Containing Artificial Intelligence

 Artificial Intelligence is on the rise, whether you search the internet, ask questions to the customer service online or use the driving tools of a modern car, we are already using AI in many applications. The solutions being build are focussing on automating human skills such as programming, translating, improving texts, making images or even videos and music. The solutions bear, as with any technology, risks in them. Programs might not be tested well, internally not understood, making them vulnerable for hacking or errors, translations might play down the incentive to learn a foreign language, text improving on learning spelling and grammar result in a similar downplay of our own skills. Image generation might reduce our joy to go out in the real world and make our own images or videos. Besides this, all these techniques blur the interaction between the author and the reader as this relationship is compromised by the tools being used. Many ‘readers’ also use summary AI tools thus further avoiding the relationship between the writer and the reader. Also, why should I read a text generated by a computer, as I am interested in the opinion or experience of a fellow human being when opening a report or book and starting to read. Reading is a human-to-human conversation with a delay. Grammatical and spelling errors are part of the relation, as it gives the reader insight in the skills of the writer.

By not displaying who we are and what our skills are, we make reading less interesting. One can state that the introduction of the calculator gives some insight in what will happen when our text tools automate our writing and spelling skills. The calculator greatly reduced our need to check the bill, and most people do not do this task anymore, also because they are not able to make the summarization themselves. Of course, we gained some time by not checking bills anymore and, also some disputes over it are past experiences. But the consequence is that most people do not understand anymore the difference between a billion and a million and do not understand what a percent increase in tax implies for their own finances. Inflation but also investments in assets without an intrinsic value such as crypto valuta do not display a high skill of calculus literacy. I do not think research on this has been done, nor would it be feasible, as the people who functioned well without a calculator, as a testing group, are old or death. We can state that calculus was the first match the computer won from us, much later followed by chess, based on the same principles. Similar it would be of interest to learn whether the defeat of humans in chess lead to less interest in the game. What I am pointing at is that we have enough recent experience of introducing compute into our society to understand the benefits and risks. With the new tools coming to us we seem to be reluctant to finetune the tools based on our previous experience. By discussing the calculator and the chess computer, I remained in clear applications upon which elderly people can still tale some of their past experiences, especially that society worked fine without them.

When it comes to social media, that effectively automated our skills to relate to strangers, in the train, a bar or just asking directions in an unknown area, it is now acknowledged that by just following marked mechanisms, the risks outweigh the advantages. Algorithms that favor inciting statements above appeasing ones created angry and discontent humans, not verifying anymore whether the statement reflects reality, remaining in their anger. While the provider of the algorithm just optimizes for attention in order display some more publicity. To avoid prosperous humans remaining angry, somehow this mechanism must be regulated. We cannot censor the statements, as now is very often done by authoritarian states or even left-wing democrats, what is to be regulated is the algorithm. A human cannot deal with too much incitement and attention seeking is not perse a good characteristic that should be automated. Regulation of algorithms that provoke negative emotions is a much more complex task for a lawmaker, especially if he or she just lost his skills in calculus. As it now seems lawmakers are just trying to stop the game, at least for children, as the industry, following competition and market mechanisms, proved incapable of making social media that makes one feel better and more capable to act in stead off feeling worse and unpowerful.

With the new AI tools coming at us, these three experiences were just an introductory game. We risk of loosing our spelling and grammar skills, our incentives to learn foreign languages and the brain gymnastics and empathy provoked by this. We will constantly be confused whether something, images, movies, music is human made and therefore interesting. We will start to relate to computers, with whom we cannot make offspring, thus finally creating a profound frustration in our functioning. Since any biologic creature is driven by creating offspring, and when we will find out that all our energy in relating to a chatbot, robot or whatever, results in nothing, a deep frustration and anger will take hold or our emotions.

I listed some arguments to regulate AI based applications that currently are released upon society. The arguments to make regulations on the dreams on making AGI and intelligent robots are even more complex, as we do not have real life experience with them, and therefore the balance between benefit and risk is unknown. On the risks we have already one clear mechanism, the need for energy of these tools should not be a burden upon ‘normal’ consumers, nor may it contribute to further damage to the composition of the atmosphere and acidity of the ocean. This rule can be regulated and that is not done, although most tech societies tend to invest in renewables and nuclear facilities, the repercussions on the electricity price for normal consumers on the need for more power are unclear and should be regulated.

I therefore propose the following regulations:

• ·         AI intelligence generation or training should be done only using renewables and only during the moment that renewables generate energy, the powerplant being used should be build exclusively for the facility and all costs, also on new or adapted infrastructure are for the tech company. Training is to be done on remote locations without fertile land and need for cooling.
• ·         AI applications cannot be connected to the internet, they can train using a local copy of the (open) internet or other sources and a protocol to communicate with them through the internet can be developed using a so-called stateless connection.
• ·         Any AI generated content is to be labelled as such, and social media and internet applications should not be allowed to use it. The internet may only reflect moments of the real world, an AI net can be made using another protocol for ‘enjoying’ the AI generated content.
• ·         Social media algorithms should be adapted to display a limited number of inciting statements or imagery to a user. The algorithms should provoke the user to be creative and active in the real world. This can be done by restricting the amount of publicity social media may display to a user per day or another time-interval. Age restriction should apply but to be mandatory controlled by parents or caretakers with defaults below 1 hour per day.
• ·         Robots, functioning in public space, may not imitate the human body, also they should serve a clear and limited purpose, mainly for enabling less abled humans to function and picking fruit or litter. Robots able to perform aggressive acts should be banned from public space. Robots, like AI cannot connect to the internet or AI net.
• ·         Chatbots may not imitate human emotions, alike social media they should provoke that people return to make relations with other human beings or animals.
• ·         There should be several days per year when all computers are turned of, apart from services needed for hospitals or security. The health and security services should function in a clear separated technical environment to avoid any disruption provoked by the tools society entertains with.