Atlas of AI Innovation

Artificial intelligence (AI) technologies are evolving at an unprecedented pace. We’re witnessing the impact on business and society. “AI has already created global change,” Intel CEO Pat Gelsinger wrote recently. “It has the potential to enable a responsible, inclusive, and sustainable future. We harness the power of AI to tackle critical global challenges like pandemics, natural disasters, and global public health. And we are developing AI capabilities and solutions to amplify human potential, enhance inclusion, and improve accessibility for people with disabilities.” Organizations in virtually every industry are investing in AI. IDC forecasts that global spending on AI-related software, hardware, and services would reach $154 billion in 2023, a 27% jump from 2022. According to Intel, enterprises will spend $300 billion alone on generative AI by 2026.

The Atlas of AI Innovation

Take a journey across the rapidly-changing AI landscape and explore unlimited opportunities to deliver positive change with the foundational technologies that drive these innovations.

#1 – AI challenges

#2 – AI value

#3 – AI opportunities

#4 – AI in the future

The past few years have borne witness to the wide-ranging impacts of global supply chain disruption. Manufacturers, suppliers, and distributors continue to face challenges, with the increasing digitalization of supply chains enhancing both opportunities and risk. “Disruptions are growing in scope and scale,” says Jackie Sturm, Corporate Vice President, Global Supply Chain Operations, Intel. “Supply chains are now deeply interconnected, so one incident — a regional disaster, a quality or financial issue, a man-made problem — can create multiple waves of disruption.”

Sturm points to three ongoing challenges:

These insights can enhance supply chain operations in many ways, including:

Cybersecurity: As interconnected supply chains give new digital access points to any one supplier, managing third-party risk becomes more critical than ever.

Tightening alignment between supply and demand.

Most supply chain leaders recognize the criticality of these factors: 74% plan to increase technology spending to improve supply chain resilience, transparency, and sustainability, according to the “2023 MHI Annual Industry Report.” AI technologies are gaining traction as a critical part of the solution to supply chain challenges. “AI can derive insights from massive amounts of data and rapidly draw conclusions from the type of datasets that humans can’t synthesize in a market-competitive length of time,” Sturm says.

AI’s potential, however, extends well beyond improving supply chain efficiencies, into areas such as safety and sustainability. “We can use AI to mine data for safety incidents that have occurred in our factories, and then drive systematic improvement plans for a higher level of health and safety,” says Sturm. “We can also harness AI to drive conservation of critical resources like water and energy to better support our local communities.”

Taking a business-led approach to AI

Looking to operationalize AI across your supply chain? Don’t let technology get ahead of the business case. To help maximize your investments, take a business-led approach:

Understand the business problem you’re trying to solve. Look at structural or process aspects that are causing inefficiencies or other challenges.

Get a business sponsor on board to remove organizational obstacles when data needs cross multiple functional domains or business groups.

Put a governance plan in place to help ensure that AI models are using data effectively and responsibly.

Improve your workforce’s data fluency skills. You don’t need an entire organization of AI experts, but people need to be comfortable accessing, querying, and analyzing data, so they’re not overwhelmed by AI’s capabilities.

Parker discusses AI's vital role in improving supply chain visibility and resilience

#2 – AI value

The past few years have borne witness to the wide-ranging impacts of global supply chain disruption. Manufacturers, suppliers, and distributors continue to face challenges, with the increasing digitalization of supply chains enhancing both opportunities and risk. “Disruptions are growing in scope and scale,” says Jackie Sturm, corporate vice president, Global Supply Chain Operations, Intel. “Supply chains are now deeply interconnected, so one incident — a regional disaster, a quality or financial issue, a man-made problem — can create multiple waves of disruption.”

“The core problem federated learning is trying to solve is the inability to bring together data siloes into one location or ML model,” says Micah Sheller, Staff Research Scientist at Intel. “It can be data from multiple consumer devices within an online workspace; a bunch of employee laptops; or even a set of institutions such as hospitals, financial services organizations, utility companies, or government agencies.” One example is the use of AI-driven federated learning in the healthcare industry to overcome today’s global shortage of medical expertise, which the World Economic Forum estimates will incur a shortfall of 10 million healthcare workers by 2030. “There are trained ML models for patient records, but there are too many variables even within one healthcare institution,” Sheller says. For example, there are quirks in how staff members take notes and how imaging equipment is positioned, plus age differences among different pieces of equipment — all of which have nothing to do with a patient’s pathology. “Applying federated learning to healthcare can account for these variables or biases and smash out meaningless signals to leave us with the underlying pathology,” Sheller says. “It’s the next step we have to take to improve patient care by assisting medical personnel, but it will require engagement across the globe to get the diverse data necessary to stamp out biases and meaningless signals.”

