Vlad Sejnoha, Partner at Glasswing Ventures, former CTO & SVP R&D at Nuance, and Kleida Martiro, Principal at Glasswing Ventures are contributing authors.

Generative AI (Artificial Intelligence) and its underlying foundation models represent a paradigm shift in innovation, significantly impacting enterprises exploring AI applications. For the first time, because of generative AI models, we have systems that understand natural language at a near-human level and can generate and synthesize output in various media, including text and images. Enabling this technology are powerful, general foundation models that serve as a basis or starting point for developing other, more specialized generative AI models. These foundation models are trained on vast amounts of data. When prompted with natural language instructions, one can use these learnings in a context-specific manner to generate an output of astonishing sophistication. An analogy to generative AI used to create images may be the talented artist who, in response to a patron’s instructions, combines her lifelong exposure to other artists’ work with her inspiration to create something entirely novel.

As news cycles eclipse one another about these advancements, it may seem like generative AI sprang out of nowhere for many business and executive leaders. Still, the reality is that these new architectures are built on approaches that have evolved over the past few decades. Therefore, it is crucial to recognize the essential role the underlying technologies play in driving advancement, enterprise adoption, and opportunities for innovation.

How we got here

The most notable enabling technologies in generative AI are deep learning, embeddings, transfer learning (all of which emerged in the early to mid-2000s), and neural net transformers (invented in 2017). The ability to work with these technologies at an unprecedented scale – both in terms of the size of the model and the amount of training – is a recent and critically important phenomenon.

Deep learning emerged in academia in the early 2000s, with broader industry adoption starting around 2010. A subfield of machine learning – deep learning – trains models for various tasks by presenting them with examples. Deep learning can be applied to a particular type of model called an artificial neural net, which consists of layers of interconnected simple computing nodes called neurons. Each neuron processes information passed to it by other neurons and then passes the results on to neurons in subsequent layers. The parameters of the neural net models are adjusted using the examples presented to the model in training. The model can then predict or classify new, previously unseen data. For instance, if we have a model trained on thousands of pictures of dogs, that model can be leveraged to detect dogs in previously unseen images.

Transfer learning emerged in the mid-2000s and quickly became popular. It is a machine-learning technique that uses knowledge from one task to improve the model performance on another task. An analogy to understand this powerful technique is learning one of the “Romance Languages,” like Spanish. Due to their similarities, one may find it easier to learn another romance language, like Italian. Transfer learning is essential in generative AI because it allows a model to leverage knowledge from one task into another related task. This technique has proven groundbreaking as it mitigates the scarcity of data challenge. Transfer learning can also improve the diversity and quality of generated content. For example, a model pre-trained on a large dataset of text can be fine-tuned on a smaller dataset of text specific to a particular domain or style. This allows the model to generate more coherent and relevant text for a particular domain or style.

Another technique that became prevalent in the early to mid-2000s was embedding. This is a way to represent data, most frequently words, as numerical vectors. While consumer-facing technologies, such as ChatGPT, demonstrate what feels like human-like logic, they are a great example of the power of word embeddings. Word embeddings are designed to capture the semantic and syntactic relationships between words. For example, the vector space representation of the words “dog” and “lion” would be much closer to each other than to the vector space for “apple.” The reason is that “dog” and “lion” have considerable contextual similarities. In generative AI, this enables a model to understand the relationships between words and their meaning in context, making it possible for models like ChatGPT to provide original text that is contextually relevant and semantically accurate.

Embeddings proved immensely successful as a representation of language and fueled an exploration of new, more powerful neural net architectures. One of the most important of such architectures, the “transformer,” was developed in 2017. The transformer is a neural network architecture designed to process sequential input data, such as natural language, and perform tasks like text summarization or translation. Notably, the transformer incorporates a “self-attention” mechanism. This allows the model to focus on different parts of the input sequence as needed to capture complex relationships between words in a context-sensitive manner. Thus, the model can learn to weigh the importance of each part of the input data differently for each context. For example, in the phrase, “the dog didn’t jump the fence because it was too tired,” the model looks at the sentence to process each word and its position. Then, through self-attention, the model evaluates word positions to find the closest association with “it.” Self-attention is used to generate an understanding of all the words in the sentence relative to the one we are currently processing, “it.” Therefore, the model can associate the word “it” with the word “dog” rather than with the word “fence.”

Progress in deep learning architectures, efficiently distributed computation, and training algorithms and methodologies have made it possible to train bigger models. As of the time of writing this article, the largest model is OpenAI’s ChatGPT3, which consists of 173 billion parameters; ChatGPT4 parameter information is not yet available. ChatGPT3 is also noteworthy because it has “absorbed” the largest publicly known quantities of text, 45TB of data, in the form of examples of text, all text content of the internet, and other forms of human expression.

While the combined use of techniques like transfer learning, embedding, and transformers for Generative AI is evolutionary, the impact on how AI systems are built and on the adoption by the enterprise is revolutionary. As a result, the race for dominance of the foundation models, such as the popular Large Language Models (LLMs), is on with incumbent companies and startups vying for a winner-take-all or take-most position.

While the capital requirements for foundation models are high, favoring large incumbents in technology or extremely well-funded startups (read billions of dollars), opportunities for disruption by Generative AI are deep and wide across the enterprise. 

Understanding the technology stack

To effectively leverage the potential of generative AI, enterprises and entrepreneurs should understand how its technology layers are categorized, and the implications each has on value creation.

The most basic way to understand the technologies around generative AI is to organize them in a three-layer technology “stack.” At the bottom of this stack are the foundation models, which represent a transformational wave in technology analogous to personal computing or the web. This layer will be dominated by entrenched incumbents such as Microsoft, Google, and Meta, rather than new startup entrants, not too different from what we saw with the mobile revolution or cloud computing. There are two critical reasons for this phenomenon. First, the scale in which these companies operate, and the size of their balance sheets are pretty significant. Secondly, today’s incumbents have cornered the primary resources that fuel foundation models: compute and data.

At the top of this stack are applications – software developed for a particular use case designed for a specific task. Next in the stack is the “middle layer.” The middle layer is where enabling technologies power the applications at the top layer and extend the capabilities of foundation models. For example, MosaicML allows users to build their own AI on their data by turning data into a large-scale AI model that efficiently runs machine learning workloads on any cloud in a user’s infrastructure. Notably, an in-depth assessment of the middle layer is missing from this discussion. Making predictions about this part of the stack this early in the cycle is fraught with risk. While free tools by incumbents seeking to drive adoption of their foundation models could lead to a commoditization of the middle layer, cross-platform or cross-foundational model tools that provide added capabilities and optimize for models best fit for a use case could become game-changers.

In the near term, preceding further development in the enabling products and platforms at the middle layer, the application layer represents the bulk of opportunities for investors and builders in generative AI. Of particular interest are user-facing products that run their proprietary model pipelines, often in addition to public foundation models. These are end-to-end applications. Such vertically integrated applications, from the model to the user-facing application layer, represent the greatest value as they provide defensibility. The proprietary model is valuable because continuously re-training a model on proprietary product data creates defensibility and differentiation. However, this comes at the cost of higher capital intensity and creates challenges for a product team to remain nimble.

