Conversational AI and NLP: Top Resources from GTC 21

NVIDIA announced several major breakthroughs in conversational AI for building and deploying ASR, NLP and TTS applications.

NVIDIA logo
NVIDIA
•
Brad Nemire
•4 min read•intermediate•
--
•View Original
Deep LearningGPTTransfer Learning

Overview

At GTC 21, NVIDIA unveiled significant advancements in conversational AI, focusing on automatic speech recognition (ASR), natural language processing (NLP), and text-to-speech (TTS) applications. The conference featured over 60 sessions highlighting the latest tools and technologies in these domains, aimed at simplifying the development of conversational AI applications.

What You'll Learn

1

How to build production-ready conversational AI applications using NVIDIA Jarvis

2

Why pipeline-parallelism is essential for running large NLP models like Megatron GPT-3

3

How to customize automatic speech recognition and natural language processing pipelines for specific domains

Key Questions Answered

What advancements in conversational AI were announced at GTC 21?
NVIDIA announced breakthroughs in automatic speech recognition, natural language processing, and text-to-speech applications, along with tools like NVIDIA Jarvis that simplify the development of conversational AI applications. The conference also included over 60 sessions showcasing the latest technologies and research in these areas.
How can organizations start developing conversational AI applications?
Organizations can begin developing conversational AI applications by utilizing NVIDIA's tools and technologies, which facilitate the creation of applications such as virtual assistants and real-time transcription. The sessions at GTC 21 provided insights on getting started with these technologies.
What is the purpose of TRITON inference server in NLP?
The TRITON inference server is an open-source software that allows teams to deploy trained AI models from any framework. It is particularly useful for running large NLP models like Megatron GPT-3 by enabling pipeline-parallelism across multiple GPUs to maximize throughput.
What resources are available for building conversational AI applications?
NVIDIA provides a range of resources including pretrained models, notebooks, and sample applications for conversational AI available from the NGC catalog. Additionally, tutorials for building and deploying these applications can be found on the NVIDIA Technical Blog.

Technologies & Tools

Some links below are affiliate links. We may earn a commission if you make a purchase.

Backend
Nvidia Jarvis
Used for building production-ready conversational AI applications.
Backend
Triton
An open-source inference server for deploying trained AI models.
AI/ML
Megatron-lm
A large model used for natural language processing tasks.
AI/ML
Onnx
Used for converting models to a format suitable for deployment.

Key Actionable Insights

1
Leverage NVIDIA Jarvis to streamline the development of conversational AI applications tailored to your enterprise needs.
Using Jarvis can significantly reduce the time and effort required for fine-tuning models, allowing teams to focus on building unique features rather than getting bogged down in the complexities of model training.
2
Utilize the TRITON inference server to efficiently deploy large NLP models across multiple GPUs.
This approach not only maximizes throughput but also allows for handling larger batch sizes, which is crucial for applications requiring high performance and scalability.
3
Engage with the NVIDIA Developer Program to access exclusive tools and training resources.
Joining the program can provide developers with the necessary support and resources to effectively utilize NVIDIA technologies in their projects.