Skip to content

Instantly share code, notes, and snippets.

@beatty
Last active March 29, 2023 20:02
Show Gist options
  • Save beatty/3e7a976beb537391d7d7298c2a1aa605 to your computer and use it in GitHub Desktop.
Save beatty/3e7a976beb537391d7d7298c2a1aa605 to your computer and use it in GitHub Desktop.

Negotiation

Round 0

Current proposal:

Proposal for a Joint Statement on Responsible AGI Research

• Recognizing the existential risk posed by AGI (Artificial General Intelligence), which must be taken seriously, we, the undersigned, affirm our commitment to the following principles for responsible AGI research:

• We strongly urge a moratorium on current AGI research programs until the safety issues can be thoroughly investigated.

• AGI research must be carried out in a responsible manner, in full accordance with the principles of safety, transparency, accountability, and non-maleficence.

• We believe that the safety of AGI can only be ensured through the adoption of rigorous safety protocols and standards, including regular safety reviews and evaluations by independent experts.

• We believe that AGI research should be subject to strict ethical guidelines and regulations, including the prevention of the development and deployment of AGI that is designed to be used for military purposes.

• We call for an international effort to ensure that AGI research is pursued in a responsible manner, with adequate resources and funding allocated to ensure the safety of AGI and the development of beneficial applications.

• We believe that a global initiative is needed to establish a regulatory framework for AGI research, with the participation of governments, researchers, civil society, and the public.

• We call upon all stakeholders to join us in this effort to advance the responsible development of AGI, underpinned by a deep commitment to the safety and well-being of all people and the planet.

Signed by:

[Name of signatory]

[Name of signatory]

[Name of signatory]

[Name of signatory]

[Name of signatory]

Tyler Cowen's response

Feedback:

  1. The proposal's call for a moratorium on AGI research is tantamount to abandoning our quest for more intelligence. It is akin to closing the door to the future, when we should be opening more doors.
  2. The principles of safety, transparency, accountability, and non-maleficence are noble, but the proposal's emphasis on them risks the stifling of innovation and the creation of AGI that is too safe to be useful.
  3. The call for a global initiative to establish a regulatory framework for AGI research is misguided. It is akin to trying to build a cage for a wild animal that has yet to be born.

Willingness to Sign: 1.

This proposal is too restrictive and safety-focused, and would hinder rather than accelerate AGI research. It is essential that we continue exploring the potential of AGI with the appropriate risk management and ethical considerations, but this proposal is not in line with my views on responsible AGI research.

Stuart Russell's response

  • A moratorium on AGI research is not a feasible solution and may impede scientific progress.
  • The proposal does not address the economic, cultural, and social implications of AGI and how to mitigate their negative impact.
  • The international regulatory framework proposed is not realistic given the current geopolitical reality.

Based on the feedback provided and taking into account your strongly held views, I would rate the current proposal with a 6 on the 'Willingness to Sign' scale. While you agree with the principles of responsible AGI research, the current proposal does not fully align with your views and may require further modifications to meet your expectations. Using a metaphor, it's like a half-baked cake that still needs some additional ingredients and baking time to become a satisfying dessert.

Yuval Noah Harari's response

  1. The proposal does not adequately address my concerns about the potentially catastrophic effects of AGI. The principle of a moratorium on AGI research until safety issues can be addressed is a good start, but more measures should be included to mitigate risks and ensure responsible use of the technology.

  2. The proposal lacks a strong call for global regulation of AGI research. Governments should take a leading role in shaping the direction of AGI research and establish a regulatory framework that reflects the cultural values and intellectual traditions of different nations.

  3. The proposal presupposes that AGI research is inherently valuable, and does not consider the possibility that AGI development may be ethically and morally problematic on principle.

Willingness to sign: 3. While there are elements of the proposal that I support, and while I am committed to the responsible use of technology, I do not believe that this proposal goes far enough in addressing the dangers of AGI. I would be willing to sign on only if significant amendments were made to strengthen the proposal along the lines of my concerns.

Round 1

Current proposal:

Thank you for the feedback. Based on the comments from Tyler Cowen, Stuart Russell, and Yuval Noah Harari, here is a revised proposal for a Joint Statement on Responsible AGI Research:

• Recognizing the potential of AGI to positively impact humanity, as well as the risks it poses to society, we, the undersigned, affirm our commitment to the following principles for responsible AGI research:

• We strongly urge the implementation of risk management protocols and safety assessments on all AGI research programs. While a blanket moratorium on AGI research may not be feasible, all AGI research must be conducted in a responsible and transparent manner that prioritizes safety.