Use cases in generative AI applications

Proper consideration of near-term application-layer use cases and opportunities for generative AI requires knowledge of the incremental value of data or content and a complete understanding of the implications of imperfect accuracy. Therefore, near-term opportunities will be those with a high value of incremental data or content, where more data or content has economic value to the business and low consequences of imperfect accuracy.

Additional considerations include the structure of the data for training and generation and the role of human-in-the-loop, an artificial intelligence system in which a human is an active participant and thus can check the work of the model.

Opportunities for entrepreneurs and enterprises in generative AI lie in use cases where data is very structured, such as software code. Additionally, human-in-the-loop can mitigate the risk of the mistakes an AI can make.

Industry verticals and use cases with these characteristics represent the initial opportunity with generative AI. They include:

Content creation: Generative AI can improve creativity, rate of content creation, and content quality. The technology can also be leveraged to analyze the performance of different types of content, such as blogs or social media ads, and provide insight into what is resonating with the audience.

Customer service and support: Generative AI can augment and automate customer service and support through chatbots or virtual assistants. This helps businesses provide faster and more efficient service to their customers while reducing the cost of customer service operations. By pre-training on large amounts of text data, foundation models can learn to accurately interpret customer inquiries and provide more precise responses, leading to improved customer satisfaction and reduced operating costs. Differentiation among new entrants leveraging generative AI will largely depend on their ability to use fine-tuned smaller models which enable a better understanding of industry-specific language, jargon, or common customer questions as a mechanism to deliver tailored support that meets the needs of each customer and to continuously refine products for more accurate and effective outcomes.

Sales and marketing: AI can analyze customer behavior and preferences and generate personalized product recommendations. This can help businesses increase sales and customer engagement. In addition, fine-tuned models can help sales and marketing teams target the right customers with the right message at the right time. By analyzing data on customer behavior, the model can predict which customers are most likely to convert and which messaging will be most effective. And that becomes a strong differentiator for a new entrant to capture market share.

Software and product development: Generative AI will simplify the entire development cycle from code generation, code completion, bug detection, documentation, and testing. Foundation models allow developers to focus on design and feature building rather than correcting errors in the code. For instance, new entrants can provide AI-powered assistants that are fine-tuned to understand programming concepts and provide context-aware assistance, helping developers navigate complex codebases, find relevant documentation, or suggest code snippets. This can help developers save time, upskill their abilities, and improve code quality.

Knowing the past to see the future

While we are still in the early days of the immense enterprise and startup value that generative AI and foundation models will unlock, everyone from entrepreneurs to C-suite decision-makers benefits from understanding how we arrived at where we are today. Moreover, understanding these concepts helps with realizing the potential for scale, reframing, and growing business opportunities. Knowing where the opportunities lie means making smart decisions about what promises to be an inspiring future ahead.

Artificial Intelligence, Enterprise, Startups

Oracle on Wednesday said it is adding new AI and automation capabilities to its Fusion Supply Chain Management (SCM) and Fusion Human Capital Management (HCM) suites to help enterprises increase efficiency across divisions.

The updates to the SCM suite, which have been made generally available, include an AI-based planning tool, an enhanced quote-to-cash process for Fusion applications, and new rebate management capabilities.

The AI-based planning tool, according to the company, is expected to aid enterprises in improving the accuracy of lead time assumptions across their supply chain through machine learning.

“The new feature can improve planning efficiency and results by identifying lead time trends, anomalies, and their potential impact with prioritized actions and resolution suggestions,” the company said in a statement.

The new planning tool has been added to the planning advisor inside Oracle Supply Chain Planning, an application that is part of the company’s SCM suite.

Since last year, Oracle has been gradually adding new capabilities, including some AI and automation features, to its supply chain suite.

Last year in February, Oracle introduced machine learning-based shipping time forecasting along with real-time analytics for the supply chain.

In October, the company released a new supply chain management application customized for healthcare companies, dubbed Oracle SCM for Healthcare. The application offers capabilities such as a supply chain planning service and added capabilities for the Oracle Procurement application to help drive down supply chain costs.

As part of the new updates to its SCM suite, Oracle is offering an enhanced quote-to-cash process across all Fusion applications.

Quote-to-cash (QTC) process is a part of the sales cycle in an enterprise that constitutes end-to-end delivery of a product or service. Typical components of the process include sales, account management, order fulfillment, billing, and accounts receivables functions.

The enhanced QTC process, according to the company, will help enterprises centralize subscription orchestration, comply with accounting requirements, improve order management, reduce costs while decreasing time to market, and improve customer experience.

“The integrated solution connecting Subscription Management (CX), Configure Price and Quote (CX), Order Management (SCM), and Financials (ERP),enables customers to quote, capture, and fulfill orders (of mixed physical goods, subscriptions, and services) more efficiently and recognize revenue accordingly,” the company said in a statement.

Oracle has added new rebate management capabilities inside the SCM suite to aid enterprises to optimize discounts or promotional campaigns targeted toward their customers.

The new capabilities automate the rebate management process, the company said, adding that automation includes rebate calculation, financial settlement, and closing customer claims.

This helps reduce administration costs and improves customer experience, the company said, adding that the new capabilities are being offered as part of Oracle’s Channel Revenue Management application under the SCM suite.

New updates to Oracle Fusion HCM suite

In addition to the new updates to the SCM suite, Oracle has announced a new application, dubbed Oracle Grow, in an effort to add more AI capabilities to its HCM suite.

“Oracle Grow is designed to enhance the employee experience and improve performance by engaging with individuals to discover new growth opportunities and empowering managers to align upskilling and reskilling with business priorities,” Chris Leone, executive vice president of applications development at Oracle Cloud HCM, said in a statement.

To be offered as part of Oracle Me that was released in April last year under the HCM suite, the AI-powered application delivers personalized insights and intelligent guidance across all interactions from Oracle Learning, Oracle Dynamic Skills, and Oracle Talent Management in one interface, the company said.

In order to provide insights, the AI engine inside the application is trained on an enterprise’s data in a controlled environment before being rolled out completely, according to Natalia Rachelson, group vice president of outbound product management for Fusion Applications at Oracle.

“We would start small, and we would see what kind of results the AI is recommending and then we work with customers to fine-tune the models to adjust to their specific sort of datasets,” Rachelson said, adding that the data normalization is done by Oracle data scientists.

This means that for new Oracle Fusion customers, rolling out Oracle Grow’s AI capabilities would require a longer time, the group vice president said.

Oracle Grow’s AI capabilities include growth experience for employees, suggestions for career paths within their enterprise, personalized development suggestions, and managerial skilling.

The AI engine in Oracle Grow, according to the company, can suggest development opportunities that workers need to adapt to changes in their role, discover new growth options, and achieve their career aspirations. “By unifying people data from across Oracle Cloud HCM, Oracle Grow provides personalized guidance on the next steps employees should take based on their responsibilities, career interests, desired skills, individual learning styles, and changes in the business,” the company said in a statement.

