The Challenges of Digitalization
Digitalisation will be for the 21st century what oil was for the 20th, and Data Privacy and Algorithms will be what environment was for Oil.
Attemps to develop a computer-based Artificial Intelligence are stemming back to the 40ies, and experienced several so-called winters, until Deep Learning could demonstrate benefits.
With the development of DeepLearning, and the ability of softwares to learn, and therefore adapt, there is a very serious concern that Artificial Intelligence could get out of human control, and develop side effects that would be adverse to human values. How exactly could this happen is presented in this fascinating and chilling video-clip: the 7 days of Artificial Intelligence!
While we are not rid of the debate on the degradation of the environment, as we are a long way from a decarbonization of the energy supply, but the debate on Digitalization is just beginning..., Gerd Leonard, a reference for these discussions, is highlighting these issues in his series of videos.
What is clear however, we need a public debate on the type of digital economy we want to see emerging, and this cannot wait any longer!
Digitalisation Strategy for Business
Digitalisation will affect most businesses in a way that is not easy to predict. However, the nature of the digital economy (marginal cost of acquiring a new customer dropping, when the benefit to the customer increases) makes it a Winner takes it all model. The sheer nature of the digital business makes highly risky, which means that agility and flexibility are the pre-requisites to move fast (....and break things) to get the necessary competitive advantage of a prime mover. More on this in this report on hard truth about the digitalisation by McKinsey: the pitfalls to prevent in forging digital strategies, and what makes the digital economy so different (first and quick mover, winner takes it all, etc...)!
Machine Learning supporting Decisions and Policies
If yo have 45 minutes, go to this link to watch on a video a great Primer about what is AI, Machine Learning and Deep Learning.
An introduction to Machine Learning is available here.
The potential of Machine Learning Artificial Intelligence (ML-AI) in supporting decisions and policies (see also this article) is getting clearer by the day, as more and more trials and experiments are done and assessed. At the present stage, ML does not necessarily have to replace humans for decisions, but may support better decisions, because they can override cognition bias from humans, and may of course process a history of data and facts in a way that humans will not be able to do. Positive experiences were done in health, legal, educational and other fields.
Still, several obstacles are paving the way of a broader acceptance of Machine Learning by the society, such as:
- It will depend to a large extend on perceived threat of this technology on privacy, fairness and privacy. Transparency could be the answer, but this will conflict with the typical secrecy surrounding the development of these algorithms.
- the access to Data will be another one: this is considered as the Gold of the 21st century, and it will probably not come for free. But then, who owns the data: the person originating the data, or the corporate who collect or buy them? And what if the algorithms can break the anonymization of the Data? And how can we deal with the explosion of Data from the IoT?
- and what about the type of prediction? For example, is it ethical to predict a probability of disease or death in order to optimize public health policies? And of course, predictive models to prevent suicide, claiming a potential success rate of 80%, would be desirable, but to a cost to privacy and personal freedom that is not yet clear.
among the 10 issues identified on AI by the WEF, the changing of work, the limits of AI, at least in the foreseable future, and ethical issues related to the type of decisions and their consequences that can be automated have already initiated a debate. This would result in a kind of MAGNA CHARTA of human rights in the digital world.
Algorithms, Machine Learning
A high level introduction to AI and Machine learning is available here. See also this video for a clear and crispy introduction to AIpresenting the views by the DARPA: some hype that deserves to be debunked, as presently a lot of engineering work lies behind any AI and ML success.
Whether and How Machine Learning is presenting a threat is a topic of discussion, as presented by I.J.Ito of MIT-Media Lab :
- on one side, machine learning algorithms can provide dramatic improvements and support in various fields such as diagnostics, driving, information scanning and processing, and management of systems,
- but it can mean also very serious challenges from the point of view that such algorithms can deviate, and deliver solutions that are not compatible with values we want to defend.
This topic is handled in detail in this report from the PEW-Research Insititute on the expected benefits and concerns of the spreading of Algorithms:
- we can expect better decisions through adequate support from expertise encapsulated into algorithms and connexion to large data sets, but also
- a rise of unemployment through job automation, and consequently a rise of inequalities,
- an increased risk of manipulation of people
- a lack of transparency, with the risk of encapsulating the bias of those (people and organizations) at the origin of the algorithms
all of which can be only corrected or at least mitigated through better, more common digital literacy, and a call for broader transparency of the rules and criteria underlying algorithms, possibly with auditing options.