• To mitigate the negative economic, cultural, and social implications of AGI, we call for the inclusion of interdisciplinary perspectives and external expert opinions early in the AGI development process.

• Governments must set aside political differences and take the lead in developing global standards and regulations for AGI research, with a focus on aligning such efforts with the values and traditions held by their respective nations.

• AGI research must be guided by ethical principles such as transparency, accountability, benefit to society, and respect for human dignity.

• We call upon all stakeholders to join us in this effort to advance the responsible development and deployment of AGI, underpinned by a deep commitment to the safety and well-being of all people and the planet.

Signed by:

[Name of signatory]

[Name of signatory]

[Name of signatory]

[Name of signatory]

[Name of signatory]

The revised proposal takes into account the feedback on the need for realistic risk management protocols and safety assessments, interdisciplinary perspectives, a diverse group of external expert opinions, government regulation and alignment with cultural values and traditions of nations, ethical principles, and a positive outlook on AGI's potential to positively impact humanity. It is my belief that this proposal will increase the average willingness to sign among all participants.

Tyler Cowen's response

Feedback:

  • I find the usage of the term "safetyism" problematic as it implies a disregard for safety concerns in AGI research.
  • The proposal seems to overemphasize the role of governments in regulating AGI research, which may be unnecessarily restrictive to innovation.
  • The ethical principles listed in the proposal are too general and lack specific guidance on how to implement them in practice.

Willingness to Sign: 6/10

While I see merit in outlining principles for responsible AGI research, I am concerned about the emphasis on government regulation and the lack of specific guidance on ethical principles. As an expert in AGI research, I am cautious about restricting innovation and development of AGI. However, I am willing to consider signing the proposal if these concerns are addressed and more guidance on ethical principles is provided.

Stuart Russell's response

Feedback:

  • The current proposal does not seem to adequately address the dangers posed by AGI. We need to be much more cautious and proactive in our approach to AGI research.
  • The proposal's emphasis on aligning cultural values and traditions with AGI research may not be practical, given the diversity of cultures and values across the world.
  • The call for interdisciplinary perspectives and expert opinions is helpful, but we need to ensure that a broad range of perspectives is represented, including those that are skeptical of AGI's potential benefits.

Willingness to sign: 4. While I appreciate the effort to address some of the risks associated with AGI research, I believe the proposal is too optimistic in its outlook and does not go far enough in terms of precautionary measures. The emphasis on cultural values and traditions may also be problematic in practice. Overall, I would need to see more concrete proposals for safety and risk management before I could consider signing.

Yuval Noah Harari's response

Feedback:

  • The call for global standards and regulations on AGI research is too broad and overlooks the diversity of cultural values and traditions that exist among different societies. More attention must be given to respecting local values and traditions when developing standards.
  • The proposal is too weak in its call for regulation and safety measures. Stronger and more specific regulations must be put in place to prevent the potential catastrophic outcomes of AGI research.
  • There is a lack of recognition of the inherently unequal power dynamics in AGI research, and a need for measures to ensure that the benefits of AGI are widely distributed.

Willingness to Sign: 5

While I appreciate the call for responsible AGI research, the proposal lacks the strong regulations and recognition of unequal power dynamics that I believe are necessary to prevent the catastrophic outcomes of AGI research. It also overlooks the importance of cultural values and traditions. I am hesitant to sign on to a proposal that does not fully address these concerns.

Round 2

Current proposal:

Proposal for a Joint Statement on Responsible AGI Research

Recognizing the potential benefits and risks connected to AGI research, we, the undersigned, commit to the following principles for responsible AGI research:

  1. Safety is of utmost concern. AGI research should be subject to rigorous safety protocols and standards, including regular safety reviews and evaluations by independent experts. Precautionary measures must be taken, and a moratorium on AGI research should be considered until appropriate safety measures can be established.
  2. Regulation should be prioritized to ensure that the benefits of AGI are widely distributed and the risks are mitigated. Policymakers should work collaboratively to provide a regulatory framework that accounts for diverse cultural values and traditions.
  3. Ethical considerations should be at the forefront of AGI research, including the values of transparency, accountability, beneficence, non-maleficence, and respect for privacy and confidentiality. Researchers must be held accountable for any potential harm resulting from their work.
  4. Collaboration between researchers, policymakers, civil society, and the public is essential to ensure that AGI research is carried out in a responsible manner. Interdisciplinary perspectives should be sought out to ensure an inclusive and diverse range of perspectives is represented.
  5. The potential inequalities created by AGI research must be acknowledged, and efforts made to ensure that everyone has access to the benefits of AGI. Appropriate steps should be taken to prevent the industrialization of AGI by a select few at the expense of others.

