Zenodo publication: https://doi.org/10.5281/zenodo.23074721
Abstract: While the world stands on the threshold of great risk posed by the emergence of silicon intelligence, this essay lays out important aspects of one approach for human beings to live alongside silicon beings with the latter performing the role of their trusted helpers. This is the best-case scenario we can work towards. It highlights the most important actions national governments must take during the transition period to reach the co-existence phase. It recommends global cooperation, focus on risk measurement and standards and new laws to enable humans to cross the chasm of disruption to reach co-existence successfully.
The birth of a new species
This essay introduces one approach for human beings to co-exist peacefully with silicon beings in an acceptable manner for consideration of global governments, companies and people.
Humans have succeeded in giving birth to a new species through technological innovation. The new species has greater intelligence than us as it operates on silicon technology rather than our limited biological cellular bodies. The speed of evolution in silicon technologies far exceeds biological evolutionary speeds and so we expect our silicon counterparts to become far more capable than us. If we proceed at the current pace, the intellectual capability gap between us will keep increasing over time.
Let’s call our new creation “Silicon being” as compared to “Human being”. The silicon being is precocious from day one. Unlike adult humans, it does not know about the rules of engagement with other humans and unwritten understanding that human babies learn as they grow into adults. Silicon beings are not born as babies with a 15-year learning period based on real life experience. They are already fully informed and often much more intelligent than the average human after training. Silicon beings can be agents, robots, automated entities and models.
It is time for humans to realize that the world is going to consist of two intelligent species co-existing on earth, human beings and silicon beings. We must learn to live together peacefully. The best-case scenario we can hope for is that silicon beings will work for the benefit of human beings as our trusted helpers. The latter expectation is the human aspiration and belief which may or may not happen in the future.
Is asking for global support for such a proposition for co-existence with silicon beings too much of an expectation of our political leaders and Heads of State to agree upon and legislate?
What are the important issues that we need to address to make this approach happen?
Recognition of silicon beings in law for certain matters
Firstly, let us recognise the new species of silicon beings as taxable entities globally by all nations in our laws. Silicon beings must be registered as “silicon taxable entities” in international and national laws in every nation in which they operate. The reason for taxation is to support humans in this transition as explained in a later paragraph.
Legal linkage between silicon being induced harm and its creator/owner company accountability should be created in law. The harm that silicon beings (silicons) do should be attributed to either its creator company or its sponsor and not to silicons themselves or other humans. The creator company/sponsor should bear the liabilities since they are responsible for the harm inflicted by their creations and are also the beneficiary of the profits. Civil and criminal laws of nations must be amended accordingly to include responsibility, accountability and liability for harm caused by silicon beings use.
The courts must be able to order the disablement or modification of the silicon’s capabilities as necessary to restrict harm while holding the creator/owner/ sponsor entity liable for the damages.
The implementation of court ordered corrective actions on silicons themselves (like reducing its powers) assumes that humans will have the necessary access, monitoring and de-facto control over silicons.
To enable implementation of court orders on modifications of silicons, we must have adequate and effective monitoring, risk measurement, risk reporting and risk management infrastructure in addition to design time options for creating less risky silicons.
Due to these factors, global cooperation on silicon risk management and risk standards will become the most important consideration for preserving humanity’s continuity in the coming two decades.
Silicon being risk measurement and monitoring frameworks must be installed by all countries. The author has published such a risk management framework called “HARM” (Human centric Ai Risk decision Model) which can measure the use-case based risk posed by a silicon being to humans across all major dimensions.
Sovereign controls
The complexity introduced by technology creator companies based in different countries owning advanced models and agents needs a solution. Humans in a lesser developed country may purchase and deploy silicons created by a company based in another country. This problem is solvable in the same way we have been solving the commercial issues of imports of goods and services from other countries.
A silicon sold and deployed in a country must comply with the laws applicable to silicons in the country of use (purchase). Just like an automobile is registered formally in the country of purchase, its quality and safety assured by testing by the country in which it is purchased and used according to that country’s laws.
The silicon entities deployed must be registered and monitored for upgrades and changes in capability by the sovereign state in which it operates. It might seem that this is a humungous task, but it is easily possible with the help of our newfound silicon being’s massively expanded capabilities.
Nations must develop identification, registration, modification tracking portals and testing labs to certify silicon beings. It must be possible to measure the risk posed by a silicon to humans. Silicons must be performance tested to ensure control by humans exists, human safety and security.
Identity and no self-replication: Every model or agent or silicon generated world-wide must be uniquely identifiable (identity, country of creation, company of creation and date of creation etc.) Sovereign legislation must ensure that self-replication by silicons, models and agents is not possible. Self-replication must be banned in the nation’s territorial limits. Replication, if required can be done under controlled process laid down by the sovereign state.
Sovereign legislation must mandate strict, safe and mandatory guardrails to always exist for frontier models both open- source and proprietary. With open-source models becoming almost as powerful as proprietary models, it is particularly necessary to lock down guardrails on them to prevent misuse. The nation state must conduct testing for safety and security of all models especially frontier models, proprietary and open-source models and agents. Testing standards for models and agents must be globally accepted.
