Agora, the AI cost killer

and AI in medicine, robots, build your agent!, computer use

Author

Martin Musiol
November 08, 2024 • Estimated Reading Time: 13 minutes

Hey there, it’s Martin.

1001 things have happened in tech/ AI. Here’s what you need to know.

✅ AI spots breast cancer - advances in AI & medicine + DEMO
✅ Robots continue advancing and might become reality sooner rather than later
✅ Agora - new protocol for multi-AI agent communication
🤖 My face-to-face exclusive workshop - Build your own AI Agent!
🔨 Many use cases open up for Anthropic’s Computer Use Feature
🍓 Worth noting: OpenAI bought chat.com-domain for $15.5M.

Before we started, we launched a premium version of the newsletter. Subscribing gives you 100% access to all content, exclusive demos, and an ad-free experience. I plan to host AMAs and develop more subscriber-requested demos.

Learn AI in 5 Minutes a Day

AI Tool Report is one of the fastest-growing and most respected newsletters in the world, with over 550,000 readers from companies like OpenAI, Nvidia, Meta, Microsoft, and more.

Our research team spends hundreds of hours a week summarizing the latest news, and finding you the best opportunities to save time and earn more using AI.

own animation

AI spotted breast cancer - AI advances into medical fields - DEMO

(Source)

A woman is "deeply grateful" to AI for detecting her breast cancer after she was initially given the all-clear following a routine scan.

AI is increasingly used as a second opinion in medical diagnosis, with several clinical trials underway. Two of many examples:

  • At Imperial College London, the AI tool "Mia" detected up to 13% more breast cancers than human radiologists alone.¹

  • Harvard Medical School’s AI model, CHIEF, achieved nearly 94% accuracy in identifying various cancers, including breast cancer, and predicting patient outcomes.²

The even better thing? You don’t have to wait for these trials to conclude—you can already get a second opinion from Grok 2. I demonstrated this in a previous demo.

Here it is again:

Analyzing medical images in-depth → challenges real-world analysis. (DEMO)

Robots continue advancing and might become reality sooner rather than later

The industry around humanoid and non-humanoid robots is making daily progress.

The image shows a middle-aged man with a bald head, dressed in a professional outfit, consisting of a dark-colored, collared shirt and a light gray blazer. He appears to be standing against a black background, which creates a strong contrast and highlights his facial features and attire. The man is captured in a profile view, looking slightly to his left with a neutral or contemplative expression. The lighting focuses on his face, casting shadows that emphasize the shape of his head and his facial contours. A small microphone is clipped to his shirt, indicating that he may be speaking at an event or giving a presentation. The overall mood of the image is serious and professional. Jeff Bezos

Physical Intelligence (startup) raised $400M from investors like Jeff Bezos and OpenAI, reaching a $2B valuation. Their AI model, π0 (pi-zero), aims to control various robots, shown performing tasks like laundry folding and table clearing.

Some envision an exciting future, saying:

  • There will be more robots than people.

  • Robots will become as commonplace as smartphones are today.

  • You can have robot bodyguards.

  • Restaurants might say, “Your robot must stay outside!”

Others, however, see great danger in it.

Are you excited about the idea of robots everywhere?

Login or Subscribe to participate in polls.

Outstanding for me is how Unitree can control their robots.

Agora - addressing "Agent Communication Trilemma"

Agora is a protocol that improves AI Agent communication. AIs communicating with AIs.

The image is a complex flowchart featuring various well-known tech logos connected by dashed arrows. The diagram appears to represent a network or a process flow involving interactions between different AI and tech systems. Here’s a detailed breakdown: ### Key Elements: 1. **Logos/Symbols:** - **Meta (blue 'infinity' logo):** Represents Meta/Facebook’s involvement in the flow. Appears several times, suggesting multiple interactions or integration points. - **Microsoft (multi-colored square logo):** Likely indicates usage of Microsoft services or platforms (e.g., Azure or Cognitive Services). - **OpenAI (GPT logo):** Used throughout the chart, signaling interaction with OpenAI’s models, such as GPT. - **Anthropic (purple 'Lambda-like' logo):** Refers to Anthropic’s AI models, possibly used for ethical or safe AI purposes. - **LangChain (black '/'):** Denotes use of LangChain, a framework for building language model applications. - **Camel icon:** Likely represents "LLM-Camel," suggesting a task or data handoff. - **Vertex AI (blue star-like shape):** Represents Google’s Vertex AI, integrated into the flow for model training or inference. - **Clip and magnifying glass (link search icon):** Implies linking or search processes, possibly denoting integration of search capabilities. - **'M' in pixelated style:** Could represent Mistral AI, a newer model in the AI space. 2. **Flow and Interactions:** - The chart has a **web-like structure** with various nodes connected by dashed arrows, indicating data flow or process handoffs between components. - The flow is non-linear, involving back-and-forth exchanges between systems like Meta, Microsoft, and OpenAI. This suggests a multi-agent environment where different models or services collaborate dynamically. - There are points where **LangChain** connects directly with OpenAI, indicating it might be orchestrating tasks involving language models. - **Google’s Vertex AI** appears centrally positioned, possibly serving as an orchestration or monitoring layer. - **The Camel icons** and links to **Anthropic** models suggest checkpoints for responsible AI usage, ensuring safety or compliance checks. 3. **Overall Structure:** - The flow has no clear start or end, implying it’s a part of a **continuous integration system** or a cyclic process where multiple AI models, APIs, and services are engaged. - The structure is decentralized, emphasizing the interaction between various **third-party APIs and AI models** rather than a single, monolithic application. ### Summary: The diagram showcases a complex integration of various AI services and platforms (Meta, Microsoft, OpenAI, Anthropic, Google’s Vertex AI), indicating a dynamic multi-agent system where tasks are handled collaboratively. It highlights orchestration (via LangChain), ethical checks (Anthropic), and the usage of diverse AI capabilities (OpenAI, Mistral). This likely represents an architecture for an advanced AI system or platform that leverages multiple models and services for comprehensive task execution and decision-making.

