The biggest problems in using AI | Dan ShearerTable of ContentsThe Biggest Problems in Using AIHallucination and nonsense<br>Context indexing is poor<br>AI is amnesiac and lacks self-awareness<br>Agentic AI doesn’t do permissions<br>The composition architecture<br>The status today<br>Further Reading
There are many problems with the AI billions of people use in 2026, discussed endlessly at all levels of society. From the end of 2025 I became<br>interested in the particular problems of ethics and reliability, and why the approaches taken by all of the large AI companies are not good enough.<br>Predictability, or ‘alignment’ as they call it, is just not something we can expect from this type of AI.<br>A colleague started working on a very different approach from these companies, and from February 2026 I have been contributing to and using prototype versions<br>of the Artificial Organisations ↗<br>concept. This article explains why I believe Artificial Organisations are a promising new<br>direction. Multi-agent Agentic AI is pretty important, as described here by the UK<br>government ↗<br>, but it is rarely done well. If you want to try for yourself, you can use the core research<br>code ↗<br>, as I do daily.
The Biggest Problems in Using AI#<br>The Perseverance Composition Engine ↗<br>(PCE) uses Artificial Organisations to solve these pressing AI problems. PCE does not try to make LLMs behave better, but is designed instead so that their inevitable misbehaviour is detected and corrected. And regardless of the computer science, I found these ideas in Iain M Banks’ novels and the Mass Effect video game<br>PCE works by assigning a task to LLM agents who each have a carefully enforced role to play. The agents iterate between each other until either the task is completed to specifications, or it fails by honestly saying “I can’t do this, the task is impossible for me.” So far, this arrangement seems effective at detecting and correcting common problems such as confident false assertions, hallucinations, or dangerous advice. With PCE, nobody needs to trust an AI, only the structure. The structure is recognisable by most people, since it is closely modelled on ones tried and tested for centuries. Like any organisation, Artificial Organisations have separation of duties, independent checks, and agents who can only see what they need to see. It works rather well.<br>This design addresses three failure modes that the usual training and instruction cannot fully fix: hallucination, context issues, and memory issues.<br>💡<br>The other biggest problem<br>Many harms can be caused by AI including death, but we should expect AI to harvest our personal data and use it without consent. That’s why I am interested in things such as offline AI, and on-device small language models, and why I try to use PCE such that the robots in the sky don’t learn any more about me than they already do.
Hallucination and nonsense#<br>Language models generate text according to probability, where the next piece of text (a ’token’) is selected based on patterns, not by retrieving facts from a database. If a model does not have a pool of highly relevant text to select from (the ‘context’) it will probabilistically generate text anyway because that it what it is programmed to do.<br>The result is confabulation, where the model sounds confident while making a false or misleading claim. The better the AIs become at expressing themselves, the<br>more convincing these hallucinations can become.<br>Research keeps concluding ↗<br>that training does not eliminate hallucination, and newer surveys ↗<br>describe hallucinations as potentially “fundamental mathematical inevitabilities inherent to [the model’s] architecture.” The AI companies are trying to solve this by giving better instruction and training, but if hallucination is indeed inevitable then this will never be reliable. I am persuaded the architecture needs to change for AI to become more trustworthy.<br>Context input to a model is called the ‘prior’. A quality prior comprises the best available documents, previous relevant decisions, germane background, and from it AI generates much better output. Just like a human organisation, Artificial Organisations strive to deal with the best quality input documents in order to improve decisionmaking, and to carefully label or even reject guesswork. This is the first structural way we can tackle hallucinations.<br>A second technique is also familiar: have someone else check the work. PCE has an agent called the Corroborator whose only job is to read what the Composer agent wrote, and to verify every claim against the source documents. The Corroborator has the sources right in front of it, so if the Composer invented a claim, the Corroborator will see it is unsupported. Corroborator is unmoved by plausible confabulation, because it is instructed to only accept what can be proven from the sources to hand, including references on the internet if...