ERP Systems, Oracle, Supply Chain Management Software

The recent mass media love affair with ChatGPT has led many to believe that AI is a “here and now” technology, expected to become pervasive in enterprise and consumer products in the blink of an eye. Indeed, Microsoft’s $10B investment in OpenAI, the company behind ChatGPT, has many people expecting a complete and thorough integration of AI into Microsoft’s product line, from Office365 to Xbox.

The company has already integrated ChatGPT into its Bing search engine and GitHub Copilot, announced that ChatGPT is now available in its Azure OpenAI service, and is looking at further integration into its Word, PowerPoint, and Outlook apps.

But is AI becoming mainstream in security? We’ve seen AI advancements in the cybersecurity world for the better part of the past decade. Companies like Cylance (acquired by Blackberry), and Darktrace, and many others, were marketing their AI-based security technology on billboards and signs at Black Hat and along the 101 near SFO in 2017 and 2018.

From my perspective in the venture world, AI penetration has barely scratched the surface of the cybersecurity market. But to do a sanity check, I recently spoke to over a dozen top CISOs, security executives, and practitioners. Their feedback confirmed my initial thoughts about AI in the early stages of the market. But more interesting to me was that these experts disagreed on where AI played a meaningful role today.

AI in the cybersecurity market

As all my experts pointed out, AI is excellent today at helping a human sort through large quantities of data, reducing “background noise,” and finding patterns or anomalies that would otherwise be very difficult and time-consuming to discover.

AI is also good at creating new threat variants and patterns based on its modeling of the past. However, AI is not adept at predicting the future, despite what some marketing materials may lead you to believe. It may help demonstrate what a future attack could look like, but it cannot produce a result with certainty showing whether a specific exploit will be unleashed.

Another broad belief among the experts was that the AI hype is ahead of reality. While every vendor talks about AI, the executives believe there little (to no) AI integration in most of the products they use today.

One prominent F500 security executive stated, “While many vendors claim the use of AI, it is not transparent to me that it is there. For example, AI might be the secret sauce within SIEM technologies or complement threat detection and threat hunting activities. But my skepticism is due to the lack of transparency.” If this skilled and experienced executive doesn’t know “where the beef is,” where is the reality today?

The perceived reality

Perception is reality, they say, so what do these industry experts perceive? Or conversely, where is today’s AI reality?

The common belief among those I spoke with is that AI is and will be valuable when large datasets are available, both for training and within the actual use case. The experts view SIEM, email phishing detection, and endpoint protection as three of the most likely segments where AI plays a somewhat more significant role today and will likely continue to provide value.

In the SIEM/SOAR category, AI plays a role today, sorting through large quantities of security event data to help humans more quickly detect and respond to threats and exploits. Splunk, in particular, was mentioned as a leading AI_enabled provider in this segment. Again, this view was not universally agreed to by the experts, but most thought that AI penetration was most likely relevant here versus other categories.

In the email filtering and anti-phishing category, large amounts of email data can be used to train systems from companies like Proofpoint and Mimecast, which effectively find many phishing attacks that arrive in an inbox. Several executives I spoke to believed that some AI was powering these products. However, at the same time, a few questioned whether AI was the driving force behind the categorization and detection.

Endpoint companies have leveraged data collected from millions of machines for years to help train their systems. Formerly, these systems produced signatures for pattern-matching across their installed base. Today these products can use AI to detect more dynamic exploits.

While no AI-based system can detect every zero-day attack (as mentioned earlier, AI can’t predict the future), these newer products from companies like CrowdStrike are perceived to close the gap more effectively.

One of the F500 executives I spoke to thought with 100% certainty that CrowdStrike was the best example of a company that demonstrated AI-delivered value. On the other hand, two of the CISOs mentioned that they had no proof that AI was really inside this vendor’s endpoint product, even though they were paying customers.

From just these three segments mentioned above, and the discrepancies in opinion, it is clear that the cybersecurity industry has a problem. When some of the top executives and practitioners in the industry don’t know whether AI is deployed and driving value, despite the marketing claims, how do the rest of us understand what drives our critical defenses? Or do we care?

Perhaps we just abstract away the underlying technology and look at the results. If a system prevents 99.9% of all attacks, does it even matter whether it is AI-based or not? Is that even relevant? I think it is, as more of the attacks we will see will be AI-driven, and standard defenses will not hold up.

AI as problem solver

Looking to the future and other security segments, AI will play a significant role in identity and access management, helping discover anomalous system access. One CISO hoped AI would finally help solve the insider threat problem, one of today’s thornier areas. In addition, there is a belief that AI will help partially automate some of the Red Team’s responsibilities and perhaps automate all of the Blue Team’s activities.

One topic was the threat that adversaries would use ChatGPT and other AI-based tools to create malicious applications or malware. But another suggested that these same tools could be used to build up better defenses, generating examples of malicious code, before bad actors actually use them, and these examples could then help inoculate the defensive systems.

Another concern is that AI-generated code, without proper curation, will be as buggy or buggier than the human-authored code that it was trained on. This creates vulnerable code at a wider scale than possible and will create new issues for AI-based vulnerability scanners to address.

A final key point was the belief that Microsoft, Google, Amazon, and others would provide the underlying AI algorithms. The smaller cybersecurity players will own the data and the front-end product that customers interact with.  But the back-end brain would leverage tech from one of the bigger players.  So, in theory, an AI-based security company won’t technically own the AI.

AI in the future

We are in the early days of AI’s penetration into our security defenses. While AI has been in the research community for decades, the technologies and platforms that make it practical and deployable have just been launched in the past few years.  But where will things be in the next 5-10 years? 

I have a clear investment thesis on AI-enabled cybersecurity solutions and believe we will see much broader and deeper enterprise penetration within the next decade. From the point of view of my experts, the general beliefs are that AI will become a reality in multiple segments, including the three mentioned above.

While the experts believe AI will play an increasingly important in every segment of security, chances are higher in areas like:

Fraud detectionNetwork anomaly detectionDiscovery of deep fake content, including in corporate websites and social media assetsRisk analysis, andCompliance management and reporting (In fact, AI will likely create a new compliance headache for organizations, as more AI-focused regulations will create the need for new processes and policies)

There is so much uncertainty about where AI resides today in cybersecurity solutions and what it does or doesn’t do. But I believe this uncertainty will drive entrepreneurs to create a new wave of products to help navigate this new frontier. This will likely go well beyond cybersecurity, covering all the software products used in an organization.

AI applications over the next 5-10 years will be fascinating, to be sure. Today’s hype may be more than the reality, but plenty of surprises will be ahead as this market evolves.

Artificial Intelligence, Security

By Bryan Kirschner, Vice President, Strategy at DataStax

Artificial intelligence is something developers are excited to work on. So much so that many enterprises give their AI systems names to better tout their innovations and aspirations to the world (Halo at Priceline or  Michelangelo at Uber, for example).

