Artificial Intelligence (AI) is driving game-changing capabilities with huge impacts on our lives: self-driving cars, algorithms diagnosing cancers, accelerated drug development, voice-activated devices, self-healing digital grids and self-replicating robots. The list goes on as AI, like any technology breakthrough, harbors secrets to applications that extend beyond our imagination. With semiconductors at the heart of AI applications, it’s important to understand AI’s role in the advance of digital technology and how European Union (EU) public policies can help businesses and people realize the full potential of AI.
Data is the lifeblood of AI
AI is a very broad field. In advanced manufacturing, AI is often seen as systems powered by algorithms that endow hardware with capabilities such as perception, reasoning, learning and semi- or fully-autonomous decision-making. The lifeblood of this sophisticated process is vast amounts of unstructured data gathered from various sources including sensors, images, voices or texts. The data collected from different sources is stored, structured and used as input for software to calculate, learn, adapt and, ultimately, command hardware to carry out tasks often beyond human physical and intellectual capabilities.
Today, more than 2 quintillion bytes of data is generated globally every day. By 2025, that number is forecast to grow 10-fold, providing vast opportunities for nascent AI applications.
Semiconductors drive the AI evolution
AI relies on semiconductor memory and computational power to store and process data. Indeed, with semiconductors at the core of data generation and playing a central role in sourcing and training AI algorithms, advances in chip design and manufacturing are key to enabling AI capabilities and enormous growth opportunities for the semiconductor manufacturing supply chain.
AI semiconductors need to operate with lower latency, higher computing power and faster performance than conventional chips. To thrive, AI devices need new materials, equipment, faster architectures, 3D configurations, more advanced nodes, new design and integration methodologies, and novel manufacturing processes. Computing demands on AI chips are orders of magnitude higher than for traditional applications. According to a European Commission publication, a fully autonomous car would need around 150 exaflops of data to train the vehicle’s neural networks in real time, whereas with the current technology this would take around 26 days for a single training site. Also, one of the greatest challenges of AI is delivering unprecedented computing power at low power, making energy efficiency key.
To tackle the technology challenges of AI and fulfill its promise, electronics research, design and manufacturing communities around the world are increasing investments in specialized AI chips. What’s more, Internet giants legendary for software development have started to develop homegrown specialized AI semiconductors, introducing a whole new crop of players with tremendous research and development power. According to UBS, the AI semiconductor market will grow to an eye-watering $35 billion by 2021, up from roughly $6 billion in 2016.
AI’s social, political and economic impacts: Governments need to introduce forward-looking policies
While industry continues to develop new AI applications for commercial gain, awareness is growing among governments in many developed and industrializing countries of the potential for AI to transform economies, societies and public policy. Some have launched AI strategies in recent years. For instance, the European Commission unveiled its AI strategy this year based on three key pillars:
- Invest in AI-related research and innovation, targeting spending of 20 billion EUR until 2020 and 20 billion EUR annually after 2020
- Prepare the European society for socio-economic changes
- Introduce an appropriate ethical and legal framework
AI to create new high-value jobs
A popular AI public policy discussion focuses on the technology’s impact on workforce composition and employment. Today many manual, repetitive, low-skilled jobs are giving way to automation, robots and AI. The inexorable advance of technology has always displaced workers, and AI innovation will be no different. This continuing trend toward greater efficiency and productivity is likely to reinforce the idea that certain tasks can be implemented more effectively by machines.
However, the technological transformation will also mean more high-value jobs for workers with the right skills in science, data, technology and engineering. And with its super algorithms, AI will serve to boost the productivity of many high-skilled employees. Governments worldwide should help alleviate societal concerns through education on the benefits of AI and by implementing workforce development programs that mitigate displacement risks.
Security concerns need to be addressed
AI will invariably process sensitive business and personal information as well as data flowing over critical infrastructure and public services such as defense, transportation, healthcare and energy, raising concerns over security.
Various approaches to security have surfaced in semiconductor manufacturing. One view is to secure data where it is stored. Another is to make AI chips domestically to maintain stringent controls over design and manufacturing sources. The European Commission launched the European Processor Initiative in March 2018. Bringing together 23 partners from 10 European countries, the initiative aims to develop a sovereign European supercomputing and data ecosystem to ensure trusted sources of high-end design and manufacturing sources for AI applications.
Balanced policy approach needed for data sharing and sensitive information protections
With the rise of AI, governments worldwide are accelerating the implementation of data economy and regulations to protect sensitive data traversing AI networks. The European Commission is making progress with initiatives such as “Towards Common European Data Space” and “Sharing Private Sector Data in the European Data Economy,” aimed at improving the efficient use of data across the EU. The semiconductor manufacturing industry supports such efforts but with a key caveat. While the industry acknowledges that data processing and sharing should be encouraged, it asserts that businesses should not be obliged to share sensitive data that might contain business secrets for the sake of building a data economy.
To help ensure personal data protection, the General Data Protection Regulation (GDPR) imposed strict requirements this year in the EU. The regulation retooled laws requiring how personal data can be accessed, stored and used in the EU. Its impact on the use of AI are not yet known. As AI drives the convergence of objects and humans, and of personal and non-personal data, questions remain about how personal data protection regulations will affect emerging AI applications.
Self-declaration and global standards best ways to address cybersecurity threats
In Europe, companies favour certifying their products or services and using self-declaration to demonstrate their conformity to cybersecurity in principle. However, as the topic rises in importance, various views on the recent EU Cybersecurity Act have emerged. One is that it mandates third-party cybersecurity certification for business that deliver products and services prone to cybersecurity threats in many areas including manufacturing.
The European manufacturing industry believes that self-declaration of conformity, voluntary approaches and industry-driven global standards should be the way to address cybersecurity concerns. Many industry leaders voice concerns that meeting the requirements of the third-party cybersecurity mandate would be too costly and slow down the launch of new products and services.
AI and semiconductors affected by foreign direct investment screening
In the interest of protecting national security, governments around the globe in recent years have been implementing screens on foreign direct investment in critical technologies including AI and semiconductors. In Europe, many Member States already screen such investment. Some believe the proposed EU-level initiative will jeopardize national efforts to attract foreign direct investment in capital-intensive sectors. While the semiconductor manufacturing industry backs efforts to secure critical infrastructure and public services against hostile foreign investment, it believes screening should not be used to block the much-needed friendly investment in Europe’s capital-intensive technologies such as AI and semiconductors and their capacity to generate high-value employment and growth.
SEMI raises awareness of AI and semiconductors among policymakers
SEMI Europe is a key voice in the AI policy debate. At SEMICON Europa, SEMI Europe is convening a special session on Semiconductors and AI, bringing together leaders from the European Commission, industry, academia, education, investment and startup companies for a multi-disciplinary look the AI debate.
Emir Demircan is senior manager Advocacy and Public Policy, SEMI Europe. He can be reached at email@example.com.