Now we have heard both strands of the AI debate. Stephen Hawking, Elon Musk and Bill Gates have warned that Artificial Intelligence is the greatest threat to humanity and now, in his book "Novacene", James Lovelock predicts that AI's will save us.
AIs are used to spot trends in data to direct advertising, recruitment etc., to replace workers, as guidance devices including guidance for military systems, as higher accuracy pattern recognition devices and by States such as China to monitor and prevent civil disobedience. As time goes by their roles will increase.
At present AIs are tools employed by corporations and public sector organisations to increase the efficiency and speed of their operations. AIs are an extension of the intentions of their owners. There is no rule in nature that describes how to create the perfect human society so where AIs are used for government and corporate functions they are always implementing human desires.
The future development of AIs will be consistent with the past development of human society. There will be two strands of development. The first strand will be to increase the profits and penetration of society by large corporations. The second strand will be to increase the control and power of the state. In the past our society has been fortunate in mounting resistance to both of these strands but AIs give big business and big government incredible power. As usual, it is not the weapons that we should fear but the tyrants who wield them.
It is conceivable that Corporate AIs can be controlled with privacy and other legislation but State AIs are another problem entirely. The UK can control the use of AIs by the State but it will rapidly be confronted by countries such as China that have turned their entire state and production machinery over to AIs. This means that an AI arms race is inevitable.
In the Cold War the West won because it had the greater production capacity. The communist tyrannies could not compete with the breadth and depth of Western development. Thus are wars won but AIs subtly change this dynamic. AIs may provide a method by which the tyrants can achieve production that is as great or greater than that of the free world whilst simultaneously placing their own populations in chains.
One of the big dangers or benefits of AIs is that they may, in the next few decades, be able to out-think human beings. Some people see this as a terrible threat where AIs seize control from free humans and others as a huge benefit where AIs seize control from wayward humans. In either case the AIs would be running the world. There is no blueprint in nature for the perfect machine or human society so AIs that run the world would be working on the basis of a constitution, a set of rules, placed inside them by their creators. A Chinese AI might ruthlessly impose Chinese Capitalism on the world and an American AI might impose the American Constitution on the planet. If some idiot programmed military AIs with nothing but search, destroy and survive they might kill us all.
Even if we do oppose AIs we are bound to develop them or be swept away by tyranny but if we develop them we are likely to hand power to Corporates and the State and so to tyranny. It is difficult to see a way out of this conundrum but here are some suggestions.
Controlling AIs and Big Data
The way to prevent multinationals from using AIs to undermine local production is to ensure that enough production occurs locally to maintain wage rates and working conditions. UK AIs should run sufficient UK industry to ensure that there is a large enough profit and tax base to maintain the wealth of the people.
The UK government must be aware that foreign powers will be developing AIs for espionage and military use and develop counter measures. Unfortunately there is a need for a national firewall because without it foreign powers, whether government or corporate, will be able to simply walk into our information space and do as they wish.
One way in which AIs can be controlled is by introducing laws governing the value of data. If a corporation or government sells or even distributes data that has not been contributed specifically for further sale then it must pay the people whose data is sold at least 80% of its nominal value and the original copyright must remain with the data contributor. Large commercial and most government data holdings should pay an annual rental to those whose derived data is held. There should be few exceptions and treating personal data as a perpetual asset will ensure that we know who has our derived data.
Systems in which AIs penetrate the home with devices such as Alexa and the Internets of Things need very careful control. Custodians of data acquired from such products must have procedures for the total deletion of that data where it travels off-site and draconian penalties for leaking or sharing that data. Only data stored on-site should be available to the police or other government body and then only after a court order.
The biggest threat currently posed by AIs in the UK is that they can manipulate social media for advertising, PR and political purposes. This is actually a fault in UK legislation which has failed to properly classify Internet Corporations: Social media companies and ISPs are publishers and printers with all the responsibilities that entails.
Utilities systems, whether they run grids or towns, should be prevented from dis-aggregating data except by court order. There should be no monitoring of individuals except with the permission of the courts.
AIs must be designed to provide the justification for their decisions so that as AIs become more competent there can always be human inspection of the grounds for AI decision taking whether that is medical diagnosis, driving a car, employing staff or even the sale of insurance.
