Jaap Arriens | Nurphoto via Getty Images
Great language models such as those developed by MicrosoftOpenai, which was placed in Openai, is scheduled to become a commodity this year amid rapid progress towards artificial intelligence agents of the next generation, the most intelligent competitors and the source, according to senior technology executives.
Last week, the Chinese company Deepseek released AI R1, Open Source Thinking Model claims the O1 Openai Model On both costs and performance. The open source indicates the programs that the source code, which is available freely on the open Internet, is provided for potential modification and distribution.
This week, awareness of the new Deepseek model is increasing Stagnate In stocks Nafidia And other technology giants, where investors are afraid of a potential reduction in spending on the powerful graphics processing units required to train and operate advanced AI's AI's work burdens.
Nafidia He lost nearly 600 billion dollars in the market value On Monday-the largest day to any company in the history of the United States.
Meanwhile, all executives and scientists at Leading Ai Labs talk about the transformation from large language models to the so -called “artificial intelligence agents” who can implement procedures on your behalf.
LLMS is the foundation technology behind artificial intelligence applications today. However, experts believe that pushing AIC AIC – which includes LLM technology – will lead to erosion this year of the value of these models.
LLMS commodity
LLMS is expected to become more than a commodity in the near future, as technology has become increasingly developed and the costs of training and operation continue to decline.
Thomas Wolf, co -founder and chief science official at Hugging Face, suggested that LLMS become more integrated into smart systems related to the company's databases.
“I think people move from this madness about the model, and thanks to the open source … many of these models will be free and available freely,” Wolf told CNBC. Huging Face is a famous warehouse for open source AI projects.
He added: “In the Internet revolution, we move from building websites as the main businesses to build original companies on the Internet-therefore, Airbnb of AI, the artificial intelligence tape.” “They are not about the model. They are about the system and how to make the model useful for tasks.”
Deepseek's R1 model is seen last week A More index llms has become an increased commodity.
R1 includes some of the main aspects that distinguish it from other models on the market. For example, the model includes a “mixed accuracy” working frame that uses a fully accurate FP32 (FP32) FP32 numbers (FP32) and low 8 -bit floating point numbers (FP8).
The latter is faster for treatment but can be less accurate. However, instead of relying on one or another, Deepseek FP8 is used for most accounts and switches to FP32 for certain tasks where there is a higher degree of accuracy.
“Since artificial intelligence becomes more efficient and accessible, we will see its use rises, and turn it into a commodity that we cannot get enough,” Microsoft CEO Satia Nadella books On the x social media platform, in a clear reference to Deepseek.
Meanwhile, Matt Kalkins, CEO of the United States Software Company, told CNBC that Deepseek's success simply shows that artificial intelligence models will become more than a commodity in the future.
“In my opinion, we will see a commodity of artificial intelligence. Many Amnesty International companies will achieve competitiveness, and a lack of distinction will be bad for the first high -spending engines.”
What are the agents of artificial intelligence?
In the past two years, a lot of focus has been on LLM products, starting from Chatbots of artificial intelligence such as Chatgpt from Openai to tools that can produce images and video content.
LLM -based obstetric applications such as ChatGPT depend on a certain amount of hand from users to constantly claim them for outputs, while artificial intelligence agents are designed to be more orientation towards work.
If you want to book a doctor's appointment, for example, you can ask ChatgPT to get information about your closest GP practices – similar to how to inquire about the web using Google search engine.
In the future, Execs Tech expects that the next generation of artificial intelligence will include agents who can know the location of the closer GP, provide the available times for your doctor, and reserve an appointment for you – all directly inside one platform. This is expected to turn the way we interact with the web.
“Over the decades since the web has been invented, the standard interface for web pages and menus has not changed much,” Bobby Bloomov, chief technology official in Akamai, told CNBC previously.
“We are somewhat live in this tyranny of connecting the screen and connecting it to web pages and the menus that you should remember how to move.”
Transformation from models to agents
Last week, Arthur Minche, CEO of Mistral, said he believed that “the focus should turn into systems” that merge both language models and contextual business data.
“This is the place where this turns,” Minsh told CNBC Arjun Kharpal in the annual gathering of the World Economic Forum in Davos, Switzerland.
“This also means that the industry you adopt will distort its experience in those systems – this is where this will be necessary to pay the value of artificial intelligence to those industries.”
Bad phases, which support him MicrosoftHe quickly became one of the most artificial intelligence companies in Europe. Founded only two years ago, it was the case The last value by investors is about $ 6 billion In June 2024.
The trend of agents has already gained a large traction. Last week, Openai launched its own agent, called the operator. Openai describes it as “an agent who can go to the web to perform tasks for you,” adding that he is trained to interact with “buttons, menus and text fields that people use daily” on the web.
Microsoft, which invested billions of dollars in Openai, The agents of the artificial intelligence itself were launched In November.
Man, and AmazonThe start of the artificial intelligence founded by Openai executive officials last year “Use of the computer”, “ It is a capacity that allows an artificial intelligence agent to use computers like humans to complete complex tasks.
Dario Ameudi, CEO of Anthropor, Andrew Ross Sorkin of CNBC in Davos last week that the use of computer should be believed to be a prelude to a “wider vision” that combines these factors with common use applications.
“We are planning this year in 2025 to create something that we call” the virtual collaborator, “said Amodei, describing it as” an agent working on your computer. “
“You can give them tasks. You can say, write this feature for this application, you will write some software instructions, you will collect the symbol, and you will check the code, and will speak to his colleagues in the recession, on Google documents, or on another platform,” the president explained Antarbur.
Amodei added that humanitarian plans to build parts of this vision gradually over the next year – while ensuring that they are “safe and controlled.”
Paul Osoulivan, chief technology in the United Kingdom and Ireland, expects verbal communication to be more important in the new world of artificial intelligence agents. It could have been said that Salsforce early in Agency AI, as she launched its Agentforce system in September before Microsoft, Openai or Anthropic.
“We will talk more,” he told CNBC in a recent interview. “The screens will become smaller, and our eyes will get worse. But it is likely to be replaced by stronger mobile devices than laptops anyway.”
