September 9, 2020
Matt Vasey, Microsoft
1. Balance the five centers of power: Government, Business, People, Civil Society, AI Assistants?
- The centers of power each have unique powers and responsibility
- Government has power to regulate and responsibly for long term social wellbeing
- Business can nimbly change capabilities to maximize economic output and is responsible for delivering value to shareholders that represent the top echelons of society
- Ideally People optimize their capabilities, effort and location to improve their personal and family situation in terms of financial wellbeing – with their responsibility primarily to their families long term outlook
- Civil Society wields the collective power of the people to effect change on governments and institutions to enable people to achieve their potential – with a intergenerational responsibility to improve the human condition
- AI Assistants are different. They are in large part an extension of their maker, though the end results can be different than the intended outcome. Today most of the AI assistants are created by business, using data and other inputs received from individuals. Important considerations include:
- Data received from people has value, and this value is being given to businesses and they are currently capturing the economic value of these
- People that use AI assistants provide data to businesses. This data increases the AI assistant’s effectiveness. This creates a feedback loop to continuously improve the capabilities of the assistant
- Augmenting human capability is the intent of most AI assistants, but a clear side effect of this is reduction of traditional work available to human workers.
- Managing bias is critical to ensure AI assistant outcomes map well to outcomes
- Ethical sourcing and use of data is essential to stricking a balance between the value AI delivers versus the cost on society it extracts
- Balancing the respective powers to ensure desirable outcomes will require new skills and thinking from each of the centers of power
- Civil Society must educate people on the benefits and risk of AI
- Ethic’s training for AI creators is critical to ensure that the benefits reach the broadest set of end users
- Users need to be educated the benefits of AI on an ongoing basis as the tech is rapidly changing and large cohorts of people can be left behind if this is not done
2. Government’s have already stepped up on the regulation of data with regulation schemes like GDPR. More needs to be done – especially in the areas of:
- Explain-ability of AI (Why did an AI come to a specific conclusion)
- Providing clear control of personal data in the hands of the creator
- Spreading the economic value of the data back to the creator of the data
3. Businesses need to have the right incentives for AI efforts that can drive a fair distribution of the benefits
- Tax and incentives combined with regulation are likely to deliver better results than
- Taxes for worker replacement, exploitive data collection, etc.
- Incentives for training and revenue sharing initiatives that spread value to a broad set of stakeholder
2. Establish a new balance of power among nations and Transform the world economy into one with opportunities for all people?
The old balance of power has been driven by the balance between 3 factors: Access to low-cost labor, energy, and raw materials, 2. Institutional and government support of industry providing access to capital, political stability, and international trade, and 3 Access to innovation that drives marginal production costs down, and worker productivity up.
The new AI Economy will be built on three pillars, and nations will need to balance their mix of labor, resources, location, and infrastructure to optimize for the new AI economy. The three pillars are:
- AI Workforce: Labor pool with broad set of skills including AI researchers, AI implementors, and AI Literate workers adept with robotics and automation
- AI Infrastructure: Data sharing, algorithm exchange, labor mobility, and low-cost energy will be needed to support continued AI innovation
- AI Social Contract: Worker displacement will create significant resistance. Retraining, guaranteed income schemes, and other transition costs will be absorbed by leading countries adopting the AI Economy
As these changes percolate thru the worlds economy the balance of power will change slowly at first, picking up speed to an eventual tipping point. This final rapid change to the AI economy will create social and economic instability.
To manage this transformation we need to promote 3 priorities:
First Priority: AI Workforce
The demand for AI and robotics savvy workers already far outstrips the current supply. As the AI Economy ramps this impedance mismatch will become a gating factor for innovation. To avoid this we need to immediately make investments in adapting our workforce. China is already creating “AI Cities” and building this AI capability with full support of the Chinese government as Kai-Fu Lee outlines in his book “AI Superpowers: China, Silicon Valley and the New World Order” . Future leaders need to build this muscle with:
• Recruit and retain (in country) the best AI talent to universities to pursue research and support undergraduate AI education. Fund postgraduate research in AI with grants, challenges, and industry focused co-innovation programs
• Broaden science, engineering and mathematics undergraduate degree requirements to include core AI education and practical experience
• Fund robotics and mechatronic training programs technical and 2 year colleges coupled with tight collaboration with industry robotics providers
China is currently investing heavily in this area ranking #1 in AI research , #1 in AI patents, #1 in AI venture capital investment, #2 in the number of AI companies, and #2 in the largest AI talent pool per the “Center For a New American Security” paper Understanding China’s AI Strategy
Second Priority: AI Infrastructure
Recent news on AI has consistently focused on the privacy concerns and bias related to data gathering and usage. From personal data being gathered for training virtual assistants to AI gauging criminal intent of shoppers it is clear that any issues remain to be solved. China has taken an expedient approach to data gathering, and could use this access to leapfrog other global AI competitors.
