The future of applied artificial intelligence: Towards a hyper-personalized and sustainable world

Business leaders are challenged to address the Sustainable Development Goals, including reducing carbon footprints and managing energy consumption costs, while also ensuring their businesses can take advantage of the rapid pace of change and new business opportunities to advance technology, particularly artificial intelligence Intelligence empowers every department.

As an Intel Ambassador, I am excited to continue our partnership with Intel on 4th Generation® Xeon® Scalable Processors , and the potential to expand AI in the economy while helping to achieve the Sustainable Development Goals.

Through built-in accelerators and software optimization, the 4th generation Intel® Xeon® built-in accelerators are proven to deliver leading performance per watt under targeted real-world workloads. This will result in more efficient CPU utilization, lower power consumption and higher return on investment, while helping enterprises achieve their sustainability goals.

We are now in the era of generalized artificial intelligence (ABI), where multi-modal, multi-task converters from Microsoft, Google, OpenAI and others enable certain deep learning algorithms to perform Vision and natural language processing (NLP) tasks. Such powerful algorithms require powerful central processing units (CPUs) and graphics processing units (GPUs) to scale to achieve good performance.

Intel’s 4th Generation® Xeon® Scalable Processors accelerate AI workloads by 3x to 5x compared to the previous generation for deep learning inference on SSD-ResNet34, and up to 2x acceleration when training on ResNet50 v1.5 using Intel® Advanced Matrix Extensions (Intel® AMX). In addition, in terms of AI performance, the fourth generation Intel® Xeon® Scalable processors provide 10 times higher PyTorch performance than the previous generation with built-in AMX (BF16) for real-time inference and training.

As we enter an era of increasingly powerful AI algorithms, such as Transformers with self-awareness and generative AI, and the rise of AI combined with the Internet of Things (AIoT), we will need 4th generation Intel® Xeon® Scalable processors deliver more efficient and powerful CPUs, allowing AI to scale very quickly and process large amounts of data in low-latency use cases while targeting energy efficiency and reduced carbon footprint as key goals.

How can artificial intelligence help achieve sustainable development and combat climate change?

Microsoft commissioned PWC to write a report titled "How Artificial Intelligence Can Enable a Sustainable Future" , covering the potential of artificial intelligence in four areas of the global economy:

• Energy
• Agriculture
• Water
• Transportation

The report’s findings show that AI has huge potential to drive emissions reductions while also increasing jobs and economic growth in the four areas explored in the report:

• Global CO2 Emissions Volume decreased by 4%
• GPD increased by 4.4%, reaching 5.2 trillion US dollars
• Employment growth, creating 38 million jobs

Greenhouse gas emission reduction potential ( Globally up to 4%) based on assumptions across all four sectors (water, energy, agriculture and transportation) and the role that AI may play in these sectors, including but not limited to precision agriculture, precision monitoring, fuel efficiency, optimized input use, Improve productivity.

In addition, the US EPA and BCG outline the benefits of a standalone 5G network (see the right side of the figure above). The resulting SA 5G network can massively expand AIoT (applied to Internet of Things devices and sensors). (artificial intelligence), and increased automated processes with machine-to-machine communication, may increase job opportunities and potentially reduce greenhouse gas emissions.

The latest Intel® Accelerator engines and software optimizations help improve energy efficiency in artificial intelligence, data analytics, networking and storage. Organizations can leverage built-in accelerators to improve average performance per watt for target workloads by 2.9x compared to the previous generation. This will lead to more efficient CPU utilization, lower power consumption and higher return on investment, while helping enterprises achieve sustainable development and carbon emission reduction goals.

Enabling Change

The 4th generation Intel® Xeon® Scalable processors improve energy efficiency due to design innovations with built-in accelerators. This allows specific workloads to consume less energy while running faster.

The results per watt are (on average) 2.9x that of 3rd generation Intel Xeon processors, while also allowing for massive scaling of workloads required for the new era of AIoT we are entering, such as inference and learning as inference and learning increase 10x better compression, 2x better data analysis, and 3x more data analysis, all with 95% fewer cores.

Another innovation is the Optimized Power Mode feature, which when enabled saves 20% of energy (140 watts on a two-socket system) with minimal impact on performance (2 impact on specific workloads) -5%).

The convergence of standalone (SA) 5G networks will dramatically increase device connectivity and ultra-low latency environments, allowing for the massive expansion of the Internet of Things (IoT), with Internet-connected devices and sensors communicating with human users and each other (machine-to-machine ). More and more IoT devices will have artificial intelligence embedded (at the edge of the network).

Additionally, Statista predicts that by 2025, there will be a staggering 75 billion connected devices, that’s more than 9 for every person on the planet! IDC Seagate predicts that the amount of data generated will nearly triple from 64 zettabytes in 2020 to 175 zettabytes in 2025, with one-third of this data being consumed in real time! Applying artificial intelligence is essential for effective management Networking and understanding data and providing near real-time responses to users is critical.

