Hello and welcome to AI. In today’s edition … companies that experiment with agents AI say technology lacks expectations; Nvidia announces her new chips and positions herself for the landscape after the depth; Elon Musk and Nvidia join the Microsoft-Blekrock AI Fund; Ai unwanted post office “roughly forcing” the internet; And FoxConn appears as a key player in the global AI race.
It is difficult to pass a day without a technological company that announces a new AI “Agent”, which he says will revolutionize the work of work and unlock unprecedented efficiency. But while manufacturers of these agents – companies love Sales power,, Amazon,, ProphetAnd the tons of startups – they are being pushed, some of their customers become skeptical that these tools can deliver, at least right now.
“Many customers report a gap between marketing and reality,” writes in Novi Report from CB Insightswhich analyzes the main points of pain that surround these products.
During March, CB Insights surveyed over 40 AI agents customers and revealed that they deal with problems with reliability, integration and security. Other recent events of the title have pointed out some of the same questions. For example, there was an increase in the arousal over Manus, which was charged as the first autonomous “general agent”, and some praised him as another deep moment for China – until the user tests discovered unreliable performance and questionable results.
The idea of a tool that can autonomously and accurately orchestrate and complete complex tasks makes sense as a goal to strive for, and it is possible that this can be achieved. But the current reality is that customers undergo uncertain water, and the hype cycle and the blurred use of the term “agent” causes confusion regarding what users can actually expect.
(UN) Reliability is top -notch
Deep The founder and executive director of Demis Hassabis recently offered a insightful description of the issue of reliability around AI agents, comparing it to an integral interest.
“If your Ai model has a 1% error rate and plan more than 5000 steps, that 1% of compounds like complex interest rates,” he said this week on a Google event, according to Computer weekly. He continued to describe that by the time when those 5,000 steps had taken the possibility that the answer was correct “random”.
For companies that need to provide accurate information and serve their own customers, it is usually not acceptable to random accuracy. CB Insights informed the reliability as the main care among customers using AI agents, and almost half cited as a problem. One buyer described that, for example, they partially process information and hallucinations from AI agent.
Customers also deal with problems with integrating agents AI into their existing systems. Lack of interoperability has long caused a headache in the company of company software, but with AI agents integration is a kind of whole point. “It was a bit of a gambling we signed up for a product where they didn’t have all the integration we wanted,” one customer told CB Insights.
New part of security risks
Security is also at the top of the Customer Care List is good reason. Connecting technology with various systems that contain sensitive information and take action to autonally open huge risks. Gartner predicts That by 2028, 25% of the company’s violations will be monitored until the abuse of AI agent from internal and external and malicious actors.
“Without adequate management, AI agents can inadvertently expose sensitive data, make unauthorized decisions or create blind compliance sites,” said Dimitri O poor, Data Intelligence and Compliance Company Big ID, for AI.
He said that the best way to which companies can safely experiment with agents AI avoiding products that are not transparent about the AI agent makes decisions. Companies should also pilot agents AI in controlled environments to detect risks and adjust as needed before scalp.
What is even Ai “Agent”?
The AI agents’ market becomes saturated, especially in certain niches such as user support and encoding. At the same time, “no one knows what the hell is AI agent” as Techcrunch He cautiously put in a story published last Friday, claiming that the term became “diluted to the point of meaninglessness.”
Each company defines “AI agent” a little different. Some generally use the term to refer to completely autonomous AI systems that can independently perform tasks, while others use it to refer to systems that follow predefined workflows. Some still offer other definitions. And some – like opening – look that they often change and oppose their own previous definitions. Many tools previously called “AI helpers” are also called “agents”.
For IT leaders, this definition chaos creates confusion and headaches in deployment. Not only is it difficult to understand what the products do and how they do, but it is also impossible to compare the benchmarks and measuring data on performance.
None of this could say that companies are not starting to see some used by AI agents. But the reminder is that they are still very Early days for this technology and hype lasts significantly ahead of reality.
And with that, here’s more news Ai.
Sage Lazzaro
sage.lazzaro@consultant.fortune.com
sagelalazzaro.com
This story is originally shown on Fortune.com
Source link