Emerging use cases for AI-driven federated learning:

Accelerating drug discovery

Mitigating financial fraud

Strengthening cybersecurity on user devices

Improving safety in autonomous vehicles

Increasing predictive maintenance models in manufacturing equipment

Personalizing customer experiences

Before capturing the benefits of federated learning, organizations need to address three critical factors: engagement, trust, and the right framework — all working in concert. “Federated learning cannot be dropped into the organization by default,” Sheller says. “IT and infosecurity professionals must engage and work together at multiple points to ensure the mechanisms being used meet required policies.” Engagement is critical, no matter the industry. In healthcare, for example, regulatory data privacy requirements must be addressed by every medical institution. Also, any organization that conducts operations in the European Union — regardless of the business it’s in — must adhere to compliance rules regarding employee and personal data, such as the General Data Protection Regulation. This compliance commonality should help in engaging as well as creating a baseline of trust. In addition, establishing trust requires transparency and execution integrity, where everyone agrees not only on the code being run but also that no one will change that code. Yet, trust must also be addressed at a higher level. For example, it is understandably difficult for rival banks to join forces to fight fraud when there are competitive issues. There must be recognition that the efforts have tangible benefits for all involved. This is where the right framework makes a difference. Software alone cannot deliver all the necessary capabilities for federated learning, especially when it comes to compute power. In addition, the framework should provide trust and transparency that enables everyone involved to see changes being made. It should also communicate those changes and support organization- and consortium-wide decision-making. “Intel brings to the table hardware security features that integrate with the Open Federated Learning (OpenFL) framework to increase trust via execution integrity,” Sheller says. “In other words, it will run and protect the code that everyone agreed on.” Rapid advances in AI and ML solutions have increased the potential power of federated learning. For example, organizations that are already running ML models to mine data can use federated learning to combine different datasets — both internally and from other organizations — to accelerate problem solving. “We want to bring people together on the issue of federated learning,” Sheller says. “Let’s collaborate on this.”

IDC's Bob Parker explains how federated learning will enable "economies of intelligence"

#4 – AI in the future

Technology permeates nearly every aspect of our lives, creating a insatiable demand for processing power. AI in particular is driving an exponential need for compute power, referred to as the "silicon demand" for AI. Welcome to the Siliconomy, a new era in which semiconductors maintain and enable modern digital success. “The Siliconomy represents the structure and conditions necessary for global expansion, where compute is foundational for greater opportunities and a better future for everyone on the planet,” says Mark Pontarelli, vice president, Corporate Strategy and Ventures at Intel. Everything and everyone will be increasingly connected across the compute continuum from cloud to edge, Pontarelli says. AI is a key part of the dynamic, because it requires massive amounts of data, sensing, and processing power to deliver actionable insights.

Silicon is the foundational piece. And demand is high:

of tech decision-makers say compute power doesn’t meet their current needs.

64%

of security leaders say their IoT and edge devices lack computing power for proper security.

73%

As many as 3,000 semiconductors are necessary to power a modern car today.

3,000

Source: Welcome to the Siliconomy

In addition, organizations will need silicon that can support their environmental sustainability goals. For example, 84% of senior IT leaders agree that technology plays an important role in their sustainable business strategies , according to Intel’s The Sustainable CTO report. “The Siliconomy is a call to action for the entire semiconductor industry,” Pontarelli says. “How can AI, for example, help drive down the cost to design and develop novel silicon? Is it too big of a leap to suggest that at some point, we’ll have GenAI-assisted silicon development?”

Business and IT leaders also have roles to play in the Siliconomy as they consider how processing power can help them meet the next wave of expansion and growth. They will need to answer questions such as:

What solutions help us compete and meet our customers’ demands? Do they require greater capacity or power?