But, as the saying goes, when it comes to the typical consumer, people don’t want to buy a quarter-inch drill; they want a quarter-inch hole. What users really care about is how well your app, website, or customer service process satisfies their goals. They generally don’t care a whole lot about the “how” under the hood.

A new survey, conducted by Wakefield Research on behalf of DataStax, hammers this home. Based on responses from 1,000 U.S. adults, Wakefield found that a majority don’t realize how often they interact with AI in products, services, and experiences. 

Take fraud detection. Nearly two-thirds (65 percent) of respondents don’t identify fraud alerts from banks or credit card companies as being powered by AI. This is, in fact, a measure of how AI consistently gets its job done so well for the consumer that it fades into the background.

How does AI make your customers feel about your organization?

Imagine a world where customers are surprised that it was even possible to detect potential fraud in real time. Imagine them feeling relief every time a legitimate credit card swipe was approved or resigned to the fact that–yet again this month–there were unauthorized charges on their bill.

AI would be front page news for all the wrong reasons, with stories about the ins and outs of balky systems that are sources of constant frustration–as well as why financial institutions were subjecting customers to such constant aggravation.

This reveals one key question you can ask to accelerate your AI strategy: how will it make your customers feel? And let’s be clear: if you are doing it right, the feelings won’t be about your AI, it will be about your organization.

In the case of fraud detection, it’s a latent confidence you’d have to poke at to bring to the surface (although it would be an obvious and utter disaster in the case of a breach). But consider these sentiments:

You knew just what I was in the mood to watch.

You really saved my bacon with the alternative product recommendation I could get the same day.

You recommended the perfect gift.

Most consumers (64%) don’t give AI the credit  for a song or movie recommendation from a streaming service. But they love the results. Eighty-seven percent find relevant recommendations “highly valuable.”

Wakefield Research/DataStax

And 60% of shoppers take advantage of relevant recommendations they come across while browsing or shopping online, including 54% of millennials, who call these “a great benefit.” Nearly 3 in 4 (72%) trust a company more when they receive relevant recommendations (including 83% of millennials). And nearly 1 in 5 (21 percent) are “extremely likely to return” after receiving good recommendations (46% of millennials say as much).

Make AI a superpower for your organization

Real-time AI has crossed two critical thresholds. First, best-of-breed tools for delivering top-quality experiences are open source and available as-a-service, on demand, to anyone. But just as important: it has become a powerful determinant of satisfaction and loyalty among consumers.

Thus, turning the tools into a competitive edge requires a vision for how you align what technology can do with the qualitative and emotional brand relationship only you can build with your customers.

Is it “you expand my horizons” or “you never let me down”? Is it “I always get more for less” or “I’m never, ever late”? (Lyft smartly flips the script on this apparent dichotomy by offering a continuum of options from “Wait and Save” to “Priority Pickup.”)

Brands are built by consistently delivering on a promise to customers. Modern AI has the power to pound away at this goal by doing so millions (or billions) of times, as often as every second of every day–while getting better throughout the process. Connecting those dots is how you can make it a superpower for your organization–even though consumers might not give AI the credit for the results.

Learn more about real-time AI here.

About Bryan Kirschner

Bryan is Vice President, Strategy at DataStax. For more than 20 years he has helped large organizations build and execute strategy when they are seeking new ways forward and a future materially different from their past. He specializes in removing fear, uncertainty, and doubt from strategic decision-making through empirical data and market sensing.

Artificial Intelligence, IT Leadership

By Bryan Kirschner, Vice President, Strategy at DataStax

In their 2020 book Competing in the Age of AI, Harvard Business School professors Marco Iansiti and Karim Lakhani make some bold predictions about the winning enterprises of the future.

These organizations, which they refer to as “AI factories,” build a “virtuous cycle between user engagement, data collection, algorithm design, prediction, and improvement,” unlocking new paths to growth as software moves to the core of the enterprise.

A little more than two years after the publication of their seminal work, data gathered from IT leaders and practitioners lend a lot of credence to Iansiti and Lakhani’s hypotheses — particularly those regarding the kind of technology architectures and strategies that engender success with AI.

The AI factory

Successful AI companies — think Apple, Netflix, Google, Uber, or FedEx — build innovative applications and, as they scale, start the flywheel of data, growth, and improvement spinning by gathering ever-growing amounts of real-time data, accessing it instantly, and tuning their predictions.

User experiences become more personal and intuitive; key decisions can be made nearly instantaneously; and predictions can occur in real-time, empowering a business to improve outcomes in the moment.

This unlocks new paths to growth: in the authors’ words, as AI factories “accumulate data by increasing scale (or even scope), the algorithms get better and the business creates greater value, something that enables more usage and thus the generation of even more data.”

For more traditional firms to achieve this kind of success requires a host of changes in both their operating models and technology profiles.

Open-source software and AI success

The State of the Data Race 2022 report is based on a survey of over 500 IT leaders and practitioners that delved into their organizations’ data strategies.

For the purpose of this analysis, responses were divided into three groups:

those where both AI and ML are already in widespread deploymentthose where AI and ML are at most in the pilot phase or early daysthose in between these two extremes, characterized as being in “limited deployment”

The study assumed the organizations with AI/ML widely in production provide useful information about the evolving shape of the “AI factory” and looked for differences across the three stages of maturity.

Iansiti and Lakhani wrote that AI factories will evolve “from a focus on proprietary technologies and software to an emphasis on shared development and open source” because the competitive advantage they enjoy comes from data they accumulate — not the software they develop in-house.

The survey data backs this up in spades. A strong majority of each of the three AI/ML groups considers open-source software (OSS) at least “somewhat” important to their organization (73%, 96%, and 97%, respectively, ordered from “early days” to “wide deployment”).

But ratings of “very” important closely track AI/ML maturity: 84% of companies with AI/ML in wide deployment describe OSS this way (22%of “early days” organizations do, and this jumps to 46% of those with AI/ML in limited deployment).

Perhaps even more striking, organizations not using OSS are a tiny minority (1%, 1%, and 7%, ordered from “wide deployment” to “early days”). But a majority of those with AI/ML in wide deployment (55%) join companies like The Home Depot in having a company-wide mandate for use of OSS.

Real-time data and AI

Consider the AI leaders mentioned above. These companies have assembled technology infrastructures that enable instantaneous changes and decisions based on real-time feedback. Relying on day-old data and batch processing to update the routing of a package to ensure on-time delivery just doesn’t cut it at FedEx.

So, it isn’t surprising that Iansiti and Lakhani report that AI factories lean into real time. “The top enterprises … develop tailored customer experiences, mitigate the risk of customer churn, anticipate equipment failure, and enable all kinds of process decisions in real time,” they say.

Much like with OSS, findings from The State of the Data Race point to real-time data (and the technology architecture that enables it) as a matter of core strategy for the AI leaders. The substantial use of this correlates with AI maturity: 81% of companies that have broadly deployed AI/ML say real-time data is a core strategy. Forty-eight percent of organizations with limited AI/ML deployment describe it as a core strategy; the figure was 32% for companies in the early stages of AI/ML.

