Below is our recent interview with Joseph M. Firestone and Mark W. McElroy, Co-Founders of Artificial Epistemics, LLC:
Q: Could you provide our readers with a brief introduction to your company?
A: Artificial Epistemics, LLC (AE) is a pioneering start-up (2026) with proprietary IP that makes it possible for AI models and agents to quality-control their outputs using principles taken from the field of epistemology. The issues it addresses are AI safety and alignment, deep (adaptive) learning, and AI sustainability, the most significant problems in AI today.
Using epistemic tools that minimize the falsity, amorality, and immorality of AI’s words and deeds, the incidence of AI-related misinformation and dangerous behaviors can also be minimized. No other company addresses the risks and consequences of AIs by killing its worst ideas before they kill us (be they about facts or values), and yet this must be done if AI is to be safe and reliable for human use.
Our core product, The Sustainability (Susty) Code, is a game-changing innovation in fact/value Knowledge Processing – a creature of epistemological AI in a category of its own. At present, it takes the form of a detailed specification that allows AIs to: 1) regulate and align their outputs with human interests by minimizing the falsity, amorality, and immorality of their words and deeds, and 2) implement long-standing principles in epistemology, still missing from leading Large Language Models (LLMs) in AI – a blind spot.
Company revenue is generated from IP licensing, training, and consulting fees. Target customers consist of: 1) leading LLM model producers (Anthropic, OpenAI, Google, Perplexity, etc.), and 2) third-party vendors of value-added Agentic AI applications (e.g., “Skill” modules callable from AI harnesses).
Q: What makes you the best choice? How are you unique?
A: AE’s Susty Code is unique in the AI world in a number of ways:
- It is the only specification that has a set of rules for comprehensive knowledge processing. We don’t have the space here to go into detail about the set of knowledge processing rules contained in the Susty Code; but there are 11 of them, each one playing a crucial role in building a multi-faceted system for testing and evaluating existing knowledge, and for producing new knowledge when needed.
- It is the only specification that offers a rule for Fair Critical Comparison (FCC) utilizing multiple criteria for comparing alternative models/theories in order to evaluate their “Truthlikeness” (for facts) and or “Oughtlikeness” (for values) by testing for falsity, illegitimacy, and non-legitimacy. While other organizations may also be developing testing procedures for knowledge claims using “Popperian Control Loops”, none of these products or specifications are nearly as comprehensive as ours. Nor do they even begin to address normative knowledge claims (i.e., for values), which is a central focus of AE’s Susty Code product.
- It is the only specification that requires AIs to systematically rely on falsificationist rather than justificationist thinking when distributing or producing universal or probablilistic knowledge claims. This means that the Susty Code provides an inherent barrier to excessive risk-taking, since the AIs using previously based knowledge would will always know that such previously produced knowledge is fallible; that decisions and actions based upon it carry risk; and that their job is to know and to specify what that risk is.
- It is the only specification that requires AIs to maintain a track record of all new knowledge produced in terms of meta-claims that provide the basis for decisions to refute, or not refute, both descriptive and normative knowledge claims.
Q: Any highlights about your latest announcement?
A: From our Press Release of 09.16.26
Artificial Epistemics, LLC (AE) announced today its latest findings from an ongoing study of prevailing epistemologies (knowledge production functions) in leading large language models (LLMs) – this time involving the specifics of Gemini and Grok. Using a distinction philosophers make between justificationist and falsificationist epistemologies, the AE team found that both LLMs (Gemini and Grok) explicitly rely on justificationist thinking when formulating responses to prompts or plans of action. In so doing, they systematically overstate the correctness of both facts and values and thereby undermine AI safety and alignment by exposing users to the risks and consequences of false or illegitimate information when used.
Our findings today are that two of the leading chatbots, Gemini and Grok, openly rely on the same notoriously fallacious form of justificationist epistemology, whereby the knowledge they produce and/or the actions they take are treated as essentially true or legitimate with certainty, thanks mainly to the authority vested in them by their makers. This more or less ignores the possibility of conflicting/contradictory evidence, and relies instead on the managerial authority of those who own or control AIs to systematically disregard the exceptions. As long as enough supporting evidence can be found, they reason, why bother looking for contradictions?
“Indeed, if what we want in AI are models prone to spreading falsehoods, hallucinations, or rogue behaviors, justificationist epistemologies are perfect for the job. If, on the other hand, what we want are AIs that routinely flag and kill their worst ideas before they kill us, the opposite is true. That is the kind of reform we must have in AI at this time. Why? Because the road to AI safety and alignment must be paved with falsificationism. Nothing less will do!
Q: What can we expect from you in the next 6 months? What are your plans?
A: Much of our work over the next few months will consist of fine-tuning our Susty Code specification so as to cement the inclusion of normative knowledge, and not just facts, within its scope. To do that, we will be building upon our long-standing commitment to Karl Popper’s falsificationist approach for assessing claims about facts by consolidating our thinking on how it can address values as well.
And since unlike facts, value claims are either legitimate, illegitimate, or neutral (and not, per se, true or false), the very name of the falsificationist doctrine itself and what it implies is out of date. We prefer “refutationism”, a term that is equally applicable to both facts and values. We will have more to say about this in the weeks ahead.
Finally, we are still very much in the throes of fund raising and have every expectation that our initial round of seed capital will be secured posthaste. The days of outcome-based reinforcement learning and its flawed grounding in justificationist thinking are clearly numbered, and we (AE) are well-positioned to usher in the vanguard of its reputationist successor – the Susty Code!
Q: What is the best thing about your company that people might not know about?
A: Three things:
- That we are the first and still only commercial player in the field of what many are now referring to as the “Popperian / Falsification Control Loops” segment of the AI industry.
- That the key to resolving the Safety & Alignment problems in AI (misinformation and rogue behaviors) arguably lies in the use of epistemological tools like the Susty Code.
- And that opportunities to partner with and invest in our company are still very much on the table. Leading LLM, “harness”, and Agentic AI developers would do well to cast a glance in our direction. The Susty Code, and other tools like it, have an important role to play in their future!