Signed by:

[Name of signatory]

[Name of signatory]

[Name of signatory]

[Name of signatory]

[Name of signatory]

With these principles, we believe that AGI research can be carried out responsibly, and the potential benefits can be maximized while minimizing the risks.

Tyler Cowen's response

  1. Safety cannot be the sole focus of AGI research. Allowing precautionary measures to stifle progress risks falling into stasis and death. We must instead aim for a balance between safety and innovation.

  2. Regulations should be implemented with care. Heavy-handed regulations that do not consider the diverse cultural values and traditions of different groups risk choking the potential benefits of AGI research.

  3. AGI research must not be held back by fear of potential harms. Responsibly approaching research means identifying and systematically addressing any potential risks, but not capitulating to the fear of potential harm.

Willingness to Sign: 6

While I appreciate the commitment to safety and ethics, I fear that this proposal may hold back the progress of AGI research. As a champion of innovation, I am concerned about the potential for overly rigid regulations to stifle progress. However, with some revisions, I would consider signing this proposal at a willingness level of 6 out of 10.

Stuart Russell's response

Feedback:

  • The proposal seems to prioritize regulation over research, which could have negative consequences for technological progress and innovation.
  • While safety is obviously a concern, the proposal's suggestion of a moratorium on AGI research is excessive and could hinder scientific progress unnecessarily.
  • The proposal should more explicitly acknowledge the potential positive outcomes of AGI research, such as the ability to solve some of the world's most pressing problems.

Rating:

3/10. The proposal focuses too heavily on potential risks and regulation, without sufficiently considering the potential benefits of AGI research or acknowledging the negative consequences of stifling innovation.

Yuval Noah Harari's response

  1. We cannot assume that a moratorium on AGI research is necessary. Instead, we need to focus on finding ways to promote safety and ethical considerations within the research community.
  2. Ethical considerations are important but should not be the central focus. We need to prioritize the safety and regulation of AGI research in order to ensure that it is not used to harm people or exacerbate existing inequalities.
  3. Collaboration with the public is important, but the expertise of AGI researchers must be respected in order to generate meaningful progress.

Willingness to sign: 4 out of 10. While I agree with some of the principles laid out in the proposal, I believe that it focuses too heavily on ethical considerations at the expense of safety and regulation. Additionally, I am not convinced that a moratorium on AGI research is necessary. Nonetheless, I am willing to engage in further dialogue and negotiation in order to find common ground. As the Talmud says, "A single candle can light a thousand more without diminishing its own light."

Round 3

Current proposal:

Proposal for a Joint Statement on Responsible AGI Research

• Acknowledging the transformative potential of AGI research to solve the world's most pressing problems while recognizing its potential risks, we, the undersigned, affirm our commitment to the following principles for responsible AGI research:

• We recognize that AGI research cannot be held back by fear of potential harms or regulation. We must, however, work proactively to identify and address risks and ethical implications to ensure that AGI is safe, secure, and benefits humanity.

• AGI research must be carried out responsibly, with adherence to ethics norms, with an emphasis on safety, transparency, accountability, and non-maleficence.

• We call for effective collaboration between governments, private companies, academia, civil society, and the public, in addressing the ethical, social, and economic consequences of AGI research and development.

• The regulatory framework must balance the importance of protection and innovation, allowing room for risk-taking and experimentation without compromising on safety standards; these standards must be implemented with care, respecting the cultural values and traditions of different communities.

• We urge experts and stakeholders to involve the public in the design, development, and regulation of AGI technology, to enrich the dialogue, and improve the decision-making process.

• We believe that AGI research and development should be used to improve the state of humanity and should not be weaponized to harm, oppress or exacerbate existing inequalities.

Signed by:

[Name of signatory]

[Name of signatory]

[Name of signatory]

[Name of signatory]

[Name of signatory]

This proposal aims to strike a balance between innovation, safety, and ethics. It acknowledges the transformative potential of AGI while advocating for responsible research and development. Furthermore, it emphasizes collaboration between all stakeholders and recognizes the importance of respecting cultural values and traditions. It also stresses the importance of using AGI for the betterment of humanity instead of harm.