The challenge 

When multiple agents collaborate or communicate, costs can escalate quickly. This issue is part of the so-called Agent Communication Trilemma, which involves balancing costs with the ability to handle a wide variety of messages and ensuring portability (i.e., minimal implementation) in agent interactions.

The solution

By enabling agents to negotiate on implemented structured communication routines, Agora significantly reduces operational costs.

The image is a line chart that compares the cost per query (in USD) across two systems: "Natural Language" and "Agora." Here’s a breakdown of the elements: - **X-Axis (Queries):** Ranges from 0 to 1000, indicating the number of queries made. - **Y-Axis (Cost in USD):** Shows the cost per query, ranging from 0.000 to 0.040 USD. - **Two Lines:** - **Natural Language (Green Line):** The cost starts around 0.035 USD per query, fluctuates throughout the query range, showing periods of increase and decrease, but remains relatively stable around the 0.035–0.040 USD mark. - **Agora (Orange Line):** Begins at a higher cost, approximately 0.027 USD per query, and quickly declines. By around 200 queries, it stabilizes below 0.010 USD, fluctuating slightly but remaining low for the rest of the queries. - **Legend:** Identifies the green line as "Natural Language" and the orange line as "Agora." ### Analysis: - The "Agora" system starts with a higher initial cost but shows a sharp decline, becoming much more cost-effective as the number of queries increases. - The "Natural Language" system has a relatively consistent cost per query, with minor fluctuations but no significant trend of decrease or increase over the query range. - Overall, "Agora" becomes significantly cheaper per query compared to "Natural Language" after the initial queries, indicating a better scalability in terms of cost efficiency.

With this, one crucial problem of upcoming AI systems is tackled, and it will boost the adoption of AI Agents that will work for us and communicate with other AI Agents on our behalf.

My face-to-face exclusive workshop - Build your own AI Agent!

Build your AI Agent! Or team of Agents (e.g., Swarm AI, CrewAI).

The image is a promotional banner for an event titled **AI Summit Seoul 2024**, specifically advertising a workshop. Here’s an in-depth breakdown: ### Layout and Design: - The banner has a sleek, modern design with a dark background that features a blue wave-like pattern at the bottom, giving it a futuristic and tech-oriented feel. - The **top section** features the title of the event in bold text: **"AI SUMMIT SEOUL 2024"** with the year highlighted in blue. Below it, a blue capsule-shaped label states "Workshop," indicating the format of the session. - The **main text section** highlights the workshop’s title: **"Build Your AI Agent: Hands-on Workshop,"** in large, bold, white font to grab attention. - Below the title, the event details are listed: - **Date and Time:** December 12, 2024, at 2:00 PM - **Location:** 2nd floor, COEX Grand Ballroom, Seoul, Korea ### Speaker Section: - On the right side of the banner is a photo of **Martin Musiol**, who is smiling and wearing a maroon sweater over a collared shirt, giving a friendly yet professional appearance. - Below his image, there is a black text box with his name **"Martin Musiol"** in bold, followed by his title: **"CEO | GenerativeAI.net."** The text is white, contrasting well with the black box, making it easily legible. ### Overall Impression: The banner conveys a professional and modern look, aimed at a tech-savvy audience interested in AI. The emphasis on a "Hands-on Workshop" suggests an interactive session where participants can expect practical, actionable learning. The clear mention of the time, date, and venue provides all essential information for attendees at a glance, while the image of the speaker adds a personal touch, increasing credibility and engagement.

I will give an EXCLUSIVE workshop around building your AI colleague who works on the problems you'll provide.

It's simple. Join the workshop, build the AI Agent that solves your problem (3h+), and leave.

Be the first in your organisation that has an actually working AI Agent.
Best about it? You have built it! 🤷‍♂️

You'll be waaaaay ahead of the curve.

I hope to see you there.

Many use cases open up for Anthropic’s Computer Use Feature

I wrote about and demoed Anthropic’s computer use. With this, Claude (the AI) takes frequent screenshots and decides what to do. It can use your mouse or keyboard to perform various actions.

Many use cases are opening up to this feature. The following one is definitely remarkable:

Watching the video, it judges what has gone wrong or what should be recommended for this construction site. Fascinating!

What other use cases can you think of?

own

That’s a wrap! I hope you enjoyed it.

Martin

Our webpage