Developers: | Nvidia |
Date of the premiere of the system: | 2020/05/14 |
Branches: | Science and education, Trade, Pharmaceutics, medicine, health care, Financial services, investments and audit |
Technology: | Speech technologies |
2020: Nvidia Jarvis representation
On May 14, 2020 the NVIDIA company provided the GPU accelerated framework of NVIDIA Jarvis which allows the companies to use video and voice data for creation of the dialogue AI services adapted for their industry, products and clients.
As noted in NVIDIA, the trend of remote work, training and development of telemedicine increased demand for specialized dialogue AI services, from client support to conversion of the speech to the text in real time and summarizations of video calls to make processes of interaction it is more productive.
"Dialogue AI opens a way to the future for many industries as applications have an opportunity to understand and interact with users to react taking into account nuances and a context. NVIDIA Jarvis will help the organizations to automate quickly overloaded client service in health care, finance, education and retail", noted Jensen Huang, the founder and the CEO of NVIDIA |
According to the developer, the applications created using Jarvis will get advantage during the work on the graphic processor NVIDIA A100 with tensor cores for AI calculations and an inferens with optimization in NVIDIA TensorRT. Work of the multimodal applications using powerful speech models and models of computer vision, in real time with delays less than 300 milliseconds became possible.
Jarvis has the full-fledged GPU accelerated program stack and tools which simplify to developers a task of creation, deployment and start of the dialogue real-time applications capable to understand terminology, specific to the different companies, claim in NVIDIA.
"IDC continues to observe rapid growth of the market of dialogue AI as the companies of all scales begin to realize the value of well trained virtual assistants and chat-bots for customer service and business development. It is expected that world costs for dialogue AI, for example, the automated assistants to a support service, will grow from 5.8 billion dollars in 2019 to 13.8 billion dollars in 2023, i.e. for 24% in a year", noted David Schubmehl, the director of AI platforms in IDC |
To offer the clients the interactive personalized interaction, the companies need to train the dialogue applications on the data characteristic of their products and requirements of their clients. But creation of service from scratch requires knowledge in the field of AI, the large volume of data and computing resources for training of models and also program providing for regular loading of new data in model.
According to the statement of the developer, Jarvis solves this problem, offering the full-fledged pipeline of deep learning for dialogue AI. Modern models of deep learning, such as NVIDIA Megatron BERT, for recognition of the natural speech enter it. The companies can adapt these models for themselves using NVIDIA NeMo, optimize for an inferens using TensorRT and unroll in a cloud or on the periphery using tables of Helm available in the NVIDIA directory of the GPU optimized NGC software.
As of May, 2020 the early access to NVIDIA Jarvis is open for limited number of the companies.