But among the advanced group, a full 61% say that leveraging real-time data is a strategic focus across their organization (four times that of organizations in the early days, and more than twice that of those with limited deployment). And 96%of today’s AI/ML leaders expect all or most of their apps to be real time within three years.

This makes sense: as an enterprise intentionally rewires its operations to make the most of AI/ML, it becomes especially important to eliminate any arbitrary architectural barriers to new use cases that require “speed at scale” anywhere in the business.

Today’s OSS as-a-service ecosystem makes that possible for everyone, freeing the future organization to make the most of its unique customer interactions and datasets.

Uniphore: A case study in real-time data, AI, and OSS

Uniphore helps its enterprise customers cultivate more fruitful relationships with their customers by applying AI to sales and customer service communications. The company relies on real-time data to quickly analyze and provide feedback to salespeople upon thousands of customer reactions during video calls.

“We have about fourteen different AI models we run in real time to coalesce the data into something meaningful for our clients,” says Saurabh Saxena, Uniphore’s head of technology and VP of engineering. “Any kind of latency is going to have a negative effect on the real time side.”

“Without the ability to process data in real-time, our solution really wouldn’t be possible,” he adds.

To get “the speed they need,” Uniphore relies on open-source Apache Cassandra® delivered as a service via DataStax (my employer) Astra DB. Its performance and reliability are key to ensuring Uniphore’s system is something every salesperson is motivated to rely on in order to be more effective in the moment.

But winning adoption among line staff points to another of Iansiti and Lakhani’s insights on the implications of AI for senior management. As the latter explained in a 2021 interview, “AI is good at predictions” — and predictions are “the guts of an organization.” Senior executives need to constantly ask, “Do I have data now to improve my prediction power — my accuracy, my speed?”

As Uniphore points out, sales forecast accuracy is something most sales leaders are concerned about. As a knock-on effect of using Uniphore’s tools, quantitative data on sentiment and engagement can flow into sales forecasts without the need for more staff time. In addition to the direct uplift that sellers experience, forecasts improve– — management to spend their time on more important things, like investing for growth, with greater confidence.

This closes the loop on Iansiti and Lakhani’s insight that AI factories can unlock a more powerful operating model over and above the benefits of individual use cases and point solutions.

Building an AI factory

Organizations that leaned into the insights in Competing in the Age of AI may have stolen a march on their competition. Judging from our survey data, they’ve been amply rewarded for doing so. The good news is that they’ve proven best practices for success — and the tools you need to accelerate your own progress on the journey to becoming an “AI factory” are ready and waiting.

Learn how DataStax enables AI-powered apps

About Bryan Kirschner:

Bryan is Vice President, Strategy at DataStax. For more than 20 years he has helped large organizations build and execute strategy when they are seeking new ways forward and a future materially different from their past. He specializes in removing fear, uncertainty, and doubt from strategic decision-making through empirical data and market sensing.

Artificial Intelligence, IT Leadership

A lawsuit has been filed against 13 current and former IBM executives, including CEO and Chairman Arvind Krishna and former CEO Ginni Rometty, accusing the company of securities fraud — bundling mainframe sales together with those of poorly performing products in order to make them appear more profitable than they actually were.

The lawsuit was filed on January 13 in the U.S. District Court for the Southern District of New York,  and seeks class action status for anyone who purchased IBM shares during the period April 4, 2017, to Oct. 20, 2021.

The complaint alleges that the company and some of its executives “knowingly or recklessly engaged in a device, scheme, or artifice to defraud, engaged in acts, practices, and courses of business conduct designed to deceive investors.”

Essentially, it’s alleged that IBM promoted its less popular cloud, analytics, mobile, social, and security products (CAMSS) products as “growth opportunities,” allowing investors to think they were much in demand when, in fact, they were being tacked onto three- to five-year mainframe Enterprise License Agreements (ELA) that were popular with large banking, healthcare, and insurance company customers.

“Defendants misled the market, engaging in a fraudulent scheme to report billions of dollars in mainframe segment and other non-strategic revenues as Strategic Imperatives and CAMSS [“Cloud,” “Analytics,” “Mobile,” “Security,” and “Social,”] revenues, enabling Defendants to report publicly materially distorted segment information,” the lawsuit states. “Defendants portrayed Strategic Imperatives and CAMSS as growing materially beyond actual growth, materially misrepresenting IBM’s shift away from its stagnant legacy mainframe segment.”

According to IBM, “strategic imperatives” are products and initiatives that provide “differentiation driven growth and value.”

IBM is also alleged to have reallocated revenue from its non-strategic Global Business Services (GBS) segment to the company’s Watson-branded AI products — a strategic imperative included in the CAMSS product portfolio — in an attempt to convince investors that the company was successfully expanding beyond its legacy business. As a result, “IBM securities traded at artificially inflated prices” resulting in financial damage to people purchasing company shares during the period covered by the lawsuit, according to the lawsuit.

In response to a request for comment, IBM emailed a statement that said, “IBM’s long-standing commitment to trust, integrity and responsibility extends across all aspects of our business operations. A similar complaint has already been voluntarily dismissed.” 

 In fact, the same complainant who filed the lawsuit last week — the June E. Adams Irrevocable Trust for the benefit of Edward Robert Adams and others who may join the lawsuit   — filed a similar lawsuit last April, then filed a notice in September moving for voluntary dismissal of the case “without prejudice,” reserving the ability to refile the suit.

The reason behind the move to abandon that case was due to disagreement with the lead law firm at the time about how to handle the case, according to The Register, which first reported on the new case filed last week. The law firm submitting the new lawsuit, The Rosen Law Firm, declined to comment.

The case filed last April alleged that IBM had bolstered its stock price and deceived shareholders by moving revenue from its non-strategic mainframe business to its strategic business segments. This previous lawsuit further alleged that by misrepresenting the true nature of CAMSS revenue, it allowed IBM executives to take home larger bonuses than they otherwise would have received.

While this new lawsuit once again alleges that IBM strategically shifted revenue, it omits the accusation related to executive bonuses.

According to the PACER electronic records system, the new case has been referred to District Judge Vincent L. Briccetti, who will have to decide whether to certify class-action status for the lawsuit.

Briccetti is currently adjudicating another ongoing lawsuit filed against IBM. In that case, filed in March last year, Gerald Hayden, an ex-IBM employee, accuses IBM of theft of trade secrets and intellectual property.  Hayden alleges that, while he worked for IBM, the company unlawfully used his proprietary business method — A2E — that he had developed to streamline enterprise sales.

Hayden’s lawsuit alleges that IBM, after promising it would protect his trade secrets, used A2E on projects that he was not working on, moving some of his clients to new projects in areas of the company including cloud and Watson — essentially transferring clients that he had attracted via the A2E methodology from GBS to newer strategic projects.

