The I-7 Experience
Opinion Piece for the OECD Forum Network
As Chair of the first meeting (Turin Sept 25, 2017) of the I-7 Advisory Board, I had the opportunity to discuss the extraordinary potential that AI and BIg Data have to catalyze economic value, help social progress and improve human performance.
The Italian G7 Presidency established the “Strategic Advisory Board to G7 Leaders on People-Centered Innovation” (I-7) (Taormina Communiqué par. 37) to provide qualified opinions on specific technological innovations which Governments might consider when designing and shaping policies. The format of the Advisory Board recommended for each G7 country and the EU, the formal involvement of a Focal Point and up to 5 innovators.
Each group produced a brief paper on each of the topics that were discussed in three parallel panels.
Two panels discussed Big Data and AI, and their potential to inform and transform governments.
Parallel Session 1. Artificial Intelligence (AI). How can AI help governments make better decisions and deliver policies and services more effectively? AI represents one of the most intriguing and promising fields of innovation and has the potential to dramatically improve the way public and private services are delivered. AI could have plenty of applications in government. As an example – applicable to e-government in particular – the private sector is already delivering “chatbots” and using NLU (Natural Language Understanding) to automate and improve services for customers. But much more can be shaped by AI: from traffic management and self-driving cars to security and emergency management. Furthermore, AI technologies may not be governable by the existing regulatory frameworks, presenting a unique policy challenge compared to prior innovations. In this session, innovators will discuss and generate ideas about how governments can take advantage of AI.
Parallel Session 2. Big Data: from regulation to active management. How can a more proactive approach to Big Data lead to smarter countries? Big Data represents the “fuel” of tomorrow’s production and a critical feed on information to AI-based services. In this session, innovators will be asked to discuss how a more proactive approach to Big Data could help tackle social and productive challenges in new ways, magnifying the effects on people’s well-being and businesses’ competitiveness. Government Big Data is usually stored in silos, and we need to find leverage points to convince the gatekeepers of the silos to collaborate. For example, should governments build national Data Analytics Frameworks as a necessary step to building “Government as an API”? Furthermore, once again, Big Data may not be governable by the existing privacy regulatory frameworks, presenting a unique policy challenge.
A third panel debated a new possible policy framework to address the changes and challenges in the job market.
Parallel Session 3. The changing nature of society: the future of work. How could innovation help deal with upcoming social and demographic changes? The present narrative is much too focused on the negative effects of technology on jobs and employment. As ever, innovation is destroying old jobs while creating new ones. In this session, innovators will highlight how new technologies are shaping the future world of labor, and identify those skills which will be more in demand, thus helping workers reap the benefits of change. (Agenda).
In coherence with its mandate, the I-7 discussed the current potential of technological advances, focused its attention on ethical issues, addressed the technology impact on skills and jobs, and identified the possible measures to reduce failure and mitigate potential downsides. Also, the I-7 drew the G7 leaders’ attention to the gap between the current potential of technological advances and the actual adoption from institutional organizations, since the digital transformation present and future challenges need transition policy and adequate public systems which are able to adjust and accompany the social and economic changes. All experts unanimously recommended that public institutions should dispel unwarranted fears about AI, and public opinion must be objectively informed about the challenges and the prospects for everyday life.
The conclusion of the meeting were collected in the Chair’s Summary.
The Summary recognized the importance for governments to become model users to tap the unprecedented opportunities offered by digital innovation to create new services with a positive impact not only on their citizens’ well-being and the job market but also on the innovation ecosystem as a whole.
Here below some of the Summary main messages.
The discussion was focused on the following 4 themes: training and education; technology; ethics and society; and jobs and income.
“TRAINING AND EDUCATION” KEY POINTS
- High Impact awareness, training and education programs about exponential trends in technology and other fields for the whole of society, including politicians, policy makers and business leaders, faith leaders, opinion leaders, teachers and professors, journalists and the media. The need to acquire fundamental and new skills and competences, such as creativity and the capacity for solving tough, complex problems, logical reasoning, or coding, become imperative for all sectors in an advanced economy. Encourage basic technology and data literacy education in primary and secondary schools.
- Educational systems should support ‘lifelong learning’ and retraining of the workforce.
- Alongside the above, education should also include the key human attributes of creativity, originality, reciprocity, responsiveness, empathy and adaptability.