“The key characteristics of the Siliconomy are that it is open, integrated, secure, and distributed,” Pontarelli said. “There’s tension there. How can something be both open and secure, integrated and distributed? That’s by design. To deliver the compute power that AI requires, we must drive higher levels of integration and efficiencies, securing data where it resides, and open to all.”

“The Siliconomy represents the structure and conditions for global expansion, where compute is foundational for opportunities and a better future for everyone on the planet.”

the direct impact of semiconductors on global GDP from 1995 to 2015.

$3 trillion

the expected growth rate by 2024 of chip units shipped worldwide.

10.4%

Semiconductors drive the Siliconomy

Source: Welcome to the Siliconomy

of chip demand will be driven by AI by 2025.

20%

people in the US are employed in the semiconductor industry, supporting 5x more jobs throughout the wider US economy.

300,000

More data, more devices: Silicon steps up to support AI innovation

The heightened expectations of consumers have been well documented. They want seamless, personalized shopping experiences no matter what they’re buying. In response, many retailers have undertaken significant transformations to get more out of their data and deliver a seamless omnichannel experience.

Chris O’Malley, Director of Marketing, Retail, Banking, Hospitality and Entertainment, Internet of Things Group at Intel, offers other use cases where AI is being effectively deployed:

Autonomous stores:

Autonomous checkouts help smaller-format retail organizations in particular deal with staff shortages. Large-format retailers are also piloting fully autonomous stores, which leverage functionalities such as smart shelving, automated checkouts, and inventory management to enable interaction-less shopping experiences.

Sustainability:

AI technologies can better align supply with demand to reduce waste. “AI offers a big sustainability play,” O’Malley says. For example, in fast food restaurants, “somewhere in the neighborhood of 45% to 50% of food is lost through spoilage, often because it has been sitting under a heat lamp for too long. Local AI data processing at the edge can deliver those insights to prevent food waste and drive greater efficiencies.”

All of these AI-driven use cases will require greater integration and analysis of data, which will likely require upgraded infrastructure, within the data center and at the edge, for storing and processing data and to address bandwidth contraints and latency challenges.

Amid considerations for a modern architecture that includes edge processing, compute will be a critical ingredient. Retail organizations need to trust that no matter how much data their AI solutions are processing, the CPUs, integrated GPUs, and even the new integrated neural processing units (NPUs) in their hardware can keep up. “You have to be sure your infrastructure is future-proof,” O’Malley says. “Don’t install something that can handle what you need right now, but that cannot handle what you’ll need 18 months from now. A flexible architecture that allows for additional compute is the way to move forward.”

Limitless opportunities for AI-driven innovation

Bringing AI everywhere

AI is transforming how we work and live. The technology is evolving rapidly, putting pressure on organizations to put the right underlying infrastructure and governance mechanisms in place to accelerate innovation, maximize value, and deploy anywhere, while keeping your business secure. Where will the AI journey take you next?

Learn More

Economics: Rising costs for multi-sourcing and finding alternative components for products is placing new financial strains on supply chain participants.

Data: Much of the massive volumes of data generated daily remain unplumbed due to inefficient systems or the inability to extract meaningful insights. This lack of visibility limits opportunities for innovation, while increasing risk.

Improving the quality characterization of materials to predict how they will perform in the factory to decrease defects and product returns.

Finding new suppliers globally that can more competitively solve product needs.

Analyzing the risks that impact supplier ecosystems.

How can we effectively and efficiently scale existing AI/ML implementations?

What is our data strategy? How will we uncover actionable insights at speed?

How can we make technology investments that are sustainable?

With whom can we partner to gain speed and capacity?

What tools will end users need for greater productivity and seamless access to data?

The next step in these transformations involves using AI to accelerate and scale innovation. Data in the retail sector is growing at an exponential pace, with a compound annual growth rate of 26.9% through 2026, according to IDC.

IDC also estimates that 88% of retailers have adopted some form of AI capability already, through various use cases, including automated marketing, demand forecasting, fulfillment, merchandise planning, customer service, chatbots, and more.

#3 – AI opportunities

Loss prevention:

“Retail organizations can layer AI technology onto existing security cameras, then run the footage through an intelligent solution at the edge to gain an understanding of how and why product is leaving the store to then take corrective actions,” O’Malley says.

Drive business success by bringing AI everywhere.

Between 5,000 and 7,000 semiconductors are needed to power an electric car.

5,000-7,000