I would be optimistic about AIs if I were optimistic about the ability of humans to analyse and control threats in a sensible manner. It is only military AIs that might produce an existential threat for humanity so, if we do nothing else, we must regulate these devices.
AIs are used to spot trends in data to direct advertising, recruitment etc., to replace workers, as guidance devices including guidance for military systems, as higher accuracy pattern recognition devices and by States such as China to monitor and prevent civil disobedience. As time goes by their roles will increase.
At present AIs are tools employed by corporations and public sector organisations to increase the efficiency and speed of their operations. AIs are an extension of the intentions of their owners. There is no rule in nature that describes how to create the perfect human society so where AIs are used for government and corporate functions they are always implementing human desires.
The future development of AIs will be consistent with the past development of human society. There will be two strands of development. The first strand will be to increase the profits and penetration of society by large corporations. The second strand will be to increase the control and power of the state. In the past our society has been fortunate in mounting resistance to both of these strands but AIs give big business and big government incredible power. As usual, it is not the weapons that we should fear but the tyrants who wield them.
It is conceivable that Corporate AIs can be controlled with privacy and other legislation but State AIs are another problem entirely. The UK can control the use of AIs by the State but it will rapidly be confronted by countries such as China that have turned their entire state and production machinery over to AIs. This means that an AI arms race is inevitable.
In the Cold War the West won because it had the greater production capacity. The communist tyrannies could not compete with the breadth and depth of Western development. Thus are wars won but AIs subtly change this dynamic. AIs may provide a method by which the tyrants can achieve production that is as great or greater than that of the free world whilst simultaneously placing their own populations in chains.
One of the big dangers or benefits of AIs is that they may, in the next few decades, be able to out-think human beings. Some people see this as a terrible threat where AIs seize control from free humans and others as a huge benefit where AIs seize control from wayward humans. In either case the AIs would be running the world. There is no blueprint in nature for the perfect machine or human society so AIs that run the world would be working on the basis of a constitution, a set of rules, placed inside them by their creators. A Chinese AI might ruthlessly impose Chinese Capitalism on the world and an American AI might impose the American Constitution on the planet. If some idiot programmed military AIs with nothing but search, destroy and survive they might kill us all.
Even if we do oppose AIs we are bound to develop them or be swept away by tyranny but if we develop them we are likely to hand power to Corporates and the State and so to tyranny. It is difficult to see a way out of this conundrum but here are some suggestions.
Controlling AIs and Big Data
The way to prevent multinationals from using AIs to undermine local production is to ensure that enough production occurs locally to maintain wage rates and working conditions. UK AIs should run sufficient UK industry to ensure that there is a large enough profit and tax base to maintain the wealth of the people.
The UK government must be aware that foreign powers will be developing AIs for espionage and military use and develop counter measures. Unfortunately there is a need for a national firewall because without it foreign powers, whether government or corporate, will be able to simply walk into our information space and do as they wish.
One way in which AIs can be controlled is by introducing laws governing the value of data. If a corporation or government sells or even distributes data that has not been contributed specifically for further sale then it must pay the people whose data is sold at least 80% of its nominal value and the original copyright must remain with the data contributor. Large commercial and most government data holdings should pay an annual rental to those whose derived data is held. There should be few exceptions and treating personal data as a perpetual asset will ensure that we know who has our derived data.
Systems in which AIs penetrate the home with devices such as Alexa and the Internets of Things need very careful control. Custodians of data acquired from such products must have procedures for the total deletion of that data where it travels off-site and draconian penalties for leaking or sharing that data. Only data stored on-site should be available to the police or other government body and then only after a court order.
The biggest threat currently posed by AIs in the UK is that they can manipulate social media for advertising, PR and political purposes. This is actually a fault in UK legislation which has failed to properly classify Internet Corporations: Social media companies and ISPs are publishers and printers with all the responsibilities that entails.
Utilities systems, whether they run grids or towns, should be prevented from dis-aggregating data except by court order. There should be no monitoring of individuals except with the permission of the courts.
AIs must be designed to provide the justification for their decisions so that as AIs become more competent there can always be human inspection of the grounds for AI decision taking whether that is medical diagnosis, driving a car, employing staff or even the sale of insurance.
I would be optimistic about AIs if I were optimistic about the ability of humans to analyse and control threats in a sensible manner. It is only military AIs that might produce an existential threat for humanity so, if we do nothing else, we must regulate these devices.
Comments