Hello and welcome to the eye for artificial intelligence. In today’s edition … companies that try artificial intelligence agents say technology lacks expectations; NVIDIA announces its new chips and new positions in the post -deep scene; Elon Musk and Nvidia join the Microsoft-Blackrock AI; Among the mullahs’ intelligence from the most recent intelligence are “imposing brute” on the Internet; Foxconn appears as a major player in the global artificial intelligence race.
Sreats a day passes without a technical company announced a new “agent” saying that he will revolutionize the progress of work and cancel the unprecedented competence. But while I like the makers of these agents Salesforcefor Amazonfor OracleAnd tons of startups – classified, some of their customers are increasingly skeptical that these tools can be provided, at least now.
“Many customers report a gap between marketing and reality,” reads new Report from CB InsightsWhich analyzes the main pain points surrounding these products.
Throughout the month of March, CB Insights wiped more than 40 agents of artificial intelligence and found that she faces reliability, integration and security problems. Other last title events highlighted some of the same issues. For example, there was an increase in the excitement on Manus, which was described as the first “general agent” independent and they were extended by some other Deepseek moment in China – until user tests revealed unreliable performance and doubtful outputs.
The idea of the artificial intelligence tool that can organize complex tasks independently and precisely, is logical as a goal for the pursuit, and this can be achieved. But the current reality is that customers pass unconfirmed water, and that the noise cycle and disorganized use of the term “agent” causes confusion about what users can actually expect.
(The United Nations) is the highest
Deepmind The founder and CEO Demis Hassabis recently presented an insightful description of the reliability problems surrounding the factors of artificial intelligence, and compared them to more attention.
He said this week in a Google The event, according to Computer weekly. He went on to describe how it was by the time when these five steps were done through the possibility of the correct answer to be “random.”
For companies that need accurate information and customer service, the random possibility for accuracy is not usually acceptable. CB Insights has reported reliability as the highest anxiety among clients who use artificial intelligence agents, indicating nearly half of them as an issue. One of the customers described obtaining partial processing information and hallucinations from the agent of Amnesty International, which he published, for example.
Customers also face problems in integrating artificial intelligence agents into their current systems. The lack of inter -operating a long -term headache has caused headaches in the world of institutions programs, but with artificial intelligence agents, integration is a kind of basic point. “It was gambling to register for a product in which they had no integration operations that we wanted,” one of the clients told CB Insights.
A new group of security risks
Safety also tops the list of customer concerns, for a good reason. The presence of a technology connected to different systems that contain sensitive information and take measures, independently open huge risks. Gartner predicts By 2028, 25 % of institution violations will be tracked to abuse of artificial intelligence agents from internal, external and malicious actors.
“Without appropriate governance, artificial intelligence agents can unintentionally display sensitive data, make unauthorized decisions, or create blind spots to comply,” said Dimitri Cerotta, CEO of Data Intelligence and Compliance Big Id.
He said that the best way to safely experience artificial intelligence agents is to avoid products that are not transparent about how to take an artificial intelligence agent. Companies must also try artificial intelligence agents in the environments that are controlled so that they can detect risks and control as necessary before scaling.
What is even the “agent” of artificial intelligence?
The market of artificial intelligence agents has become saturated, especially in specific outlets such as customer support and coding. At the same time, “Nobody knows what an artificial intelligence agent against hell” Techcrunch It was frankly put in a story published last Friday, on the pretext that the term became “reduced to the point of no meaning.”
Each company determines the “artificial intelligence agent” slightly differently. Some generally use the term to indicate completely independent artificial intelligence systems that can carry out tasks independently, while others use them to indicate the systems that follow the pre -defined workflow. Some offer other definitions. Others – such as Openai – often show the change and conflict of their previous definitions. There are many tools that were previously called “artificial intelligence assistants”, as well as “agents”.
For information technology leaders, this identification chaos creates headaches confusion and publishing. It is not only difficult to understand what products do and how they work, but it is also impossible to compare standards and performance standards.
Nothing of this means that companies do not start to see some Benefit from artificial intelligence agents. But it is a reminder that this is still very The first days of this technology, the noise is run before reality.
However, here is more news of artificial intelligence.
Hakim Lazaro
Sage.lazzaro@consultant.fortune.com
sagelazzaro.com
This story was originally shown on Fortune.com