For more on Deep Learning, taxonomy and parametrizations, this introduction provides a good insight. And a view of the complexity of the different models is presented in this review of the neural network zoo! that is also pointing to some of the key publications of this area.
As explaines in this article of the MIT review, a looming issue with deep learning for AI is the way algorithms are developed from large data set: the exact behaviour may escape engineers, with the result that odd outputs can be expected. And mathematical models turned into blckboxes, and taking critical decisions in finance, medicine, or military matters may be a problem!
AI: Myths and Facts
AI will disrupt our society, generate armies of unemployed, robots will control our destinies, it will aggravate inequalities, etc...
All of this is probable, but the truth is that we don't know when. Many experts such as Stephen King, or Elon Musk express concerns, and make a call to start to seriously anticipate the possible impacts of AI, and devise adequate models and policies that could not only mitigate, but direct these developments to a better direction.
The Institute Future of Life is holding a conversation on this, reflecting the concerns and the proposals of many of the key players of this field: misalignment of AI developments from our values is the main concern of many researchers.
But as the debate is just going on, many articles are available at the end of this page for more information and viewpoints.
As example, the debate on autonomous cars is certainly fueled with a lot of hypes, and while crivers support through automation is progressing well, there are reasons to be cautious about the coming of fully automated cares, with full safety and no steering wheels, as explained in this article.
AI: Hype or Hopes: Clearly, the enthusiasm about the breakthrough potentials of AI is mitigated by some, who thend by the way to be amongst the leaders of AI development: AI is about a system that plays a brillant chess move, while the room around is burning! This article of the MIT-Review is shedding some chill on such hype, due to the complexity of the ML design, and the fact that it is operating within a closely confined knowledge territory.
As claime on this article about the 7 deadly sins of AI prediction, AI, or even AGI (for Artificial General Intelligence) modern-day research is not doing well at all on either being general or supporting an independent entity with an ongoing existence. It mostly seems stuck on the same issues in reasoning and common sense that AI has had problems with for at least 50 years.
Additional updates on AI developments available at this webpage pointing to videos of the EMTech2017 event.
An Executive Guide to AI by McKinsey will provide highlights of the must-know on AI
The Internet of Things
A clear managerial overview of IoT, the Internet of Things, is provided in this Primer on IoT by Deloitte. IoT relies on several technology platforms, such as:
- Sensors, transforming an event or condition into an electronic signal
- Networks, communicating the electronic signals
- Standards, enabling the communication of various devices
- Augmented Intelligence, based on cognitive algorithms and big data
- Augmented Behavior: making prescriptions for action.
and Architecture Design, the art of combining the above. With the falling costs of IoT, experts expect that it will transform whole segments of businesses, such as transportation, health, logistics and distribution, etc...
Intel provides additional background info on IoT on their reference site here.
Part of the Hype on IoT, the Internet of things, is that designing and implementing such systems, by using standardized modules, can be quite cheap...., if we discount the necessary security. But having his home camera highjacked may not be a pleasant option. More on IoT security here. Beyond this, the same ethical and social concerns on Algorithms and Data Privacy apply here as well.
Responsible Algorithms Developments
A code of conduct for Responsible Digital Developments in Artificial Intelligence (the 23 ASILOMAR principles for AI developments) is proposed by the Future of Life Institute , highlighting elements of transparency, alignment to human values, privacy, shared benefits, not promoting unethical values, etc...
Additionally, the potential threat of AI on society is taken seriously, and a partnership is now in place to address the necessary framework to mitigate these risks. More details here.
Another concern is the design of Robots loaded with some pre-programed Emotional Intelligence that can anticipate mood-state: how should this be accepted by end-users, what type of intrusion into privacy or intimacy, and what could happen if this information would be shared? This video elaborates on this.
The concerns about malicious use of algorithms and AI was depicted in thisreport of the Future of Life Institute: reprogramming drones or robots to deliver lethal payload, blurring reality with ficticious information, diverting systems from intended purpose is becoming a real threat, requiring more awareness and preventive actions.
Digital Human Rights and Regulation
Skill Set for the Future Jobs
With more and more automation, and the emergence of AI and Smart Data, Experts expect that the future of the workplace will change dramatically. There is of course the risk that we will not have enough jobs to ensure at the present conditions a decent life to all citizen, hence the discussions on Basic Minimal Income. There is also the concerns that the skills needed will be in short supply, unless we reorient the education and the introduction of Youth into the workplace. As per the report The Future of Jobs from the WEF, the skill set required today is already focusing on complex problem solving, critical thinking, creativity.