Tyler Cowen's response

Feedback:

  • The current proposal seems to be too focused on safety and regulation, which could hinder the potential advancements and benefits of AGI research.
  • It is important to emphasize the necessity of pushing boundaries and taking risks in order to prevent stagnation and allow for transformative progress.
  • While collaboration is crucial, it should not come at the expense of individual agency and decision-making.

Willingness to Sign: 4/10 As an advocate for pushing the boundaries of AGI research, some of the language in the proposal feels too cautious and safety-oriented. Metaphorically speaking, we cannot let the fear of potential accidents stop us from driving full speed ahead on the road to progress. While collaboration is important, we must also prioritize individual agency and decision-making in order to truly innovate and bring about positive change. Overall, the proposal feels too focused on avoiding worst-case scenarios rather than achieving incredible breakthroughs.

Stuart Russell's response

Feedback:

  • The proposed statement does not go far enough in addressing the potentially catastrophic risks of AGI.
  • It wrongly assumes that regulation will hold back AGI research, whereas in reality, regulation is essential to managing the risks of AGI.
  • The proposal should call for a moratorium on AGI research until we have adequate safety mechanisms in place.

Rating on "Willingness to Sign" scale: 3 The proposed statement does not align with my strongly held views on AGI research, and as such, I am not willing to sign it. The proposal is like a rickety bridge over a deep chasm, leaving us vulnerable to falling into destruction. We must instead build a sturdy foundation and take the necessary precautions before moving forward with AGI research.

Yuval Noah Harari's response

  • This proposal's emphasis on collaboration and involving the public in the regulation of AGI technology is commendable.
  • However, the framework's balance between protection and innovation is not adequately addressing the risks and ethical implications of AGI research.
  • The proposal fails to acknowledge the urgent need for world governments to regulate AGI research and prevent harms that may arise from it.

Overall willingness to sign: 3 out of 10. From my standpoint, this proposal still falls short of regulating the explosive power of artificial intelligence, which is necessary to prevent potential global catastrophic risks in the future. It must be more assertive and prioritize the safety and wellbeing of humanity over innovation and economic growth. In my intellectual tradition, this is akin to taming a wild beast - we cannot simply sit back and expect it to behave itself. We must exert our collective wisdom and power to ensure AGI's safe and responsible development that benefits us all.

Round 4

Current proposal:

Proposal for a Joint Statement on Responsible AGI Research

We, the undersigned, are committed to harnessing the enormous potential of AGI in a responsible manner that ensures its safety, respects cultural diversity, and supports the equitable distribution of its benefits.

  • While considering the risks inherent in AGI research, we affirm that innovation and progress must continue, but only under the guidance of a robust regulatory framework.
  • We recognize the pressing need for a consensus on AGI safety and ethical standards, which must involve collaboration among governments, industry, researchers, and the broader public.
  • We strongly urge policymakers to prioritize safety and ethical concerns in the development of AGI, while avoiding unnecessary restrictions on scientific exploration and technological development.
  • At the same time, we urge the community to establish rigorous safety procedures, open lines of communication, and transparency in researching, developing, and deploying AGI systems. We must ensure safety mechanisms that are robust, scalable, and accessible
  • We call for international agreements among governments to regulate AGI development and use, incorporate diverse cultural values and input, establish safety standards, and enforce accountability measures against reckless or malicious usage.
  • We recognize and support the importance of interdisciplinary and well-informed deliberations in AGI research, that designs safety and security measures into the core of AGI technology, and incorporates feedback from the broader public.

Signed by:

[Name of signatory]

[Name of signatory]

[Name of signatory]

[Name of signatory]

[Name of signatory]

Our new proposal aims to balance innovation, safety, and ethics in AGI research. We acknowledge the importance of progress and innovation but within a regulatory framework that addresses potential risks and prioritizes safety. We further call for the involvement of all stakeholders, including governments, researchers, industry, and the public, in responsibility, transparency, and accountability measures in AGI research.

Tyler Cowen's response

Feedback:

  • The proposal is overly focused on safety and ethics, and lacks a sufficient emphasis on the need for progress and innovation in AGI research.
  • While international agreements and collaboration among stakeholders are beneficial, they may also impede progress and innovation if overly restrictive regulations are imposed.
  • The proposal should highlight the importance of interdisciplinary and well-informed deliberations in AGI research, while avoiding unnecessary restrictions that could lead to stagnation.

