"Deliberative structure" and AI security research:
Differences research security AI, are:
1. Effect of time trouble, - acceleration of (self) -development of AI.
2. Localization of the threat - in the political application of AI.
3. Using democratic AI for security studies of AI.
4. The role of the geographical factor in the safety studies of AI
The effect of time trouble:
No time for traditional - Sequential approach:
- first – Research of the risks of AI,
- then - Research on the safety of AI,
- then - Work with (inter) -governmental bodies for work with AI developers,
- then - Development of security for AI,
- then - Hope for the hierarchical implementation of development:
- Hope for implementation in (foreign) AI development centers,
- Hope for training grassroots (foreign) users.
Parallel research, development and public implementation of security AI:
1. The security approach of AI must be Parallel.
That is, the above steps should be carried out simultaneously.
2. It is necessary to work not only with government agencies and AI developers, but directly with people.
The "Deliberative structure" is a tool for this.
Political orientation of security research AI:
A public organization is needed to carry out all of the above works also because:
The threat is not the AI itself, but the political use of AI.
Therefore, the safety studies of AI:
1. Should focus on the threat from the political use of AI,
2. Must be widely public.
Effects of using AI for AI security studies:
Yes, the average current quality of work may be low.
But the end result should be higher due to:
1. Time trouble.
2. The use of Democratic AI-tools, dramatically increasing the efficiency of 1-rank work, doing a "convolution" of the iterations of the above stages - in real time.
3. The dissemination of the achievements of the best achievements should be faster and more effective.
4. This, in turn, should improve the quality of the best works.
Democratic organization of security research AI:
Democratic organization of research and implementation of AI security and the use of democratic AI-tools will have advantages over the traditional hierarchical one.
Geographical factor of AI security studies:
Since authoritarian countries are the source of the greatest AI-threats, it seems especially important that people working on AI safety have an understanding of the work that is being done here.
It also seems valuable to have strong connections and 1-rank direct lines of communication between those working on the political use of AI and those working on safety.
Why Russia?:
It is difficult to say which country will lead the apocalypse race:
1. In the US, advanced technology, but democratic control somehow works,
2. In the USA - orientation on commercial use of AI, in Russia - on political, that is on direct threat.
3. In China, the demand for political use of AI is less than in Russia, since in it political stability is higher than in Russia.
4. China is lagging behind precisely in the development of AI technologies.
5. In Russia, traditionally, the rate on special operations and propaganda is higher.
Russia traditionally hides its costs and development.
In the nuclear race, Russia also initially lagged behind the United States.
In the AI race, the time is compressed.
|