Industry executives and experts share their predictions for 2023. Read them in this 15th annual VMblog.com series exclusive.
Coming in 2023: Plug’n’Play Enterprise AI
By Justin Boitano, VP, Enterprise and Edge Computing at NVIDIA
Let’s face it, AI is really hard. Enterprises large and small for the past half-decade have struggled not only to understand where AI might fit into their business objectives but also how to create and implement algorithms for speech AI, recommendation systems, inferencing and other pretrained models once they do.
While AI is expected to drive over $10 trillion in future incremental value, across all industries, yet in a recent survey, IDC found that only 31% of respondents are saying their organizations have AI in production.*
In 2023, “plug and play” AI comes to the rescue. Rather than trying to build a team to do it themselves or stitching together piecemeal components from multiple vendors, enterprises will turn to integrated offerings that can be purchased and customized to any business, application or budget.
Intelligent virtual assistants, AI audio transcription and digital fingerprinting for cybersecurity anomaly detection are just a few of the AI workflows to be offered in simplified packages that can be built, adapted, deployed, managed and serviced in just hours, and then continually updated year after year.
Enterprises that adopt AI solutions to support advanced customer engagements in contact centers, help customers quickly choose the right product, and offer other applications that transform and improve the customer experience will be positioned to increase appeal to a younger generation.
Millennials – today’s 25-40-year-old cohort – are set to be the largest, most profitable demographic for customer-facing industries like banking, retail, telecom and travel. There are 1.8 billion millennials, representing 23% of the world’s population, and they’re used to low-latency business response times and service. This next generation of decision makers also will be embedded within companies looking to upgrade enterprise systems such as security or transportation.
Today’s problem-solving for enterprises increasingly focuses on the digital-minded customer – those with fingers flying across phone screens who are more comfortable talking to a digital assistant than a person, and who harbor notions that stepping into a bank (or even having cash in their pockets) is quaint.
Demand for options to simplify enterprise AI adoption has continued to rise. 2023 will be the year these come to market and take off.
Once Unobtainable, AI Goes Mainstream in 2023
As recession looms around the world, many executives understand integrating AI into operations in 2023 will become a requirement for achieving objectives and saving costs.
The race to bridge a disconnect between understanding and implementing AI is well underway. Customizable, fully supported AI solutions will be widely available for use cases that help drive sales, reduce costs, grow customer satisfaction and enhance operational efficiency.
Pre-Built, Customizable AI Workflows Get AI Up and Running Across the Enterprise
While AI workflows up to now have needed a dedicated R&D team just to get going, the new workflows enable anyone in IT to build and deploy AI.
An enterprise-ready, streamlined road to AI adoption will enable organizations to use integrated software to create and deploy predictive models that automate business processes and gain faster business insights with key AI applications such as conversational AI, recommender systems, computer vision and more.
All these workloads will be readily available in integrated, fully supported solutions that enable a hassle-free path to becoming an AI leader, even for IT teams with little or no AI engineering expertise.
Enterprises in 2023 will have new growth opportunities made possible by AI, especially in many core capabilities. As enterprise demand rises across industries, these streamlined and supported workflows will be among the most sought after by enterprises:
- Data Science: Data science is at the foundation of nearly every AI application. Data scientists are the architects of AI who build the scaffolding and infrastructure needed to transform raw data into AI pipelines. Accelerated data science libraries are key to ensuring these experts can work efficiently, and, in 2023, more enterprises will seek to update legacy data systems with options that can finally be modernized and accelerated.
- Speech AI: Speech AI-based applications enable customers to converse with devices, machines and computers to simplify and improve their lives. The technology has been a boon to employees working in call centers, and, in the year ahead, we’ll see call-center operational efficiency continue to grow in importance. As a subset of conversational AI, speech AI includes automatic speech recognition (ASR) and text-to-speech (TTS) to convert human voices into text and, conversely, generate a human-like voice from written words. These capabilities enable powerful technologies like virtual assistants, real-time transcriptions, voice searches and question-answering systems – and we’ll see more breakthroughs in this area in 2023.
- Pretrained Models: Some of the most famous large-scale pretrained models such as BERT and GPT, have become performance milestones for AI workflows. With pretraining objectives and huge parameters, large-scale pretrained models can capture knowledge from massive amounts of labeled and unlabeled data. The adoption of these models is the foundation for quickly accomplishing downstream tasks rather than building models from scratch. These “out-of-the-box” models are trained on a variety of domain-specific data for hundreds of thousands of hours.
- Real-Time Inferencing: Inference is at the heart of production AI, driving applications across all industries. But as data scientists and engineers push the boundaries of what’s possible in computer vision, speech, natural language processing and recommender systems, AI models are rapidly evolving and expanding in size, complexity and diversity. The majority of deployed inference scenarios will include complex pipelines or ensembles of multiple models. To streamline development amidst this growth, organizations have to adopt a full-stack approach to AI inference. This will equip them to keep pace with evolving machine learning and deep learning frameworks and offer a standardized approach to both accelerated and non-accelerated computing in the cloud, on premises and at the edge, keeping the cost of AI inference low. In 2023, we will see stricter latency thresholds for real-time inferencing of the entire pipeline and ensemble, across applications.
- Enterprise Security: Cyber threats are growing in number and sophistication. Organizations are responding by harnessing the power of AI to perform real-time security analytics and out-of-band detection of malicious activity to find threats within massive amounts of data. They’re building systems that embody a zero-trust, security-everywhere architecture that accelerates security processing and offloads it from the host CPU. Using unsupervised learning, AI workflows will enable threat detection to achieve comprehensive data visibility. This improves security by helping enterprises uniquely fingerprint every user, service, account and machine across the network to detect anomalous behavior.
The rapid evolution of AI makes it difficult for organizations to build and maintain a high-performing, scalable enterprise infrastructure that can keep pace with advancements.
Many companies are finding that operationalizing enterprise AI requires an ongoing investment to ensure employees are up to speed with constant software updates, security alerts, compatibility issues and more. This high-touch, high-resource reality of enterprise AI maintenance and support can be what hinders companies from achieving full enterprise AI maturity.
2023 is the year we’ll see the world’s leading AI companies address their challenges with simplified hybrid cloud offerings that are easy to customize, easy to use and easy to support. These offerings will meet the full stack of AI-adoption obstacles with a full stack of solutions.
*IDC Press Release, IDC Survey Illustrates the Growing Importance of Purpose-built AI Infrastructure in the Modern Enterprise, February 2022
##
ABOUT THE AUTHOR
Justin Boitano is the VP enterprise and edge computing at NVIDIA, leading the company’s enterprise accelerated data center business.
