Risk Management of AI urgently requiredJapan require to play a coordinating role for competition and　regulation

This article is written by Mr. Izumi Harada , President, Crisis & Risk Management Society of Japan and Executive Fellow, Institute for International Socio-Economic Studies（NEC Group）

Just one year ago I wrote an article “Light and Shadow of Artificial Intelligence” in this column. Since then AI (Artificial Intelligence) especially deep learning has been rapidly developed, improved and deployed widely in the society.                                                                                           

On the other hand if the development is more progressed, there would be more and more risks apparent and visible in the society.

In this issue I would like to summarize possible social risks caused by AI from two aspects.

 

Influence on military affairs and employment
Initially I would like to raise social risks caused by the development of AI itself. There are still pros and cons and divided argument on so-called singularity issue that Artificial General Intelligence (AGI) will surpass human intelligence.

There is emerging opinion that further development of AI should be stopped, but, the opinion might not be taken nor realized in short time under current situation where human potential itself has not yet been thoroughly studied and there is big argument that it would be too early to conclude whether AI surpass human intelligence and potential.

On the other hand the specialized AI (narrow AI) could be practically developed out of our reasonable control.

Especially AI weapons developed under AI military revolution could be out of our control as there is no global agreement to prohibit autonomous lethal weapons in UN meetings as it involves global conflicts of interest between nations.

Unemployment problem is also popularly and widely raised since many years ago. Unskilled and simple labor could be replaced and normal business skill and even many expert skills could be replaced by AI. From the experience in the past there would be such social unrest as re-employment of the individual spreading in the short term. For the society total there would be a big change in the re-distribution of wealth, productivity improvement, reduction of working hours, and relief for vulnerable groups. These issues, however, need time to be realized and in place.

Risks to Human rights and Privacy are also big issues. The more people demand safety and security, the more people become under tight control and surveillance. This is true for whatever social regimes are. Wether the liberty and freedom can be secured and the government authority can be under people’s proper control depends on the people’s suitable political power to evaluate and judge them and also on maintaining the freedom of speech. The maturity of democracy would be tested and questioned. People are required to maintain proper power not to yield to the intentional propaganda represented by fake news and fake videos.

Technologies are advancing too fast so that legal system can never be properly organized nor people can recognize its advancement. There is frequent argument how the responsibility should be defined for the accident of autonomous driving vehicle. It is indispensable that free diversified and comprehensive initiatives from integration of humanities and sciences are arranged to discuss and study what it should be.

 

Secondly there is social risk caused purely by technical problem of AI itself. Utilization of a large amount of data and information in the area of machine learning would give advantages and benefits to platformers represented by GAFA in the monopoly of data and induce unfair trading as a consequence.

Supervised learning in the machine learning sometimes presents error and deflection which would increase possible discrimination and prejudice. It is often pointed out also that there would be possible errors of algorithm there. There is, therefore, a risk that wrong judgement would cause adverse social effects.

On the other hand deep learning is having some risk of “black box” where the reasoning of judgement is not necessarily shown or given, and, we are obliged to accept whatever way or measure deep learning has selected to achieve or meet the given target.  Sometimes there will be problems caused because it is not possible for us to make proper prior checking of the issue. There will be further risks and concerns that data is altered or falsified by cyber attacks and even instruction itself to the systems might be altered.

 

Involvement of human being indispensable
AI is never versatile nor almighty. AI makes wrong judgement and there is a risk that the system runs out of our control. Therefore, it is indispensable that human being should be involved at the final stage in such cases as essential and influential judgement and decision.

Since Asilomar AI Principles back in the year 2017 to AI principles agreed by G20 members in June this year many efforts and attempts have been made by international organizations and enterprises to agree how to cope with problems possibly caused by technology development of AI. It is, however, regrettable that these tentative agreements have not been properly kept or observed.

As discussed above it is difficult or impossible to eliminate possible social risks given by AI. President Putin once commented that those who control AI would control the world. His comment can be verified simply by the fact that two nations, USA and China who are struggling for global supremacy are both desperate in developing AI.

Under the circumstance like this it is definitely required that our nation, Japan, takes initiative in developing and spreading human-centric AI which is safe and secure, and is not widening inequality while Japan is protecting freedom and human rights.

It is also required that Japan takes leadership in promoting global cooperation of mitigating and reducing the social risks caused by AI with proper balance between competition and regulation.

(Profile) Izumi Harada

Graduate School of Keio University
Currently in service as Executive Fellow for Institute for International Socio-Economic Studies, moved from NEC Research Institute and the Association for the Promotion of International Trade, Japan, and also in service currently as Part-time Lecturer of Waseda University.
63 years of age, born in Tokyo