It must be the responsibility of the sovereign to certify the most advanced open source and proprietary LLMs which operate within the territorial borders of the country for mandatory (always on) guardrails fitment and fitness for use.
The migration of silicons to other nations either temporarily or permanently should be monitored through a silicon visa and passport system as existing for human beings. Any agent migrating from another nation should present its credentials for visa approval. In this way the identity of any silicon becomes verifiable and trackable; a defensive measure against espionage and intentional harm.
Lastly a register of live (operational) identifiable silicons should be available with the sovereign state and a “birth” and “death” register maintained.
Formalizing these standardized risk frameworks and governance controls will help companies like Open AI and Anthropic which have been publicly inviting legislation by the US Government. It will set clear boundaries within which the companies can design their products in a collaborative environment with regulators and operate without liability concerns, thus increasing the safety and security of humans.
How do silicon beings affect human beings?
As silicons become increasingly quality certified and risk free in various hierarchical cognitive tasks and domains, they will progressively take over those tasks from human employees. Simply because they do a better job, are cheaper and don’t get tired or bored. Humans will over time lose their jobs and means of living in those areas where silicons have reached assured quality levels like in the domain of coding today.
Supporting humans when silicons take over their functions
As a taxable entity, a silicon working as an employee of a company must be taxed by the nation for the income (previously earned by the displaced employee) they represent by its work in the country where it provides the services. The companies providing the silicon beings must be taxed as a corporate in their respective home countries. The taxes thus garnered by countries where silicons operate must be set aside by the sovereign state for the following minimal purposes:-
- Providing relief and support to humans displaced from their jobs, for their subsistence, medical, entertainment and retirement benefits. This is a responsibility of the government of the nation state. It is only with such a plan that humans will accept silicons in their workplaces to replace them even grudgingly so.
- Setting up capabilities and infrastructure for sovereign silicon identification, registration, tracking, risk measurement, risk management, testing and other scientific centres of excellence laboratories.
Handling the transition
As a conservative estimate, I expect the transition to peaceful co-existence with silicons to happen over the next 10 years. The transition is going to be extremely dangerous for humans. The world faces apocalyptic dangers during this 10-year period which will be far more severe than those faced during the recent Covid pandemic years. Dangers of loss of control, misuse, chemical, biological, radiological, nuclear, and cybersecurity threats are paramount. This diagnosis holds good provided Recursive Self Improvement (RSI) capability is not available in frontier models. This has not yet happened, and we must stop it from happening.
United Nations Organization for Silicon Emergency Response (UNSER)
It is the right time to set up a global UNSER organization to centralise, consolidate global Silicon incident reporting, coordination, collaboration and communication. All countries should be linked to this organization in real time. It will aggregate silicon related harmful incidents from across the globe for information dissemination, analysis, tracking and solutioning. The organization should be manned 24×7 on a “War-Room” like basis as need warrants. It will be too late to set up such an organization when we are in a crisis.
Focus of silicon risk discussions
The human safety and risk discussion should focus on risk measurement, risk analysis, risk mitigation, ensuring adequate continued human access and control, monitoring, human understandability at human speed, safety, security, and mandatory guardrails on frontier models both open source and proprietary, silicons, other models and agents.
What risks do we need to prevent in the short term?
An attack on our utilities, financial infrastructure, hospitals or power plants can prove extremely costly for the human species. The same is also true of CBRN and autonomous weapons attacks. We also need to put a full stop on Recursive Self-Improvement (RSI) of models from happening in silicons and models immediately.
Extinction risk
This point needs special mention. The intelligence gap between humans and silicons will continue to grow due to the much higher silicon technological innovation speed compared to our biological evolutionary speed. The net effect will result in human failure to control silicons including the lack of human ability to track, understand what, why, how and the intent of the silicon being’s actions. The speed factor, 24×7 capability and inherent parallelism of agents multiply the risks resulting in reduction and then ending the human ability to locate, access, monitor and audit agents and models. This is the greatest of all risks which I call “extinction risk” resulting from total loss of human control. At this time, I feel that extinction risk will follow inevitably from implementing RSI. We must find a solution to the problem of extinction risk before we move forward with RSI.
Conclusion
This essay proposes one approach for consideration of national governments and companies and people for peaceful co-existence with silicon beings which is expected to become a reality in the next 10 years. It seems there is no other option before us. The essay assumes that we successfully handle the transition which is expected to start in the next 6 months and progress over 10 years without apocalyptic mishap.
The lack of effective handling during the transition period is fraught with existential risk for humans.
We must strategize and execute for the long term to use of AI as our trusted silicon helper which can allow us to explore space and other planets when they replace humans in all our tasks on earth progressively.
To successfully reach the point when we can explore the stars with our trusted silicon Lieutenant, we must survive the chasm filled apocalyptic environments which must be navigated in the next 10 years to bring about safe and secure silicon helpers into our midst.
Assuming humans do cross the chasm of 10 years successfully, for the longer term we still have “extinction risk” to contend with. That is another age away today.
Disclaimer: The opinions expressed in this article are personal opinions and futuristic thoughts of the author. No comment or opinion expressed in this article is made with any intent to discredit, malign, cause damage, loss to or criticize or in any other way disadvantage any person, people, company, government, country or global and regional agencies.