Additionally, this new era will enable us to measure, analyze, evaluate, and dynamically respond to environments, whether it is healthcare, energy, transportation in smart cities, manufacturing, and more. Artificial intelligence capabilities and reasoning performance will be key to success in the era we are entering.

Intel® Xeon® Scalable processors deliver more network computing with lower latency while helping to maintain data integrity. When using NVMe over TCP, storage I/O per second (IOPS) is increased by 79% and latency is reduced by 45%, using Intel® Data Stream Accelerator (Intel® DSA) to accelerate CRC32C error checking compared to unaccelerated software error checking.

In an article titled "Harnessing the Power of Artificial Intelligence to Reduce Carbon Emissions and Costs," BCG predicts that applying AI technology to corporate sustainability goals could reduce 2.6 to 5.3 billion tons or 1 to US$3 trillion in added value.

The process to achieve this goal includes:

• Monitoring emissions
• Predicting emissions
• Reducing emissions

BCG believes that industries with the greatest potential for reducing greenhouse gas emissions due to the application of artificial intelligence include: industrial products, transportation, pharmaceuticals, consumer goods, energy, and utilities.

Intel’s vision is to accelerate sustainable computing from manufacturing to products to solutions to enable a sustainable future. Organizations can help reduce Scope 3 greenhouse gas emissions by selecting 4th Generation Intel® 2.8 billion gallons of water were recycled in 2021. For the avoidance of doubt, it is noted that the statistics provided in this paragraph relate to Scope 3 emissions related to embodied carbon that do not impact operational carbon emissions, however, Scope 3 also includes operational carbon where servers play a larger part in the equation huge comparison.

Examples of use cases applying AIoT to sustainability include:

• Sensors detect whether a home is occupied, turning off lights and air conditioning, or lowering the temperature to a lower level.
• The sensor can close the window of the switch when the heating is running.
• Predict problems before they occur, such as water main breaks, monitor unexpected outages, traffic congestion points and try to reroute traffic, or modify traffic light sequences to reduce congestion
• In agriculture, in Computer vision is applied on drones to determine when crops are ripe for harvesting to reduce wasted crops and to check for signs of drought and insect infestation
• Near real-time analysis of deforestation, illegal logging.
• Renewable energy, drones applying computer vision from deep learning algorithms can inspect wind turbine blades and solar farms’ solar panels for cracks and damage, extending asset life and increasing power generation .
• Use machine learning algorithms to optimize energy storage to maximize the operating performance and return on investment of battery energy storage.

Rolnick et al. published a paper in 2019 titled “Fighting Climate Change with Machine Learning,” co-authored by leading AI researchers including Demis Hassabis, Andrew Y Ng, and Yoshua Bengio. author, explains the potential to reduce emissions by applying artificial intelligence across a company's entire manufacturing operations, from the design stage to generative design and 3D printing, supply chain optimization and prioritizing low greenhouse gas emission options to improve factories with renewable energy supplies Energy consumption, improving efficiency by detecting emissions (including predictive maintenance) and taking follow-up actions to reduce emissions from heating and cooling and optimize transportation routes.

4th Generation Intel® Xeon® Scalable processors also feature power management tools for more control and greater operational savings. For example, new optimized power modes in the platform BIOS can save up to 20% of outlet power for selected workloads with less than 5% performance impact.

In addition, the paper by Rolnick et al. explains how companies deal with retailers' unsold inventory problems. It is estimated that the fashion industry spends $120 billion on this every year! This is both an economic waste and an environmental waste. Targeted recommendation algorithms to match supply and demand, and the application of machine learning to predict demand and production needs may also help reduce this waste.

In the world of AIoT, customers can walk along a commercial street or shopping mall, and machine learning algorithms can provide personalized product recommendations based on stores nearby.

Both retail and manufacturing examples require near real-time response from AI algorithms, which is why accelerators within the CPU are an important factor in delivering enhanced performance.

The world of AIoT will require the ability to operate in power-constrained environments and respond to user needs in near real-time.

Intel enables organizations to dynamically adjust to save power as computing needs fluctuate. Intel® Xeon® Scalable processors have built-in telemetry tools that provide critical data and AI capabilities to help intelligently monitor and manage CPU resources, building models that help predict peak loads in the data center or network, and when demand is lower Adjust CPU frequency to reduce power usage. This opens the door to greater power savings, the ability to selectively increase workloads when renewable energy is available, and the opportunity to reduce the data center's carbon footprint.

Additionally, only Intel offers processor SKUs optimized for liquid cooling systems with an immersion cooling warranty add-on to help organizations further their sustainability goals.

Artificial intelligence will be everywhere, permeating the devices and sensors we use, enabling hyper-personalization at scale and near-real-time instantaneous responses to customers and users. However, to take advantage of these opportunities, business leaders need to ensure they invest in the right technology that meets the needs of the business and its customers.

We are entering an era of near-instant response, where customer engagement and dynamic response are a necessity in a world of machine-to-machine communications.

Intel® Advanced Matrix Extensions (Intel® AMX) allows efficient scaling of AI capabilities in response to user and network needs.

