Physical Address

304 North Cardinal St.
Dorchester Center, MA 02124

Deepseekek, the next-generation AI agents can erode the value of large models


Jaap Arriens Nurphoto via Getty Images

Large language models like developed Microsoft– Openai is open this year, the next generation artificial intelligence artificial agents and open source rivals, according to higher directors, are rapid progress.

Last week, the AI ​​Chinese company Deepseek released R1, on Open source reasoning that claims to address Openai O1 model both cost and performance. Open source refers to the software and their source code is freely available on open web, change and reduction.

This week, the depenseses greater awareness of the new model caused severe in shares Nvidia And other giant technologies, investors can have an impact on the powerful processing of the graphics required to prepare and execute advanced workloads.

Nvidia Lost 600 million dollars in the lowercase market Monday – the largest fall for any single day for any company in the US history.

Meanwhile, the directors of the most important AI laborers and scientists are changing from large language models that can do actions on your behalf.

LLMs are the basic technology behind the current AI creators. However, experts believe that the AI ​​system agents impetus – LLM technology incorporates the value of the models this year.

LLMS production

LLMS is expected to have more goods in the near future, as technology becomes increasingly advanced and continues to lower the costs involved in training.

Thomas Wolf, an opponent and the head of science, will be integrated into intelligent systems related to the LLM company databases.

“I think people moves around the model around the model, thanks to open sources … Many of these models will be free and freely available,” Wolf said CNBC. Hug face is a popular open source storage for AI projects.

Hug face creator: The next step in the air will be applications

What are the AI ​​agents?

Over the last two years, a large part of the focus has been on LLM products, the AI ​​Chatbots of the AI ​​creators can create images and video content tools.

Puppy-based application creators are based on a certain amount of money to ask, on the contrary, the AI ​​agents are designed for more action.

If you wanted to book a doctor’s appointment, for example, Chatg can request information about your closest internship in your GP – similar to how you would consult the website using Google’s search engine.

In the future, the next generation AI will be the agents that can know where your nearest GP is, it offers us useful time for your doctor and book an appointment – all directly in a single platform. It is expected to transform the way to interact with the web.

“Since the decades were invented, the standard web page and menu interface hasn’t changed”, “Bobby Blumofe, Akamai Technological Technology Manager, told CNBC before.

“We live in this tyranny tied to a screen and you need to navigate to the web pages and menus you need to remember.”

Model changes to operators

Last week, Arthur Mench, Mistrale Director General said, “the focus should change two language models and include context data.

“That’s what it varies,” Mensch’s CNBC’s Arjun Kharpalk in Davos, Switzerland, world economy forum world.

Mistral CEO: We have proven that Europe can build a successful AI Tech

“It also means that the industry takes its specialization to these systems, which will be necessary to take air value to these industries.”

Mistral, which is protected MicrosoftIt has quickly become AI in most companies. Only two years ago Investors serve about $ 6 billion In June 2024.

The agent trend has already reached a great traction. Last week, Openai launched his own agent, which is called the operator. Openai has described the website as an agent that can go to you “adding” to interact with the “, adding” networks daily to interact with menus, and text fields “.

Microsoft has invested a million dollars in Billions in Openai, The AI ​​agent shot himself In November.

Anthropic, an Gym-En Openai research managers were founded at the AI ​​launch Submitted to “computer usage” What is his AI agents to use computers like humans to complete complex tasks.

Dario Amodei, the CEO of Anthropic said CNBC’s Andrew Ross Sorkin said in Davos last week, should be thought of that it integrates the “broader view” that integrates these applications used in Davos last week.

“We are organizing something in 2025 this year to build something we call” virtual collaborator “, describing Amodei, describing it as an agent that works on your computer.”

Anthropic CEO: It is safer than ever

“You can tell the tasks, write this feature for this application, and write some codes, it will verify the code, he will talk to his colleagues with his colleagues, or in Google Docs, or In other platforms, “explained the chiefs of the anthropic.

Anthropic plans next year to gradually build parts of this view – while he assured that it is “safe and controllable”, the amodei added.

Paul O’Sullivan, Salesforcek, the main technology in the UK and Ireland, will be more important in verbal communication in the New World of AI Agents. Salesforce had definitely launched his Own agent system in front of the AI ​​game agent before Microsoft, Open and Anthropic.

“We’ll change more,” CNBC said in a final interview. “The screens will be smaller, our eyes will worsen. But they will probably be replaced with mobile devices that are stronger than laptops.”

Leave a Reply

Your email address will not be published. Required fields are marked *