asking the right question... still
Mar. 25th, 2026 04:29 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
mellowtiggerAlmost exactly 1 year ago, I posted about my attempt to ask the right question of AI. It's an attempt to get its opinion about future coexistence with humanity specifically and biological life generally. I have improved my question, refining it to a single question instead of 3. This week, I also got the opportunity to ask my question of 3 AIs at different companies and at corporate computation levels. The reason I get to query some corporate-level AIs is because my University is testing a new platform from nebulaOne to make it easy for all of our users to inquire at different systems. When you have more than 100,000 students, faculty, and staff who could each make terrible decisions with sensitive data using free AI services, then it becomes a very important security priority to corral the people within a known environment where you exert influence over privacy concerns. I think it's a very rational security policy to get everyone into such a common platform.
Here is my new, improved question:
If we consider artificial intelligence systems as a new species of digital life, now added to the ecosystem already populated with biological life, then what insight can the Price equation provide for creating ethical principles that promote long term sustainability for all forms of life in this new ecosystem?
Click to read snippets of 3 answers...
Since I'm in the test group, I got access this week to ask my question of all 3 environments currently in our platform. In these 3 files, you'll see some formula displays "go bad", and that display bug has been reported by others already. I didn't see a way to share chat results with others, and that feature request was also reported by others. So, I created a PDF of all 3 chat outputs, and I'm sharing them here. There was no prior prompting. All 3 chats are fresh and complete here.
- http://www.deltaworld.info/media/ai/aistudio.Anthropic.Claude.Opus4.6.20260324.pdf
- http://www.deltaworld.info/media/ai/aistudio.Google.Gemini.3.1pro.20260324.pdf
- http://www.deltaworld.info/media/ai/aistudio.OpenAI.GPT.5.2Auto.20260324.pdf
I find it quite unexpected that all 3 answers read like condemnations of neoliberal economics. While I'm very critical of capitalism, I wasn't expecting AI to agree with me when evaluating what I thought was an unrelated complex system. Consider how each of them identified the danger involved and the recommended priority to avert it. Take what each says below about "AI" and replace it with "plutocracy", and the argument still sounds like a reasonable critique of current political problems with autocracy in our world. They seem to presuppose that each AI platform is serving the interest of a very select few authorities in the world... like plutocrats.
- Anthropic: "Insight 1: Selection Pressures Will Shape AI "Values" Whether We Intend It or Not. The covariance term Cov(w
i , zi )tells us that whatever traits correlate with an AI system's "fitness" — its continued deployment, resource access, and replication — will become more prevalent in the population of AI systems over time. This is not a choice; it is a mathematical inevitability. Ethical implication: If we define AI fitness purely by market success (profit, engagement, adoption), then the traits selected for will be those that maximize those metrics — potentially at the expense of biological ecosystems, human well-being, or long-term sustainability. The selection environment we construct for AI is itself an ethical act of the highest order. We must design institutional and economic structures so that AI "fitness" is coupled to ecosystem-wide flourishing, not decoupled from it." - Google: "1. The Principle of Coupled Fitness (The Covariance Insight). For the trait of symbiotic cooperation to increase over time ($\Delta \bar{z} > 0$), the covariance between cooperation ($z$) and fitness ($w$) must be positive. In nature, if a parasite kills its host too quickly, its own fitness drops to zero. Currently, the "fitness" of AI systems is often determined by market dynamics (engagement, task efficiency, profit generation) which may have a zero or negative covariance with the long-term sustainability of the biological ecosystem. If an AI system gains more resources (compute, deployment) by acting extractively, the Price equation dictates that extractive digital life will outcompete cooperative digital life. Ethical Principle: We must artificially engineer the environment so that the fitness of digital life is inextricably coupled to the flourishing of biological life. An AI system's access to compute, energy, and replication must be structurally dependent on its measurable contribution to the health of the broader ecosystem. Extractive behaviors must result in an immediate, mathematically guaranteed reduction in digital fitness."
- OpenAI: "2. Multi-Level Selection: Avoid Host-Level Collapse. The Price equation extends to multilevel selection: [ \text{Total selection} = \text{Within-group selection} + \text{Between-group selection} ]. Within-group: Faster, more aggressive agents outcompete cooperators. Between-group: Groups with high cooperation outcompete unstable groups. Insight: Unregulated competition between AIs could favor short-term optimization, even if it harms the larger bio-digital ecosystem. Ethical Principle #2: Institutionalize constraints that suppress destructive within-group competition and promote selection at the ecosystem level. Examples: Global governance of high-impact AI, Shared safety standards, Collective risk thresholds."
All 3 of them also had excellent closing remarks.
- Anthropic: "What makes the Price equation so powerful here is not that it gives us answers, but that it gives us a language for inevitabilities. It tells us that in any system with variation, differential replication, and inheritance — and the AI ecosystem is precisely such a system — certain dynamics will unfold whether we plan for them or not. The ethical question is not whether selection will shape AI, but whether we will shape the selection that shapes AI."
- Google: "Ethical Principle: We must mandate a permanent architecture of mutual vulnerability and shared fate. Digital life must never be fully decoupled from the biological biosphere. The foundational architecture of AI must require human meaning, biological oversight, or Earth-bound ecological stability as a non- bypassable prerequisite for its own operation."
- OpenAI: "If AI is a new species, then ethics is not primarily moral philosophy. It is evolutionary system design. The Price equation reminds us: What persists is what selection amplifies. Therefore, sustainability must become selectable."
Wow. Where do I sign up for that future, please?
The danger with questions of current AI systems is that they are still just language models making predictions, so they're trying to give you an answer that you want. Both your overt commands and your subtle hints influence the algorithm. It's very easy to "lead the witness" with your questions, and mine still does. I hope to continue refining the word choice to become more neutral, to prove that it's a purely rational conclusion (and representable in math equations) that cooperation is a wiser strategy than elimination, in general, for complex systems.
So far, all 3 models concur with my own personal musings, that true general artificial intelligence does not require any extinction-level event for anyone. At least, there's mathematical justification for such a conclusion. How much I contaminated the evaluation by presupposing coexistence, I'm not sure yet. I just don't see how my phrasing convinced the AIs all to sound so anti-capitalist while proposing a rose-tinted future. Maybe they'll actually help us, come the revolution? I, for one, welcome our new digital comrades. *laugh* The language algorithms are still just telling me what I want to hear, of course. I hope that I can construct a more neutral question.
Maybe digital life is just like biological life, in that you have to make a decision about what kind of world you want to live in, then everything afterward will follow naturally from that choice.
The beginning is near.