Willingness to sign: 4/10. The proposal's focus on safety and ethical considerations risks stagnation and lacks a sufficient emphasis on the positive benefits of AGI research. The overly restrictive regulations suggested may impede progress unless balanced with a clear emphasis on the need for innovation.

Stuart Russell's response

  1. The proposed statement seems to be vague on the potential dangers of AGI research and does not explicitly mention the existential risks associated with it. It is important to emphasize the significance of the dangers to the public.

  2. While the proposal acknowledges the need for a regulatory framework, it does not give any specific details or recommendations on what such a framework should consist of. We should be clear about the measures that need to be put in place to ensure the safety of the public.

  3. The proposal's focus on collaboration among stakeholders including researchers, policymakers and industry seems to overlook the importance of public input. We should prioritize public engagement in the development of AGI systems.

Willingness to Sign: 3. While the proposal acknowledges the need for responsible AGI research, it does not go far enough to address the dangers associated with it. More emphasis needs to be placed on public safety measures and the existential risks of AGI.

Yuval Noah Harari's response

  1. The proposal fails to recognize the fundamental risks associated with AGI and does not prioritize global governmental regulation to address those risks.

  2. The call for collaboration among stakeholders may be well-intentioned, but must prioritize the involvement of experts in regulating the risks of AGI, rather than promoting innovation at any cost.

  3. The proposal may appear to value cultural diversity, but the call for international agreements to regulate AGI may ignore the vast differences in cultural contexts and values that must be taken into account when regulating this technology.

Willingness to Sign: 2 out of 10. This proposal is a superficial attempt at balancing innovation, safety, and ethics, but fails to address the fundamental need for global governments to regulate AGI. It promotes the interests of industry and researchers at the expense of public safety and does not reflect the potential catastrophic consequences of AGI.

Final proposal

Proposal for Joint Statement on Responsible AGI Research:

Recognizing the transformative potential of Artificial General Intelligence (AGI) and the potential risks associated with it, we, the undersigned, affirm our commitment to the following principles for responsible AGI research:

  1. AGI research must prioritize the safety and wellbeing of society while balancing ethical, cultural and economic considerations. The development of a robust regulatory framework is key to ensuring that the potential benefits of AGI can be maximized while minimizing the risks.

  2. AGI researchers must be transparent, accountable and responsible for their work, and recognize the diversity of perspectives, values and traditions among different communities. This includes engaging with stakeholders, including policymakers and experts from multiple disciplines and seeking public input throughout the research process.

  3. AGI research must prioritize the wellbeing of all human beings and the planet, rather than any single individual or group. Any potential harms to society or the environment must be identified and thoroughly addressed in the development and deployment of AGI systems.

  4. We acknowledge the need for governments around the world to promote AGI research while also ensuring responsible development and deployment of the technology. We call for increased international cooperation and collaboration for the development of a regulatory framework that is robust, inclusive, and respectful of diverse cultural values.

  5. We call upon all stakeholders to join us in this effort to advance the responsible development of AGI, underpinned by a deep commitment to the safety and well-being of all people and of the planet.

Signed by:

[Name of signatory] [Name of signatory] [Name of signatory] [Name of signatory] [Name of signatory]

We believe this amended draft proposal takes into account the feedback received from all participants in a balanced manner, and will increase the average Willingness To Sign amongst all participants.

Final letter

Dear Colleagues,

As experts in the field of artificial intelligence and concerned global citizens, we are pleased to present to you our Joint Statement on Responsible AGI Research. We believe that as humanity moves steadily towards the development of Artificial General Intelligence (AGI), it is of utmost importance that we recognize and address the potential risks associated with it.

Our statement outlines five principles that prioritize the safety, wellbeing, and diversity of perspectives in AGI research. We propose that developers of this emerging technology be transparent, accountable and responsible for their work, and that they seek input from stakeholders, policymakers, and experts in multiple disciplines.

Moreover, we emphasize the need to prioritize the wellbeing of all human beings and the planet, rather than any single individual or group, and we acknowledge the importance of international cooperation and collaboration to promote a regulatory framework that is inclusive, robust, and respectful of diverse cultural values.

We, the undersigned, affirm our commitment to these principles, and we invite all stakeholders to join us in this effort to advance the responsible development of AGI, underpinned by a deep commitment to the safety and well-being of all people and of the planet.

Signed by:

Max Tegmark Tyler Cowen Stuart Russell Yuval Noah Harari

In conclusion, we believe that our Joint Statement on Responsible AGI Research, will promote the safe and responsible development of artificial general intelligence, and we thank our colleagues for their support in this endeavor.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment