One year ago Nvivia spent about $ 1 billion in acquiring two Israeli software companies. The first deal for running: AI, who has used a solution to make the most of the power processors processors in NVIDIA; And the second for dechi.ai with a “shrug” technology of creative AI models allowed to use less computing power. While claiming to run: AI has a high jump by regulatory hurdles until the approval of eight months, the claims of dechi.i has completed the shortcomings.
Dechi.ai, “Globes” can be revealed to be united with NVIDIA software operations, at a time of launching the company’s depths, the launching of the “equipment of equipment” in recent months. These models are called “intelligent agent.”
Unlike language models in recent years, which are generally intended for training and operating staff in progressing staff in progress and care for those who have employees and may be replacing some of them. Nvidia discusses AI agents who wrote software code, comparing scientific studies and participate in accounting tasks such as payroll management in today’s payroll management.
The paper at Dis.AI in progress
In recent months, Nvidia announced several agreements. For example, in Servicenow, using the model claims to deploy customer service and employee support agents; An agreement with business systems in the systems of Israeli systems and SAP and Salesforce to develop agents who can also respond to questions related to business information.
To make customers – the software giants – to build and operate intelligent agents themselves, NVIDIA itself has launched firms of committed tasks with complex tasks. The model, based on training from Meta’s Llama, NVIDIA’s Llama Nemotron, and sold free as open sporters in company fees.
Related articles
Today, VP Geneative AI Software in Nvidia for Enterprise Kari Briski, reveals dechi.i paper in new development. Dechi.ai helps the number of parameters and bits used for the operation of language models, so that they can process AI operations faster to process power. Briski tells “Globes,” “The deci uses a neural architectural software, which is permitted to the accuracy of 49 billion, which is also the accuracy of models to increase in 169
Until last year, Dechi is one of the few companies in Israel to develop personal language models. It was founded by Dr. Yonatan Geifman, Prof. Ran El-Yaniv, and Jonathan Eliel, some of whom were google executives who were discouraged by the high costs dictated by ai processing due to the expensive hardware needed, and were looking for a solution from the software world that could accelerate the use of ai and allow more software developers to benefit from it. Before selling for $ 300 million, Dechi raises $ 55 million insultors in insult and the fund fund emerges.
Helped Derekeek model
Briski explained the choice of NVIDIA to keep the open source model. He said, “Our model of Nimotron is based on Llama 3, because it is the average open model of business customers, and we find that there are many customer programs.”
Brisky also pointed out that Llama is not just the star of its new open model. AI Higant used new and successful r1 models, which won praise from Pir Altman and Pather Gates, etc. “We use synthetic data from R1, but we control its data,” he explained.
In March, Nvidia CEO Jensen Huang presents Deresteek and the same as a large growing machine for the company, because the chips of chips. It came two months after Dreeseek was launched by $ 600 million out of the NVIDIA market a day.
Published in Globes, Israel News in News – en.globes.co.il – On May 22, 2025.
© Copyright in Globes Publisher Itonut (1983) Ltd., 2025.