“IBM thus used A2E’s value proposition to drive IBM’s claimed reinvention of itself as a leader in the hybrid cloud computing industry and as an invaluable consultant to the financial services,” according to the lawsuit. “To add insult to injury, after stealing Plaintiff Hayden’s proprietary A2E business methodology and stripping him of his client base, IBM shortly thereafter terminated Plaintiff for ‘lack of work.’ “

(Additional reporting by Marc Ferranti.)

IBM, Legal, Technology Industry

Next generation chatbots are now writing poetry and giving math lessons, but these smart applications have a bigger job to do. Advanced chatbots simulate human interaction via complex artificial intelligence (AI) processes, or conversational AI. As business-ready systems, conversational AI is joining mainstream tech to deliver strategic benefits to customers and employees. For companies looking to adopt or expand their use of conversational AI, there’s quite a bit to understand and consider. 

Now that humans and machines are talking to each other, decision-makers will need clarity around the capabilities—especially as they vet various products and platforms. It helps to start by defining some key terms.

Artificial intelligence (AI): A wide-ranging category of technology that allows computers to “strive to mimic human intelligence through experience and learning.”[1] Common AI applications involve analysis of language, imagery, video, and data. Machine learning (ML): In its definition of AI, Gartner cites ML as one of AI’s notable “advanced analysis and logic-based techniques,”[2] whereby computer systems can learn from their experiences without explicit programming. Natural language processing (NLP) focuses on machine reading comprehension through grammar and context, enabling it to determine the intended meaning of a sentence.  Known for applications such as voice-to-text and language translation, NLP uses AI and often ML to enable a computer to understand spoken or written human language. Natural language generation (NLG) focuses on text generation, or the construction of text in English or other languages, by a machine and based on a given dataset.Conversational AI: This advanced application of NLP is what allows people to have a spoken or written conversation with a computer system. At their best, conversational AI systems closely match human conversation—passing a measure called the Turing test.[3] Here’s how it works from a technical perspective: During the automatic speech recognition (ASR) stage, a person may ask a question and the application converts that audio waveform to text. During the NLP phase, the question is interpreted, and the device generates a smart response. Finally, the text is converted back into audio for the user during the text-to-speech (TTS) stage. 

A Quick Rundown of How Conversational AI Works

Asking a smart phone whether it’s going to rain, telling a virtual assistant to play ’90s hip hop, requesting a navigation system give directions to a new sushi restaurant—each are examples of interacting with conversational AI. By speaking in a normal voice, a person can communicate with a device that understands, finds answers, and replies with natural-sounding speech.

Conversational AI may seem simple to the end user. But the technology behind it is intricate, involving multiple steps, a massive amount of computing power, and computations that occur in less than 300 milliseconds. When an application is presented with a question, the audio waveform is converted to text in what’s known as the automatic speech recognition stage. Using NLP, the question is interpreted and a response is generated. At the next step, called text-to-speech, the text response is converted into speech signals to generate audio. 

Why Customers and Employees Prefer Conversational AI
Most people have experienced the frustration of talking to a legacy chatbot, and perhaps even resorted to anger or shouting “Representatitive!”. But once chatbots are enhanced with conversational AI capabilities, research shows customer satisfaction rates to be three times higher, attributed to shorter wait times and more accurate, consistent customer support.[4]

For employees, conversational AI can reduce stress and boost productivity by handling most low-level tasks and easing their day-to-day human-machine interactions. This frees up staff for other valuable and higher-level functions, benefiting customers and increasing morale.

Overall, for companies, the benefits may seem obvious: more productive staff and better customer service leading to increased productivity as well as higher customer satisfaction and retention rates. An additional benefit comes from the learning and training of models that continually improve and enhance employee and customer experiences.

Conversational AI in Action, From Retail to Healthcare to Real Estate

In constant search of competitive advantage, companies are increasing their investments in AI to the tune of a projected $204 billion by 2025.[5] Across industries, the technology promises to deepen customer insights, drive employee efficiency, and accelerate innovation. 

In retail, conversational AI is giving shoppers a streamlined experience with call centers and customer service interactions. As the clunky chatbots of yore are replaced with savvy AI chatbots, customers can quickly get their questions answered, receive product recommendations, find the proper digital channel for their inquiry, or connect with a human service agent. 

In healthcare, applications for conversational AI can support telehealth patient triage to identify potential medical conditions. Systems can also be trained to securely manage patient data—making it easier to access information such as test results or immunization records. And the technology can support patients who are scheduling an appointment, checking on insurance eligibility, or looking for a provider.

In real estate, conversational AI tools are being applied to the time-sensitive lead generation process, automating functions for accuracy and efficiency. Chatbots are also handling initial conversations to assess what a customer is looking to buy or sell. Given AI’s ability to handle thousands of calls per day, a program can be integrated with the customer relationship management system, or CRM, to create more positive experiences.

Five Questions to Ask Before Deploying a Conversational AI System

Once a company is ready to explore a conversational AI project, there will be groundwork. Here are five essential questions—and clues to finding the answers.

What kind of hardware do you need? The answer depends on the application scope and throughput needs. Some implementations rely on ML tools and run best on high-performance computing. Others may be more limited in scope. In any case, Dell Technologies Validated Designs offer tested and proven configurations to fit needs based on specific use cases.Which user interface options will your project support? Whether it’s a text-only chatbot or the more user-friendly voice interface, the decision must be based on what’s best for the customer and the budget.What platforms will be supported? Determine how customers might access the chatbot—via mobile app, web, social media—and think about whether to integrate with popular voice assistants. Will you build your own or rely on a vendor? Doing it in-house requires expertise and time but offers more control. If selecting a vendor, consider whether one vendor or multiple vendors will be needed for the end-to-end system. What kind of infrastructure will you need? This depends on whether the implementation will be hosted in a private or public cloud service. For those hosting in their own data centers, likely for compliance or security reasons, be sure the vendor’s systems are designed specifically to meet the speed and performance for conversational AI. 

As consumers become more familiar with AI, using it to create art and pay bills and plan their workouts, the technology holds greater professional promise. Conversational AI is already supporting a number of essential business functions—a boon for customers, staff, and the bottom line. Executives can set the foundation for their own advanced chatbots and other applications by ensuring their IT systems are ready for innovation. 

Read the Guide to Conversational AI for Financial Services, and explore AI solutions from Dell Technologies and Intel

IT Leadership

What is the future of analytics and AI? And how can organizations thrive in an era of disruption? We asked Bryan Harris, Executive Vice President and Chief Technology Officer of analytics software company SAS, for his perspective.

Q: What is your advice to technology leaders for improving organizational resiliency?

A: Right now, we are all in a race against disruption and data. Customer buying habits have changed. Work-life balance has changed. The financial climate has changed. So, how do you establish a data-driven culture to identify and adapt to change? Or, in other words, what is the learning rate of your organization?

This is why executing a persistent data and analytics strategy is so important. It allows you to create a baseline of the past, identify when change happens, adapt your strategy, and make new, informed decisions. This process is your organization’s learning rate and competitive advantage.

Q: How can AI and analytics help business and technology leaders anticipate and adapt to disruption?

