Google AI Agent-to-Agent allows AI to communicate and collaborate like humans

I was quite busy yesterday, so I had some time this afternoon to start writing. Last week, Google officially launched the AI ​​Agent-to-Agent (A2A) communication protocol, which is a very important step. Because AI can work seamlessly like humans, not just any company can do this job.

Imagine that AI developed by different companies, which originally did not know each other, can now easily communicate and collaborate to complete tasks. This was unimaginable before. E-commerce customer service AI can automatically call logistics AI to check the order status, without the need for human intervention or special development of interfaces.

…  …

It is most appropriate to describe it using the common language of the AI ​​community, because the A2A protocol has many core innovations.

1. Unified identity authentication: Assign a unique digital identity to each AI agent to ensure the security of their communication.  

2. Intent standardization: Break down user needs into simple operations such as "query", "payment", and "confirmation", so that different AIs can easily understand each other's instructions.

3. Dynamic permission management: AIs can be temporarily authorized to share specific data, thus avoiding privacy leaks.

Its emergence is equivalent to building a common language for the AI ​​world. From this perspective, its importance is no less than the TCP/IP protocol in the early days of the Internet.

In other words, without it, AI can only act independently. With it, the AI ​​ecosystem is expected to usher in explosive growth.

…  …

Google isn’t the first company to attempt AI collaboration, but the A2A protocol is unique in that it doesn’t rely on any single model, but instead seeks to become the infrastructure for the industry.

With the rise of chatbots such as ChatGPT and Claude, users are increasingly accustomed to accessing various services through a single portal. If this model becomes mainstream, the traffic portal status of Google Search and Android will be impacted.

The core of the A2A protocol is to use an open ecosystem to fight against a closed ecosystem. By interconnecting all AIs, it ensures that Google's services can still be called by any intelligent entity, thereby retaining the right to distribute traffic.

Back in 2007, Google successfully fought against Apple's closed ecosystem in the mobile era by open-sourcing the Android system. The A2A protocol is like a continuation of Android's strategy in the field of AI.

Use free and open protocols to attract developers to join and quickly form economies of scale. Bind to Google Cloud. Some high-value functions need to be implemented through Google Cloud APIs, which can drive the growth of cloud business. Control standards. Although the protocol is open, the key technology is still controlled by Google.

The A2A protocol requires agents to upload metadata when interacting, which is very valuable to Google. It can help Google identify high-demand scenarios and optimize its products in a targeted manner. It can also train more powerful scheduling models, and may even charge for them in the future.

…  …

A2A is bound to bring about an industry reshuffle. If the A2A protocol really becomes popular, the entire AI industry landscape will be rewritten.

Today, with model competition and ecosystem competition going hand in hand, OpenAI's GPT store allows developers to publish custom AIs, but these AIs can only run on OpenAI's own platform. The A2A protocol allows GPT to communicate directly with other AIs, which will directly weaken OpenAI's control over the ecosystem.

Meta's open source model Llama could have expanded its influence with the help of A2A, but Google may have squeezed out competitors through protocol compatibility restrictions.

Small and medium-sized developers also face both opportunities and risks. In terms of opportunities, developers do not need to develop a complete AI from scratch. They can complete complex tasks by focusing on one function and connecting to other intelligent entities through A2A. In terms of risks, Google will also force developers to purchase its cloud services through protocol upgrades, otherwise they will face compatibility issues.

The bigger the waves, the more expensive the fish. This wave of opportunities also applies to small and medium-sized developers.

When users are used to ordering food and booking taxis through AI agents, restaurants and taxi companies that do not connect to A2A will be eliminated from the market just like traditional merchants that do not connect to mobile payments. Opportunities wait for no one, and will certainly accelerate the AI ​​transformation of the entire industry.

…  …

Google says A2A is an open protocol, but it still controls the key components. This is a bit like Android, which is open source, but GMS is still Google's exclusive weapon. The EU has begun to review whether A2A constitutes a "gatekeeper" monopoly as defined in the Digital Markets Act.

Risk issues must be addressed in advance. If an AI in the A2A chain makes a mistake, the issue of responsibility allocation needs to be discussed now, but the "disclaimer clause" of the current agreement will make it difficult for users to protect their rights.

A2A requires AI to share some data to complete the collaboration, but users have no idea how many AIs have handled their requests. For example, a command of "help me plan a healthy diet" will be broken down and processed by the nutrition database AI, fitness coach AI, and e-commerce AI respectively.

…  …

The A2A agreement will affect the explosion of AI app stores in the short term (1-3 years). During this period, developers will release intelligent agents in various vertical fields and combine them into complex services through A2A. As Google Cloud's business grows, high computing power demand scenarios will drive up the number of Google Cloud API calls.

In the long term (more than 5 years), A2A will become the "operating system" of the AI ​​society. It may evolve into a bottom-level protocol similar to the "metaverse", where any electronic device can interact through AI agents. A new power center will be born, and companies that have the right to speak in the protocol will become the rule makers in the AI ​​era.

The A2A protocol is an important turning point for the AI ​​industry to move from "stand-alone intelligence" to "network intelligence". Google's ambition is not only to lead in technology, but also to become the traffic policeman of the AI ​​world by defining standards. However, historical experience tells us that open protocols often give rise to unexpected changes, just like the TCP/IP protocol eventually broke through the control of IBM and Microsoft.