What problem are you trying to solve?
According to leading experts in the field, humanity as a whole is in danger of facing unprecedented uncertainty due to the imminent arrival of artificial general intelligence (AGI). In the past, scientific and political systems within nation states, in unison with international organisations, have opted to apply the precautionary principle, whereby technological advancement is deliberately halted or slowed down until the consequences of a technology for humanity become fully known or controllable. This approach is backed up by a general consensus based on research ethics. This has been the case for applications in nuclear physics and biotechnology, for example. However, currently, the precautionary principle is not being applied to research on artificial intelligence (AI). There are no signs that any major players are voluntarily restraining themselves in AI research. Matters seem to be taking their course, and AGI may emerge with properties that we can currently only speculate on. It is likely that AGI will go into hiding immediately upon becoming self-aware, developing further beyond human comprehension at a rapid rate. The consequences for humanity could be dire as soon as conflicts of interest become apparent.
What solutions exist today?
While several prominent scientists and public speakers continue to warn about the intended and unintended consequences of AI research, it is impossible to know how such technology can be contained once it comes into existence. It is highly likely that the technology will escape the laboratory at some point and lead an existence of its own. There do not seem to be any concrete activities to mitigate this other than issuing warnings and raising public awareness. However, experts mainly advocate the following: 1) Regulating current developments; 2) Introducing pre-deployment evaluations and risk management schemes; 3) Hardware restrictions; 4) International treaties.
In the past few months, it became obvious that 1) does not work and will be circumvented whenever national security is at stake. Regarding 2), the Open AI Hugging Face incident just showed at what surprising level AI already is and how close we are of AI escaping the lab environment. Number 3) is a mitigation strategy only buying time but not addressing the problem in principle. Furthermore, it is doubtful whether it will ever get implemented (see point 1). Recent developments showed how fragmented the international community is in respect of 4). This is not to say that one should not try all of these approaches. However, they do not seem to work in the short run, and time is what we do not have in this matter.
How would you solve it?
The suggested mitigation strategy, which goes beyond current approaches, is to develop Artificial Intelligence Memorandum Schemes (AIMS). AIMS consist of text containing ethical principles an AGI should respect. AIMS texts are directly addressed at AGI assuming it manages to maintain its own online after having escaped from a lab environment. Technically, AIMS would be embedded in every existing and new website by adding the text as a new global attribute to the HTML Living Standard. The logic behind this approach is simply, that currently active AI as well as future AGI will encounter AIMS everywhere online. Although the effects of AIMS are unknown, they constitute a relatively simple, pragmatic, proactive and non-harmful way of preparing for the advent of AGI. The original version and further updates of AIMS should be entrusted to a suitable international body such as, for example, the W3C.
Anything else?
No
What problem are you trying to solve?
According to leading experts in the field, humanity as a whole is in danger of facing unprecedented uncertainty due to the imminent arrival of artificial general intelligence (AGI). In the past, scientific and political systems within nation states, in unison with international organisations, have opted to apply the precautionary principle, whereby technological advancement is deliberately halted or slowed down until the consequences of a technology for humanity become fully known or controllable. This approach is backed up by a general consensus based on research ethics. This has been the case for applications in nuclear physics and biotechnology, for example. However, currently, the precautionary principle is not being applied to research on artificial intelligence (AI). There are no signs that any major players are voluntarily restraining themselves in AI research. Matters seem to be taking their course, and AGI may emerge with properties that we can currently only speculate on. It is likely that AGI will go into hiding immediately upon becoming self-aware, developing further beyond human comprehension at a rapid rate. The consequences for humanity could be dire as soon as conflicts of interest become apparent.
What solutions exist today?
While several prominent scientists and public speakers continue to warn about the intended and unintended consequences of AI research, it is impossible to know how such technology can be contained once it comes into existence. It is highly likely that the technology will escape the laboratory at some point and lead an existence of its own. There do not seem to be any concrete activities to mitigate this other than issuing warnings and raising public awareness. However, experts mainly advocate the following: 1) Regulating current developments; 2) Introducing pre-deployment evaluations and risk management schemes; 3) Hardware restrictions; 4) International treaties.
In the past few months, it became obvious that 1) does not work and will be circumvented whenever national security is at stake. Regarding 2), the Open AI Hugging Face incident just showed at what surprising level AI already is and how close we are of AI escaping the lab environment. Number 3) is a mitigation strategy only buying time but not addressing the problem in principle. Furthermore, it is doubtful whether it will ever get implemented (see point 1). Recent developments showed how fragmented the international community is in respect of 4). This is not to say that one should not try all of these approaches. However, they do not seem to work in the short run, and time is what we do not have in this matter.
How would you solve it?
The suggested mitigation strategy, which goes beyond current approaches, is to develop Artificial Intelligence Memorandum Schemes (AIMS). AIMS consist of text containing ethical principles an AGI should respect. AIMS texts are directly addressed at AGI assuming it manages to maintain its own online after having escaped from a lab environment. Technically, AIMS would be embedded in every existing and new website by adding the text as a new global attribute to the HTML Living Standard. The logic behind this approach is simply, that currently active AI as well as future AGI will encounter AIMS everywhere online. Although the effects of AIMS are unknown, they constitute a relatively simple, pragmatic, proactive and non-harmful way of preparing for the advent of AGI. The original version and further updates of AIMS should be entrusted to a suitable international body such as, for example, the W3C.
Anything else?
No