Nobody knows what is an artificial intelligence agent against hell

Photo of author

By [email protected]


Silicon Valley is optimistic about artificial intelligence agents. The CEO of Openai Sam Altman said You will “join the workforce” this year. The CEO of Microsoft Satya Nadella predicted that the agents It will replace some knowledge work. The CEO of Salesforce Marc Benioff said that Salesforce’s goal is to be.The number one provider for digital work in the worldThe “agent” services “the agent”.

But no one can agree on what is the agent of artificial intelligence He isexactly.

In the past few years, the technology industry has boldly announced that “artificial intelligence agents” – the latest tonn words – will change everything. In the same way as Chatbots, Amnesty International Chatgpt Openai He gave us new ways to flatten the information, and the agents will mainly change how we deal with work, and demand CEOs such as Altman and Nadella.

This may be true. But this also depends on how one defines “agents”, which is not an easy task. It is very similar to other terms related to the Acting (for example, “multimedia”, “AGI” and “AI” itself, the terms “Agent” and “Agentic” have become reduced to the point of no meaning.

This threatens to leave Openai, Microsoft, Salesforce, Amazon and Google, and countless companies that build a collection of full products around agents in an embarrassing place. Amazon agent is not the same Google agent or any other seller, and this leads to confusion – and frustration of customers.

Ryan Salva, the chief director of the producer in Google and the former Copilot leader, said he came to “hate” the word “agents”.

Salva told Techcrunch in an interview: “I think our industry overcomes the term” agent “to the point that it is almost meaningless,” Salva told Techcrunch in an interview. (It is) one of Bevis.

The definition of the agent is not new. In last year, former Techcrunch correspondent Ron Miller to request: What is the agent of artificial intelligence? The problem that he identified is that all company building agents approach technology differently.

It is a problem recently deteriorated.

This week, Openai Publish a blog post These factors are determined as “automatic systems that can accomplish tasks independently on behalf of users.” But in the same week, the company issued Developer documents These factors specified as “LLMS are equipped with instructions and tools.

Leher Pathak, the API Openai Product Marketing Command, said later in A. After x It understood the terms of “assistants” and “agents” to be substitute – which increases the distortion of water.

During, Microsoft blogs Try He distinguishes Between agents and artificial intelligence assistants. The previous, which Microsoft calls “new applications” for “a world working with the same Amnesty International”, can be customized for a specific experience, while assistants only help in general tasks, such as formulating emails.

AI LAB ANTHROPIC The HodgePodge is treated directly the agent’s definitions. in Blog postAntarbur says that factors “can be defined in several ways”, including “fully autonomy systems that work independently over extended periods” and “mandatory applications that follow the pre -determined workflow.”

Salesforce has the most definition of Amnesty International. According to the program giant, Agents “A type of system (…) can understand customer inquiries and respond to it without human intervention.” The company’s website lists six different categories, ranging from “simple reflexive agents” to “tool agents.”

So why is chaos?

Well, agents – like artificial intelligence – is a mysterious thing, and they are constantly evolving. Openai, Google and Perplexity have just started charging what the first agents consider – Openai workerand Google Mariner ProjectAnd Shopping agent in confusion – And its capabilities throughout the map.

Rich Villars, GVP of Worldwide Research at IDC, indicated that technology companies “have a long history” of strict abuse of technical tariffs.

“They are more interested in what they are trying to accomplish,” said Villars Techcrunch, especially in advanced fast markets. “

But marketing is also the blame in a large part of it, according to Derro Ng, the founder of the AI ​​Deeplearning.ai learning platform.

“The concepts” of the agents of “Amnesty International and the functioning of the” agent “had an artistic meaning”, but about a year ago, the marketers and a few major companies obtained it.

Jim Rowan, head of Amnesty International, says to DELOTTE, the lack of a unified definition of agents is an opportunity and challenge alike. On the one hand, ambiguity allows flexibility, allowing companies to allocate agents to their needs. On the other hand, this may lead – and it can be said that it is already – to “unbalanced expectations” and difficulties in measuring value and returning to investment from agents.

Rowan said: “Without a unified definition, at least within the institution, it becomes difficult to determine performance and ensure fixed results.” “This can lead to various explanations for what artificial intelligence agents must provide, which may complicate the goals of the project and the results. In the end, although flexibility can push creative solutions, the most uniform understanding would help institutions to move better in the natural scene of the artificial intelligence agent and increase their investments to the maximum.”

Unfortunately, if the detection of the term “artificial intelligence” is any indication, it is unlikely that the industry will gather about one definition of the “agent” any time soon – if any.



https://techcrunch.com/wp-content/uploads/2023/02/GettyImages-1065679054.jpg?resize=1200,849

Source link

Leave a Comment