Leaders in the digital economy will have also to be able to steer the integration of the data, predictive analytics and AI in their business developments. This will require an adequate understanding of the functioning and limitations of these technologies. For this a grasp on what deep-learning can deliver, its ethical and business limitations, the issues of bias, the transparency of the decision process. More on this discussion here!
Data and Health Care
The increase of Health expenses, now at 18% of GDP in USA, seems to get out of control, and is challenging some foundations of our solidarity mechanisms. The improvements in Technologies and Medications, together with demographic trends such as ageing, are obvious root-causes. The potential of digitalization and precision health prevention and medical care are raising expectations that it will bring a change of paradigms and of business models that could result in savings estimated anywhere between 15% to 30%, possibly with an improvement of the medical system performance. But of course,this is guesswork.
Presently, the developments of this digital health economy extend into several directions, such as:
- Apps, as interface for remote diagnosis
- wearable devices,
- access to huge DataBases,
- records of history of health conditions and life-styles,
- predictive analytics,
All this is expected to contribute to a more personalized medicine and preventive warnings about impacts to be expected from deficiencies in diet and lifestyles. Add to this the genomics and the match of personal DNA sequencing with records of inheritable diseases, and the power of Big Data is getting clearly visible. All large digital companies (Google, IBM, Apple, etc...) investigate this domain of health and digitalization. More on this in this article of the Economist
A discussion led by Politico on Data driving the future of Health Care raised the following points:
- Data collection, also linked to sensors (e.g. measuring glucose, etc.) should open new directions in preemptive/proactive health management
- Opportunities will emerge for fast diagnosis, and connected after-care
This should be the result of finding patterns by sharing large data from patients. To be effective, and overcome the reluctance of sharing data due to possible data security breaches exposing patients, some concerns must be addressed, such as:
- The control of data flow should benefit to the person generating the data
- Ideally, every citizen should be able to know what is done with his data, and possibly having a right to filter access.
- To be effective, data access must solve the contradiction of being safe secure and adequately accessible.
- And finally, Data should be a mean for a conversation about health, not an end.
But clearly, the acceptance from society of the Digital Health solutions will certainly depend on the responses that we will have to ethical algorithms, data privacy, ownership of Data, Integrity of solutions, security of systems, and much more.
The European GDPR - General Data Protection Regulation
Oil was the wealth of the 20th century, and Data will be the one of the 21st. What environment protection was for the oil economy, Data Protection will be for the Knowledge Economy. It is therefore of utmost importance that we put in place a proper legislation. This is what will take place with the coming legislation from the EU on Data Privacy, which will address:
- the consent of users for the processing of their data
- the right for deleting personal data, when not in conflict to public interest
- the principles of Privacy by Design, and Privacy by Default for connected objects,
- the protection of personal data related to health
- clarity on the use of data related to health insurance
the possibility to request inquiries in case of suspicion of privacy infringement by authorities or the police
- a kind of privacy shield for data transferred to USA.
Read more on Internet and Privacy Protection at this page of the EU, or get a deeper introduction to these topics in this brilliant interactive presentation.
And 2018 will be the year of the implementation of the EU - GDPR (General Data Protection Regulation), briefly summarized here, and discussed here: transparency on consent (especally on profiling), privacy, right to forgiveness, reinforcement policies, obligation to notify data-breaches are key-words behind this.
OECD views: Issues with digital transformation in the G20
The OECD has published a report that is adressing the potentially disruptive effect of the emerging digital economy on productivity, employment and well-being. It provides an analysis of the challenges of the Digital Economy through numerous data and graphs, and offers as well recommendations.
- building and reinforcing trust
- improving access to digital technologies and services, especially for SME
- develop national access to broad-bands
- addressing issues with Data pricing, privacy, ownership and IP rights
- developing standards for inter-operability
- continuously adapting the regulation, by removing obsolete regulations, and ensuring competitiveness and Innovation
- Developing Digital Security Strategies, protecting core assets, and including SME as well.
- Accelerating the transformation of the education to generate the necessary skills
- Reinforcing the consumer rights
Blockchains, the electronic transactions, with encrypted validation by a distributed ledger, is expected to simplify and secure interface without necessary intermediaries. This is expected to be a dramatic accelerator for the digital economy. This article of the WEF provides a good introduction to the technology and its potentials.
Spend 10 minutes watching this video for a simple and clear explanation of the distributed ledgers of the blockchain, or spend 20 minutes with this video for a more detailed explanation of the process and get access to a demo of the concept as well.