Dramatically accelerate artificial intelligence capabilities on the CPU with Intel® Advanced Matrix Extensions (Intel® AMX). Intel AMX is a built-in accelerator that improves the performance of deep learning training and inference on 4th generation Intel® Xeon® Scalable processors, ideal for workloads such as natural language processing, recommendation systems, and image recognition.

4th Generation Intel® Provides performance and energy efficiency benefits across workload types. 4th generation Intel® Xeon® Scalable processors deliver superior AI training and inference performance with all new accelerated matrix multiplication operations.

Additional seamlessly integrated accelerators accelerate data movement and compression for faster network connections, increase query throughput for more responsive analytics, and offload scheduling and queue management to dynamically balance multiple load between cores. To enable new built-in accelerator capabilities, Intel provides the ecosystem with operating system-level software, libraries, and APIs.

Performance improvements for 4th generation Intel® Xeon® Scalable processors include the following:

• Use fewer cores and more Run cloud and network workloads with fast encryption. Intel® QuickAssist Technology (Intel® QAT) using RSA4K on the open source NGINX web server increases client density by 4.35x compared to software running on CPU cores without acceleration.
• In the open source RocksDB engine, using Intel In-Memory Analytics Accelerator (Intel® IAA), data decompression throughput is increased by 1.91 times compared to software compression on the core without an acceleration solution, improving the database and analytical performance. Using Intel Data Stream Accelerator (Intel® DSA), memory-to-memory transfers are increased by 8.9x compared to previous generation direct memory access.
• For 5G vRAN deployments, new instruction set acceleration increases network capacity by up to 2x compared to the previous generation.

With the expansion and scale of SA 5G network, security has become a key issue in the AIoT era.

Whether deployed on-premises, at the edge, or in the cloud, enterprises need to protect data and comply with privacy regulations. 4th Generation Intel® Xeon® Scalable processors open new opportunities for business collaboration and insights, even with sensitive or regulated data. Confidential Computing provides a solution to help protect data in use through hardware-based isolation and remote attestation of workloads. Intel® Software Guard Extensions (Intel® SGX) is the most researched, updated, and deployed data center confidential computing technology on the market today, with the smallest trust boundaries of any confidential computing technology in the data center today. Developers can run sensitive data operations within the enclave to help improve application security and protect data confidentiality.

Intel’s Bosch case study provides examples of security applications in the IoT space.

This case study observes that access to raw data sets is ideal for developing AI-based analytics. This example illustrates how Bosch’s autonomous vehicle division uses Gramine, an open source project running on Intel SGX, to reduce risks associated with data or IP leakage.

By the end of this decade, we are likely to experience a significant increase in the number of advanced electric and autonomous vehicles (EV/AV) on the road.

As more renewable energy enters the grid, battery storage will become even more important. Powerful CPUs with built-in accelerators can help machine learning techniques scale across battery storage facilities to optimize the availability of energy and battery performance. This is relevant for edge and network scenarios with power and battery constraints, such as electric vehicles and power-optimized devices in smart homes and manufacturing facilities.

In this world, massive hyper-personalization enabled by AIoT will allow for instant interactions with customers in near real-time and increase efficiency, thereby reducing waste, as machine learning and data science will be able to create data from Better predict customer needs from large amounts of data.

It is conceivable that users engage in retail or entertainment activities during their commute, and EV/AV recognizes passengers through deep learning computer vision and personalizes the car environment (entertainment, etc.) based on user profiles. Self-driving cars will adapt to different users and allow users to use their time efficiently (work, entertainment). However, even before the arrival of more advanced EV/AV, there are many opportunities for businesses to leverage near real-time interactions with customers while reducing waste in the AIoT era. Such as better matching supply and demand, improving demand forecasting, identifying and matching supply chain and manufacturing processes.

4th Generation Intel® Chance.

This vision of scaling and enabling secure AIoT aligns with my personal vision to apply AI and related data analytics and digital technologies to achieve sustainable development goals while delivering a truly hyper-personalized world at scale, Enables businesses to truly respond to customer needs under real-time conditions and further customize products based on individual customer needs.

Starting this year, we will enter an exciting new era in which, for the remainder of the decade, artificial intelligence will rapidly expand to the devices and sensors around us, as well as remote cloud servers that will Continue to remain important for training algorithms, acting as a data lake, and enabling historical data analysis to improve AI learning outcomes, improve service personalization, or identify opportunities to further improve organizational operational efficiency.

We will be able to measure and assess emissions and energy consumption around us, identify waste and reduce inefficiencies.

AI algorithms throughout the network edge will require energy-efficient CPUs to run in power-constrained environments and achieve a reduced carbon footprint. 4th Generation Intel® Xeon® Scalable processors enable organizations to expand AI capabilities, deliver hyper-personalization at scale, and more effectively manage internal operations at the edge, while also helping to achieve security and sustainability goals.

The above is the detailed content of The future of applied artificial intelligence: Towards a hyper-personalized and sustainable world. For more information, please follow other related articles on the PHP Chinese website!