A: We are creating data that is outpacing human capacity. So the question becomes: how do you scale human observation and decision-making? AI is becoming the sensors for data in the world we’re in right now. So, we need analytics, machine learning and artificial intelligence (AI) to help scale decision-making for organizations.

Q: What best practices do you recommend around developing and deploying AI?

A: When we talk to customers, we first show them that the resiliency and agility of the cloud allows them to adapt quickly to the changing data environment.

The second step is lowering the barrier of entry for their workforce to become literate in analytics, modeling and decision-making through AI, so they can scale their decision-making. Everyone has a different maturity spot in that curve, but those who achieve this outcome will thrive – even in the face of disruption.

I recommend the following best practices:

Think about the ModelOps life cycle, or the analytics life cycle, as a strategic capability in your organization. If you can observe the world faster, and make and deploy insights and decisions as part of AI workloads faster, you can see the future ahead of time. This means you have a competitive advantage in the market.Innovate responsibly and be aware of bias. We give capabilities and best practices to our customers that allow them to understand the responsibility they have when scaling their business with AI. And we are taking a practical approach to helping customers adhere to the ethical AI legislative policies that are emerging.Ensure explainability and transparency in models. You won’t have adoption of AI unless there is trust in AI. To have trust in AI, you must have transparency. Transparency is critical to the process.

Q: What does the future hold for AI and analytics?

A: Synthetic data is a big conversation for us right now. One of the challenges with AI is getting data labeled. Right now, someone must label, for example, a picture of a car, a house or a dog to train a computer vision model. And then, you must validate the performance of the model against unlabeled data.

Synthetic data, in contrast, allows us to build synthetic data that is statistically congruent to real data. This advancement represents a huge opportunity to help us create more robust models — models that aren’t even possible today because conventional data labeling is too challenging and expensive. SAS lowers the cost of data acquisition and accelerates the time to a model.

If, because of this innovation, they get insights about the future, companies gain a competitive advantage. But they must do it responsibly, with awareness of the bias that AI may inadvertently introduce. That is why we provide capabilities and best practices to our customers that allow them to understand the responsibility they have when scaling their business with AI.

For more information, download the SAS report – “4 Winning Strategies for Digital Transformation” – here.

Artificial Intelligence

Contact centers are evolving rapidly. The days of single-channel, telephony-based call centers are long gone. This old model has given way to the omnichannel customer experience center.

In legacy call centers, the customer’s pathway through sales or service was relatively linear. Call in, speak to an agent, and (hopefully) resolve the issue. In this system, the manager’s focus was strictly on ensuring there would be enough well-trained staff to handle every call as efficiently as possible.

Nowadays, however, the customer journey is more complex, and the path to successful customer experience (CX) may weave its way through various channels, touching both human and robot agents along the way. Today’s managers must not only build an adequate staff, but they must also choose the right solutions to effectively meld together technological and human elements to deliver a near-flawless CX. 

Although many solutions have proved important for managers seeking to create successful contact centers, none are more important than the cloud and conversational AI. You might think of these as the twin pillars of success for today’s contact centers. However, as we’ll discuss here, they’re not sufficient on their own. There’s a third pillar to consider: quality assurance, or dedication to ensuring a finely tuned customer experience at every stage in the customer journey.

The cloud makes the contact center omnipresent

It looks like we’ve reached the tipping point for cloud adoption in contact centers. Deloitte reports that 75% of contact centers plan to migrate their operations to the cloud by mid-2023, if they haven’t already done so. IDC forecasts that investments in cloud solutions will account for 67% of infrastructure spending by 2025, compared to only 33% for non-cloud solutions. Genesys, a major contact center provider, recently announced that, going forward, it will focus its efforts on its Genesys Cloud CX software rather than its on-premises solutions.  

Considering the cloud’s potential, it’s not surprising to see that it’s taking over. Fundamentally, the cloud allows contact centers to keep pace with the changing expectations of employees and customers simultaneously.

The pandemic quickly changed what both groups were looking for. Employees came to expect more accommodating remote work arrangements, and those expectations have held strong even in 2022. According to research by Gallup, only 6% of workers who can do their jobs remotely actually want to return to a full on-site arrangement. Expectations for CX, meanwhile, have continued to rise to new heights, whether in terms of omnichannel service or personalized experiences.

The cloud makes it much easier for contact centers to meet these expectations. Without the need to rely on legacy, brick-and-mortar infrastructure, remote agents can deliver service to customers from anywhere at any time. Plus, the cloud more effectively facilitates seamless omnichannel service delivery and efficient software updates.

From setup to ongoing execution, the cloud is simply easier to manage. With no telecom hardware to purchase, installation and setup happen more quickly. And contact centers can rapidly scale up and down as needed, and when needed, allowing them to effectively manage costs.

The net effect of these benefits is that the cloud creates a new kind of contact center — one that’s omnipresent to deliver a modern customer experience from anywhere and to anyone.

Conversational AI transforms CX

One of the key benefits of moving to the cloud is the availability of conversational AI that can power self-service solutions. This technology, which is indispensable to chatbots and IVR, enables bots to interact with customers in natural — even human — ways.

Thanks to powerful components of AI, such as natural language processing and machine learning, bots are increasingly able to provide much of the service customers seek. In fact, in today’s self-service economy, conversational AI allows consumers to solve many of their own issues. Even more, the machine learning capabilities of AI allow it to easily and quickly collect customer data and use it to personalize the service experience. Unsurprisingly, organizations that employ conversational AI see a 3.5-fold increase in customer satisfaction rates.

That boost in customer satisfaction stems not only from offering personalized self-service, but also from organizations making the most of their human service. While bots handle many of the simpler requests, they reserve agents’ time for handling more complex matters. Ultimately, companies that deploy them can improve customer service while also cutting costs by between 15% and 70%.

This AI-powered CX transformation is already well underway in many industries. Banks use conversational AI to power customer self-service with simple tasks, like money transfers and balance inquiries. Hotels employ it to offer streamlined booking and concierge services. And retailers put it to work engaging customers in more personalized ways.

These are only a few of the basic benefits that forward-thinking companies can gain from deploying conversational AI. Its more advanced forms will power a new kind of proactive CX in the years ahead, shaped by powerful tools like sentiment analysis. 

True success requires a third pillar: quality assurance

Although critical for today’s contact centers, those two pieces are incomplete without the third pillar of quality assurance.

The expanded service capacities enabled by the cloud and conversational AI add new layers of complexity to a contact center’s CX delivery. Cloud migration, for instance, often involves bringing together many disparate legacy systems and remapping the entire customer journey. It requires extensive testing and mapping to make sure it’s done right. 

And as powerful as conversational AI is, it still requires a lot of human guidance to ensure it’s doing its job correctly. Without the capacity for that guidance, IVR or chatbot solutions may cause more CX problems than they solve. They can also be more costly — defects discovered in the IVR or chatbot production environment are much more expensive to undo than they would be when discovered in design.

The best way to provide cost-effective quality assurance is through a robust set of testing solutions that can work with any cloud, IVR, or chatbot solution that a contact center uses. As a platform-agnostic CX assurance solution, that’s exactly what Cyara is designed to do. 