“TECHNOLOGY” KEY POINTS
- APIs are the nervous system of the 21st century. Silos that prevent data integration are greatly reducing the power of high-variety data to deliver deep insights. To overcome these, we encourage collaboration on common open APIs in sectors of shared importance and promote access to and exchange of data.
- The development and maintenance of trust among citizens and businesses is critically important for the effective implementation of AI solutions. The corollary of trust is trustworthiness. All of those that participate in the development of such systems should recognize the importance of behaving in a trustworthy fashion. This includes consistent observance of best practices to maintain security and privacy. AI technology itself can be used to enhance security and privacy.
- G7 countries should acknowledge the importance of innovative applications of AI for government work and the opportunities for collaborations between them.
- Governments around the world are implementing machine learning and AI to develop and deliver services. We encourage sharing amongst G7 Countries of best practices and lessons learnt from these examples.
- G7 Countries encourage research in key scientific, technological and societal fields to provide an environment conducive to the emergence of innovation and development of a digital economy. Advancement in security technology is a key priority.
“ETHICS AND SOCIETY” KEY POINTS
- The transformational nature of Big Data will require continuous adaptation of all key actors involved – governmental bodies, businesses, citizens – in their multiple roles of data generators, data analysts, and end users, and a different conversation on the concepts of digital privacy, data ownerships, and digital security, and accountability in the sharing and use of the data.
- Policy makers and providers of AI based services should consider how they can be delivered in the most inclusive fashion.
“JOBS AND INCOME” KEY POINTS
- The new technology platforms have created new working environments and the so called ‘gig’ economy. Governments need to consider how to recognize and encourage these new platforms and to consider the terms and conditions under which this new workforce operates.
- It is important to monitor and measure what kinds of jobs are created, changed and lost. This will help businesses and policymakers to constantly update the attributes needed for modern workplaces.
- In the era of increasing automation, we need to find better ways of measuring the ‘gig’ economy, non-paid labour, volunteering, to improve social inclusion and civic engagement.
To test the potentials of AI, the I-7 decided to experiment with an AI Generated Chair’s Summary. From a technical point of view, the model contains over 1 Billion free parameters and it is trained on an HPC cluster with a total of 16 GPUs (Graphical Processing Units). It was used a customised version of the open source software Tensor2Tensor (https://github.com/tensorflow/tensor2tensor) to implement the model and was trained for over 60 thousand iterations on a recent dump of the English Wikipedia.
For each page the text was extracted and truncated to the length of 2048 words. Then a selection of random pages were bundled in batches of 8192 words and fed to the neural network. Synchronous training was used. The model, with no prior knowledge of the English language nor of any other concepts, receives sentences and tries to predict the next word in the sentence; it merely tries to learn to guess the next word in the sentence by observing millions of sentences. The model was trained for over a week. At this point the model was already capable of producing original generic sentences in English. To further narrow down the learning space the training over was extended to a selected list of 20 books coming from the broad search “artificial intelligence” done in Amazon as of September 2017, as this corpora reflects a good approximation of what a group of experts can achieve to learn for the task of answering the questions posed during the I7.
The results were quite amazing! See for yourself.
In the future artificial intelligence will replace human work — and may represent a potential long-term future in AIs. If worker roles can be perfectly changed without anymore human intervention, we'll be in the AI winter of madness, but nature does appear to require a digital lifesaving regime for a biological human being, and this may allow the prospect of human workers to continue exist indefinitely. Artificial intelligence can help government deliver policies and services more effectively than with human capabilities and abilities, that also encompass deterrence and analysis. The worst scenario is probably excessive or inappropriate programming, which is why the government have to work to create a perfect deterrent. Future workers will use artificial intelligence to power existing technologies, or to eliminate redundant pre-defined factors such as genetic bias and simple algorithms. What we will call "general artificial intelligence"? We might propose "a superintelligence whose well organized function is not conceivable". As well we might point out that a suitably engineered intelligence might look human even if it's not very intelligent. These hypothetical "smart" intelligent beings that will be selected (or at least worthy) will already supersede human intelligence.
For information on the I-7’s meeting https://teamdigitale.governo.it/en/i7.html
 NB The statements in bold report the incipit of the sampling; there was some human post editing to fix basic grammar nuances.