1. National data sharing programs are critical to building up the training sets needed to build the next generation of AI — the west will need to resolve privacy and ethics concerns quickly
2. Open algorithm exchange as well as commercial algorithm exchange will be a critical success factor. IP laws will need to adapt to facilitate a flow of ideas and ML models
3. Labor mobility, and low-cost energy will be needed to support continued AI innovation
Already we are seeing businesses form just to close these data gaps, as detailed in this IEEE article: IEEE Spectrum Article but these models present much commercial friction
Third Priority: AI Social Contract:
Recent global political and economic conditions have heightened nationalism and immigration concerns around the world. This is in part due to the start of worker displacement from automation in the workplace. The leading AI Economics will address these social issues with an AI optimized social fabric or safety net to avoid political friction to change
Priority 1: Retraining, upskilling, and education: Much of the early impact of automation will be on low and medium skilled labor. Jobs that include delivery, sorting, driving, lifting, inspecting will be early targets for automation. New America’s work Shifting Gears on Automation shows that workers that live outside of urban areas face even steeper odds when faced with automation.
1. AI Optimized Vocational Education: Much of the focus on AI education has been on advanced math and science. The AI Social Contract will need to have vocational education that equips workers for emerging fields: Advanced Manufacturing, Mechatronics, servicing and repair of robots, etc. NGO and Community Colleges need funding to continue and grow programs like Seattle Goodwill’s Youth Aerospace Program.
2. Diversion of young workers from declining industries and job functions is critical. Retraining and job changing incentives to fully employed individuals in declining fields will save longer term government expenditures to retrain older and less flexible workers. Policy prescriptions to employers to hire late in career workers to fill current vacancies can help in this area as well as solving for some of the older workers employment needs after being made redundant.
3. Retraining and upskilling of middle aged workers presents the most challenge since they have many years left in the workforce, and can be less flexible to change industries, and frequently will pick training options that are related to their past employment making the risk of redundancy higher. A key success factor is receiving a career assessment to aid displaced workers in making retraining decisions(source: Atlantic).
Priority 2: Today just 48% of Americans support the idea of providing automation displaced workers with basic income, while nearly the same percentage of workers face the risk of job loss to robots (Source: Northeastern University /Gallup Survey, Sept 15th-October 10th 2017). It is reasonable to assume that public support will reach a plurality as automation impacts become more clear to workers. As public support builds, policy makers will need to consider the following as they implement these policies:
1. Guaranteed income schemes will need to be reserved for workers that are unable to reskill
2. Guaranteed income scheme participants should be prioritized as backfill for younger workers that are diverted
3. These schemes should be time limited to make the program more appealing to workers that are close to Social Security and Medicare eligibility
4. These programs will need to provide Medicaid eligibility as a stopgap until the worker and their family reaches standard Medicare eligibility
About the author:
Matt Vasey is a Senior Director responsible for AI Business and Corporate Development at Microsoft. He is focused on expanding the ecosystem of technology partners, standards bodies, and other innovation enablers that are required for the new generation of AI Applications, Services, and Systems that serve both individuals and businesses. Technology interests and expertise include Cognitive Workplace Automation, Robotics, Mixed Reality, Virtual Assistant Capabilities, Vision AI, Content Intelligence, and Edge/Fog AI.
In addition to his work at Microsoft, he serves as the Chairman of the OpenFog Consortium, Member of the Steering Committee of the Industrial Internet Consortium, Board member at the OPC Foundation, and on AI advisory boards at A3 Automation, Myplanet, and startups in the AI and IoT field.