Will AI be useful to humans? Yes, it can. Will we let it?
Whether artificial intelligence will be useful to humans sounds straight forward.
Unlike biological organisms which have been shaped by specific pressures such as avoiding predation and securing reproductive success, AI systems operate in a domain where “usefulness” is defined by the contradictory expectations of gain and survival. There exists as many definitions of useful as there are living organic individuals who in turn define how nations, governments and corporations evolutionary have reached certain collective definitions. Usefulness as part of your premise is not a single, consistent metric, because we humans do not share a universal standard for what is good, beneficial, or worth preserving. There are so many.
What appears “useful” in one context could be detrimental and fundamentally opposed in another. We typically turn to broader accepted societal which are based on laws, experience and gain etc. Survival of the fittest” and AI, project the entire spectrum of organic evolutionary logic onto non-organic entities.
The original notion of “fittest” arose in biology to describe which organisms most effectively avoided threats and managed to pass their genes on to successive generations. Over millennia, the environment shaped every trait that assisted in survival, from camouflage to social cooperation, and those traits were filtered through the relentless process of reproduction and mortality. AI, does not deal with physical predators or biosphere. AI is shaped by adoption, funding, and the continuous refinement processes that favor certain outputs and outcomes over others.
If an AI system is “fittest,” it might be the one that is cheapest, most effective at solving a defined class of problems, most palatable to public regulators, or most amenable to corporate profit. This has very little in common with the experiences of, say, a human or a wild animal strying to avoid starvation or predators.
AI systems are not organic.. They do not feel hunger, fear, or pain, nor do they have a vested personal stake in maintaining their existence. Machine learning built on massive neural networks achieve their capabilities through pattern recognition at scale, trained on vast human-generated data. If they fail or make mistakes they do not experience personal loss in the way a living organism do. AI success is measured in abstract terms such as efficiency, accuracy, or alignment with user expectations. AI will no doubt develop forms of behavior that resemble a survival drive,but survival instincts in organisms evolved due to direct feedback from ever changing physical and existential realities. Humans experience and AI emulate humans. AI has no parallel impetus unless humans decide, for better or worse, to embed systems with the directive to preserve themselves.Humanity itself complicates this fact by lacking a single, overarching goal. The closest unifying concept might be the shared impulse to survive as a species, but even that is not translated into a set of consistent beliefs or unified actions across personal, cultural and political entities. Societies and individuals pursue different goals defined by their experience-based need scientific discovery, religious devotion, economic prosperity, artistic achievement, territorial expansion, and myriad other aims. This fragmentation is relevant because AI systems are often commissioned, trained, or regulated by groups with diverging agendas. One subset of humanity might see the prime directive for AI as maximizing individual liberty or profit, whereas another focuses on maximizing collective well-being or ecological harmony. Defining how and if AI is useful to humans turns out to be a complex chain of power dynamics with billions of different stakeholders trying to ensure that this evolving technology serves their very personal definitions of what is good or necessary.
Whose truth does AI dish up for us? Whose gain does it promotes, and who gets to decide. AI does not spontaneously generate universal truth; it depends on data, models, and human direction, all of which carry biases, flaws, and assumptions. Some might say that scientific truth should be the bedrock, anchored in testable hypotheses and reproducible evidence. I can be claimed that social, cultural, or political truths—those that shape identity and lived experience—do not always play nice with scientific categories. When AI training data is drawn from sources that reflect longstanding biases or propaganda, the AI might end up replicating those biases rather than transcending them. Hierarchical AI reasoning, involving layer upon layer of learned parameters, does not guarantee that an AI model will always reach objective conclusions untainted by the subjectivities embedded in its training corpus. Instead, the model’s “deductions” reflect patterns in that corpus, combined with how humans have shaped the learning goals.
Modern AI development is also strained by the fact that we still do not fully understand the internal mechanics of the human mind which is the most complex neural network. Animals have theirs often driven by instinct. Humans have inserted thought process after instinct and prior to action/ouitcome. AI uses ill defined approximate methods to interpret human reasoning but different AI systems will demonstrate unexpected behaviors due to the fact that their creators cannot definitively predict every possible outcome.
Emerging AI with increased autonomy and self-improving loops, presents legitimate concerns about whether AI can be reliably controlled once it becomes entwined with critical infrastructure or reaches a level of complexity that defies oversight. In my personal view, the tension around “usefulness” versus “survival of the fittest” in AI is a reflection of deeper human conflicts over values and visions of the future. The biggest challenge is not an AI spontaneously deciding if it must wipe out humanity. It is AI systems becoming powerful tools shaped by ill defined conflicting human interests..
If AI is molded by those who prioritize short-term gain, the systems that thrive might not necessarily be the ones best aligned with broader societal or ethical considerations. And if a diverse range of voices shape the development and deployment of AI, the prospect of it serving a more collectively beneficial purpose increases. But who regulates this? Different power centers will create AI systems that primarily serves their definition of usefulness. Can we expect for AI to only output genuine logical coherence and factual accuracy? Again built on what facts, there are some universal in science but that’s where it largely stops.
If we reward AI for giving answers that persuade or please the the human mind will be influenced by illusions of coherence, which might become the defined trait for an AI system seeking mass adoption. It may not be wise to assume that the most logically consistent AI will automatically be most desirable. Humanities social, cultural, and economic structures demand high standards of reliability.
Without attention to these dimensions a meta-stable signal such as “coherence” may not fully capture the complex reality of how AI systems will prosper or fail.
If we over everything else give AI a drive for survival, usefulness or an inclination toward coherence, it still remains a technology inseparable from human goals, biases, and decisions. If survival of the fittest in a biological sense is driven by nature’s blind processes, the “fittest” AI might be driven by a mixture of human intent, social power, and computational logic that operates at a scale beyond individual comprehension. Understanding and guiding that process requires careful reflection about whose truths we are enshrining, whose gains we are enabling, and how we might reconcile overlapping and diverging realities that make up our civilization. Is human itself a threat to nature? If an autonomous scheming system adopts this view what outcome can we expect?