AI & Society
What happens when intelligence is no longer ours alone?
I follow developments in AI research and work on the philosophical and societal questions they raise: consciousness, human judgement, and the authority we assign to systems we may not fully understand.
A system that helps someone write a letter presents one set of questions. A system that helps an employer assess an applicant, or a researcher decide which result deserves confidence, presents others. At some point we have to ask what has been delegated, who can challenge the conclusion, and who carries the consequence.
Keep the questions distinct
Capability, agency and an inner life.
The foundation model
The trained computational model can generate, represent information and perform increasingly demanding tasks. Its competence alone does not establish a self or a point of view.
The agentic system
A model can be placed within a system that adds tools, memory, goals and permission to act. That is functional agency. What it can do depends on the architecture and access it is given.
Possible subjectivity
Whether anything is experienced from within is a separate question. Fluent language, self-description and autonomous behaviour do not settle it.
Judgement and public responsibility
Can we still answer for the decision?
Consider the applicant who is told that a system has assessed them unfavourably. A person may have approved the process, but can that person explain the result, inspect the relevant evidence and correct a mistake? Being present somewhere in the chain is a thin form of responsibility if there is no practical ability to question what happened.
I am interested in what meaningful oversight would require in workplaces, education, public institutions and research. AI can be useful without becoming the final authority on which purposes deserve to be pursued or whose interests should count.
A developing research question
When AI helps research AI.
The oversight question becomes sharper when AI helps design experiments, write code or assess results in AI research itself. A researcher may gain speed while finding it harder to inspect every consequential step.
The question I am developing is how to distinguish useful assistance from a transfer of epistemic authority—the authority to decide what should count as knowledge. What evidence can people independently check? Where can they challenge the system? Which decisions are they actually retaining?
This inquiry follows what systems do and what people permit them to do. Claims about inevitable runaway progress would need evidence of their own.
Consciousness and inner experience
Could anything matter to a machine from within?
We know experience from our own lives: pain hurts, grief changes the day, music can absorb us. An AI system can describe these things persuasively. Whether that description is accompanied by experience remains an open question.
We should take the possibility seriously enough to investigate it and the uncertainty seriously enough to resist easy declarations. The question needs independent argument and evidence; neither impressive performance nor a confident claim about what machines cannot be provides the answer by itself.
Two inquiries, increasingly in conversation
The human questions remain with us.
Sacred Ideas and Incandescence ask how conscious beings might live meaningfully in a material universe. AI & Society examines what changes as intelligence becomes available in systems whose inward status is uncertain. Each inquiry has its own work to do. They meet in the decisions we make about trust, responsibility and the value of experience.