With a powerful solution like Cyara, businesses can speed up cloud migration, correct voice quality issues, load-test IVRs, and performance-test chatbots, regardless of which solutions they use. They can even run more advanced chatbot tests to see how well they follow natural human conversation flows and recognize various speech patterns.

This kind of quality assurance allows contact centers to jump to the cloud and deploy conversational AI with confidence, knowing that both will push their CX forward. Together, these three pillars provide a firm foundation for contact centers of the future.

Ready to get started? Cyara can provide assurance for your cloud migration so you can start building these pillars. Reach out to get started today.

Digital Transformation

Nvidia used to be just a graphics chip vendor, but CEO Jensen Huang wants you to know that the company is now a full-stack computing service provider, and that he may be an artificial construct.

With such lofty ambitions, Nvidia is moving into the cloud, delivering both hardware and software as-a-service. At the company’s GTC Fall conference last week, Huang showed off a few new toys for gamers, but he spent most of his keynote speech outlining the tools Nvidia offers CIOs to accelerate computing in the enterprise.

There was hardware for industrial designers in the new Ada Lovelace RTX GPU; a chip to steer self-driving vehicles while entertaining passengers; and the IGX edge computing platform for autonomous systems.

But it wasn’t only hardware. Software (for drug discovery, biology research, language processing, and building metaverses for industry) and services including consulting, cybersecurity, and software- and infrastructure-as-a-service in the cloud were there too.

Huang punctuated his keynote with demos of a single processor performing photo-realistic, real-time rendering of scenes with natural-looking lighting effects, an AI that can seamlessly fill in missing frames to smooth and speed up animation, and a way of training large language models for AI that allow them to respond to prompts in context-dependent ways. The quality of those demos made it at least somewhat plausible when, in a videoconference with journalists after the keynote, the on-screen Huang quipped, “Don’t be surprised if I’m an AI.”

Joking aside, CIOs will want to pay serious attention to Nvidia’s new cloud services play, as it could enable them to deliver new capabilities across their organizations without increasing equipment budgets. In an age when hardware costs are likely to climb and the industry’s ability to pack more transistors into a given area of silicon is stalling, challenges still exist for many.

“Moore’s law is dead,” said Huang, referencing Gordon Moore’s 1965 statement that the number of transistors on microchips will double about every two years. “And the idea that a chip is going to go down in cost over time, unfortunately, is a story of the past.”

Many factors are contributing to the troubles of chip makers like Nvidia, including difficulty obtaining vital tooling and the rising cost of raw materials such as neon gas (supplies of which have been affected by the war in Ukraine) and the silicon wafers chips are made from.

“A 12-inch wafer is a lot more expensive today than it was yesterday,” Huang said. “And it’s not a little bit more expensive, it is a ton more expensive.”

Nvidia’s response to those rising costs is to develop software optimized so customers get the most out of its processors, helping redress a price-performance balance. “The future is about accelerated full stack,” he said. “Computing is not a chip problem. Computing is a software and chip problem, a full stack challenge.”

Fine-tuning NeMo

To underline that point, Nvidia announced it’s already busy optimizing its NeMo large language model training software for its new H100 chip, which has just entered full production. The H100 is the first chip based on the Hopper architecture that Nvidia unveiled at its Spring GTC conference in March. Other deep learning frameworks being optimized for the H100 include Microsoft DeepSpeed, Google JAX, PyTorch, TensorFlow, and XLA, Nvidia said.

Nvidia Hopper

NeMo also has the distinction of being one of the first two Nvidia products to be sold as a cloud-based service, the other being Omniverse.

The NeMo Large Language Model Service enables developers to train or tailor the responses of large language models built by Nvidia for processing or predicting responses in human languages and computer code. The related BioNeMo LLM Service does something similar for protein structures, predicting their biomolecular properties.

Nvidia’s latest innovation in this area is to enable enterprises to take a model built from billions of parameters and fine-tune it using a few hundred data points, so a chatbot can provide responses more appropriate to a particular context. For example, if a chatbot asked, “What are the rental options?” it might respond, “You can rent a modem for $5 per month,” if it were tuned for an ISP; “We can offer economy, compact and full-size cars,” for a car rental company; or, “We have units from studios to three bedrooms,” for a property management agency.

Such tuning, Nvidia said, can be performed in hours, whereas training a model from scratch can take months. Tuned models, once created, can also be called up using a “prompt token” combined with the original model. Enterprises can run the models on premises or in the cloud or, starting in October, access them in Nvidia’s cloud through an API.

Omniverse Cloud

Nvidia’s Omniverse platform is the foundation of the other suite of cloud services the company offers.

Huang described the platform as having three key features. One is the ability to ingest and store three-dimensional information about worlds: “It’s a modern database in the cloud,” Huang said. Another is its ability to connect devices, people or software agents to that information and to one another. “And the third gives you a viewport into this new world, another way of saying it’s a simulation engine,” Huang said.

Those simulations can be of the real world, in the case of enterprises creating digital twins of manufacturing facilities or products, or of fictional worlds used to train sensor networks (with Omniverse Replicator), robots (with Isaac Sim), and self-driving vehicles (with Drive Sim) by feeding them simulated sensor data.

There’s also Omniverse Nucleus Cloud, which provides a shared Universal Scene Description store for 3D scenes and data that can be used for online collaboration, and Omniverse Farm, a scale-out tool for rendering scenes and generating synthetic data using Omniverse.

Industrial giant Siemens is already using the Omniverse platform to develop digital twins for manufacturing, and Nvidia said the company is now working on delivering those services to its customers using Omniverse Cloud.

Omniverse Farm, Replicator and Isaac Sim are already available in containers for enterprises to deploy on Amazon Web Services’ compute cloud instances equipped with Nvidia GPUs, but enterprises will have to wait for general availability of the other Omniverse Cloud applications as Nvidia managed services. The company is now taking applications for early access.

Nvidia is also opening up new channels to help enterprises consume its new products and services. Management consulting provider Booz Allen Hamilton offers enterprises a new cybersecurity service it calls Cyber Precog, built on Nvidia Morpheus, an AI cybersecurity processing framework, while Deloitte will offer enterprise services around Nvidia’s Omniverse software suite, the companies announced at GTC.

As Nvidia works with consultants and systems integrators to roll out its SaaS and hardware rental offerings, that doesn’t mean it’s going to stop selling hardware outright. Huang noted that some organizations, typically start-ups or those that only use their infrastructure sporadically, prefer to rent, while large, established enterprises prefer to own their infrastructure.

He likened the process of training AI models to operating a factory. “Nvidia is now in the factory business, the most important factory of the future,” he says. Where today’s factories take in raw materials and put out products, he said, “In the future, factories are going to have data come in, and what comes out is going to be intelligence or models.”

But Nvidia needs to package its hardware and software factories for CIOs in different ways, Huang said: “Just like factories today, some people would rather outsource their factory, and some would rather own it. It just depends on what business model you’re in.”

CIO, Cloud Management