should agree on is that to claim a technology is going to revolutionize our world we should be able to quantify and measure it: “In God we trust, everyone else get data.”

Economic productivity is the strongest incentive and most important measure in our world. War, peace, happiness, misery, advancement, and failure are all somehow linked to money/value/economy/incentives. Unless the AI “revolution” can clearly show significant economic productivity, the hype will cool off.

Our obvious claim is that productivity mostly will come from AI Agents. And since our world is a capitalist one, and eventually the “market” corrects itself, most efforts and resources on AI will converge to AI Agents. But why?

AI Agents is a fancy way of saying “a software that autonomously analyzes/reasons and makes decisions on your behalf”. It is a simplistic and shallow definition of an amazingly complicated technology, but the keywords are “on your behalf”. This is the productivity gain: it could be tiny gains or large gains. However small it is, the power of compounding is explosive (think 2^200 is large enough just like 200^200; both are unimaginably large). AI Agents will be highly scalable and hence a tiny productivity gain with scalability leads to massive results because of compounding.

We might be in a euphoric time now, but we should not look at the short-term booms and busts, it is very clear that over the past 100 years, science and math (and AI) are moving in an exponential direction.

Will AI cure cancer next year? Will we have AI agents that perfectly take over our needs without our involvement? Will enterprises replace part of their workforce with AI agents in few years? Unlikely. BUT, the trajectory points to that, whether in 5 years, a decade or two, this is bound to happen.

And now to the most quoted 3-letter acronym on the AI realm in Twitter/X: “AGI” (Artificial General Intelligence).
First, we confess we do not fully understand what it means despite the countless hours we listened to Lex Fridman interviewing great minds and asking them about AGI (and consciousness)!

The world we know today is a result of hundreds of millions of years of evolution, with the human brain reaching reasoning and intelligence relatively recently. The level of advancement and intelligence today is a product of millions of human minds building on each other’s work, successes, and innovations. AGI, similarly, will be achieved through a plethora of AI Agents building on the progress and advancements of each other simultaneously and progressively, this is currently the most plausible approach.

Whether it is economic productivity, or the holy grail of “AGI”, it will all converge to AI Agents!

Hackathons

We are planning to launch two hackathons in the next few months, one will be in-person and the other online. The hackathons will be focused on AI Agents and there will be rewards for winners. Ideas and suggestions are welcome.

Events

Meetups and webinars on Agentplex, AI Agents, and the AI Agents Global Challenge will be announced soon. Ideas and suggestions are welcome.

Update on summer internship program

As communicated in the previous issue of this newsletter, the application deadline is 15 May 2024; applicants should submit a resume that details their relevant work and educational experiences and cover letter stating their interests, qualifications and areas of interest in developing the AI Agents ecosystem to apply@mm01.ai.