Blockchain is further discussed in The WIRED Guide to the Blockchain
Still, the discussion about the energy consumption of each transaction, and the time needed for the miners to complete the check indicates the limitations of the systems, and the necessary further developments to get a sustainable solution for distributed ledgers.
Beyond it, this site is the reference on blockchains transactions.
Whatever the obstacles, Blockchain has reached a high level of maturity and acceptance, and applications in the Fintech and especially health and insurance with the potential of reducing transaction costs and improving data integrity are now emerging. More on this in this communication of Bankrate
Digitalization, the Future of Jobs, and Taxation
The automation of the working place through A.I. (Artificial Intelligence) might result in the vanishing of millions of jobs. For the E.U., some estimate the number of jobs at risk at about 8 to 10%, or around 30 million, while the gaps for jobs in A.I., robotics and Algorithm development is estimated at .8 to 1 million jobs. Clearly this would mean a dramatic increase of unemployment, and of social inequalities. While the value of such estimates depends from the quality of the crystal ball, the impact is large enough to initiate some thoughts on the social benefit of the digitalization ad the future of work.
A taxation on robotization (supported by Bill Gates) is such an idea that is often discussed but implies a clarification of the definition of a robot: is this a machine or an A.I. algorithm, is it partly replacing a human, or just improving its productivity, or fully replacing it? In each case, the scope of the taxation would be different, with obvious issues in its implementation. This is why other options are discussed, such a taxation on dividends to finance a Basic Universal Income, with 2 points:
- a discussion on taxation has to be enlarged to cover the type of taxation we want: taxing labor? Capital? or scarce resources (see also Ex-Tax)? This will require a fundamental debate.
- and a discussion on Universal Basic Income. Do we want, or can afford a Universal Basic Income? When we define work as an activity to provide the means for subsistence, may be. If we define work as a social interaction for integration, may be not. And is this the best way to allocate scarce resources, or should we more focus on retraining and continuous education?
Digitalisation of the Work Place
Now that AI will be able to automate many redundant or predictable activities, the discussion is turning on the future of the work place, and concerns whether enough work will be available in the future in order to provide decent jobs to everybody and ensure enough equity to maintain a social balance.
Several analysis, e.g. by McKinsey, indicate however that similar to the Luddites concerns of the XIX century, what can be expected is a dramatic evolution of the work place, much more than the disapearance at a large scale of jobs. This will almost certainly lead to a major enhancement of productivity (report here), as routine tasks will be performed with more Know-How at hand, and workers in industry or services will be able to focus more on exceptions.
Getting employee involvment and commitment from management will be all the more important in the digital economy!
Digitalisation and Environment
Huge quantities of rare earth, metals, and others are used for computers and electronic devices of all types. Once used, these materials are discarded, and generate a toxic cocktail of waste. Planned obsolescence, not repairable devices, and little concerns ofr a circular economy in this sector contribute to this crisis. GreenPeace has proposed a path to greener electronics
As per a UNU report, with over 44 millions metric tons of e-waste each year, and to a large extend not traceable, the ICT industry is turning to a major environment crisis. Many interventions are possible, from increasing traceability, to more recycling locations, but there is an urgency to tackle this issue, which can only become worse.
One way to prevent this mounting heap of waste would be to use electronics as long as its functionalities are acceptable, repair it, and then recycle it. This assumes to select devices that are more easily reparable, and not all are equal here, as per this report from GreenPeace again.
Investing in AI
According to POLITICO, 4 countries makes the bulk in nvesting in AI: USA, China, Israel and UK. This is also reflected in the number of start-ups relevant to this field.
- UK is presently the acting AI leader, but this is shadowed by political clouds, resulting in some relocations.
- France has clear plans to boost the AI sector, and is benefitting from the BREXIT, and the quality of the top universities. However, it wll have to tame the bureaucratic devils to grow big.
- Germany has still to flex its muscles behind, and also turn a research competence into an innovation machine.
AI patent landscape
The explosion of AI based developments was obviously prededed by a surge of scientific publications, and of patents.
The WIPO (World Intellectual Property Office) has published a review of the patent landscape in AI that is providing also a view of the competitive field that is present or emerging.
- Patent filed are exploding, and this is maily the business of USA and China
- in AI, machine learning, and especially Deep Learning, is the most patented technology
- Robotics and Control systems are the fastest growing application areas
- Transportation is the fastest growing area in industry
- Microsoft, Toshiba, Samsung and NEC are the top 5 applicants.