No one knows what an AI agent is

Silicon Valley is bullish on AI agents. Openai CEO Sam Altman said the agent This year, I will be joining the workforce. Microsoft CEO Satya Nadella predicted the agent It replaces specific knowledge work. Salesforce CEO Marc Benioff said Salesforce’s goal is “The world’s largest provider of digital labor” via the company’s various “agents” services.

But no one seems to agree with what the AI ​​agent agrees teeththat’s right.

Over the past few years, the tech industry has boldly declared that AI “agents” (the latest buzzword) will change everything. Like an AI chatbot AAIP of opic Agents who claim CEOs like Altman and Nadella offer new ways to express information.

That may be true. But it also depends on how you define “agents”, which is not an easy task. Like other AI-related terms (“multimodal”, “AGI”, “AI” itself), the terms “agent” and “agent” are diluted to pointless points.

This threatens to leave Openai, Microsoft, Salesforce, Amazon, Google and countless other companies that build an entire lineup of products around agents in nasty places. Amazon’s agents are not the same as agents from Google or other vendors, leading to confusion and customer dissatisfaction.

Ryan Salva, senior director of Google’s products and former leader of Github Copilot, said he “hate” the word “agent.”

“I think our industry is overloading the term “agent” to the point of making it almost pointless,” Salva told TechCrunch in an interview. “(It’s) one of my pet’s pee.”

The agent-defined dilemma is nothing new. Former TechCrunch reporter Ron Miller, last year’s work I asked:What is an AI agent? The problem he identified is that almost every company’s construction agents approach technology in a different way.

This is a problem that has recently gotten worse.

This week, Openai Blog post published This agent is defined as an “automated system that allows tasks to be accomplished independently on behalf of the user.” However, in the same week, the company released it Developer Documentation This agent was defined as “LLMS equipped with instructions and tools.”

Leher Pathak, Openai’s API product marketing lead, said later I’ll post it on x She understood that the terms “assistant” and “agent” were interchangeable.

meanwhile, Microsoft’s Blog Please try Distinguish Between the agent and the AI ​​assistant. The former, which Microsoft calls the “new app” in the “AI-powered world,” can be tailored to have specific expertise, but assistants simply help with common tasks such as drafting emails.

AI Lab Humanity deals with agent-defined Hodge Podge a little more directly. in Blog posthumanity states that agents can be defined in several ways, including both “a fully autonomous system that operates independently over a long period of time” and “normative implementation following a predefined workflow.”

Salesforce has perhaps the broadest definition of AI “agents.” According to the software giant, The agent is “The type of (…) system that allows you to understand and respond to customer inquiries without human intervention.” The company’s website lists six different categories, ranging from “simple reflective agents” to “utility-based agents.”

So why is it confusion?

Well, agents – like AI – are ambiguous and they are constantly evolving. Openai, Google, and Prplexity are just beginning to ship what they consider to be their first agents – Openai Operator, Google Project Marinerand Perplexity Shopping Agent – And those features are all over the map.

Rich Villars, IDC’s global research GVP, noted that technology companies have a “long history” of not strictly following technical definitions.

“They care more on a technical level about what they are trying to achieve,” Villers told TechCrunch. “Especially in a rapidly evolving market.”

However, according to Andrew Ng, founder of AI Learning Platform DeepLearning.ai, marketing is largely responsible too.

“The concepts of AI ‘agent’ and ‘agent’ workflows used to have technical implications,” Ng said in a recent interview.

The lack of a unified definition of agents is an opportunity and a challenge, says Jim Rowan, head of AI at Deloitte. On the one hand, ambiguity allows for flexibility, allowing businesses to customize agents to suit their needs. On the other hand, this can lead to “false expectations” and difficulties when measuring value and ROI from an agent project.

“Without a standardized definition, at least within an organization, it’s difficult to benchmark performance and ensure consistent results,” Rowan said. “This can result in a variety of interpretations of what AI agents should provide, and complicate the goals and outcomes of the project. Ultimately, flexibility can promote creative solutions, but a more standardized understanding will help companies better navigate the AI ​​agents’ landscape and maximize their investment.”

Unfortunately, if the elucidation of the term “ai” is any indication, it appears unlikely that the industry will soon merge one definition of “agent” any time soon.