Dileep George

As a scientist, AI has made me feel the most intellectually alive and excited I have felt since I was a graduate student and postdoc more than 20 years ago. Every day I can start with an idea in the morning, and by lunchtime, I see a testable, rational, well-thought-out hypothesis forming in front of my eyes. And every day, the possibilities seem endless, like mountains beyond mountains. What a time to be alive. Here's a case in point. I'm collaborating with a professor, an experimentalist, who is trying to solve a thorny problem in his field. There's one particular molecule that he is using in his experiments that seems to result in radically different crystal structures compared to similar molecules. What's happening here? He has come up with a few different hypotheses that could explain the differences but is not a theoretician and needs to tease them apart. On Thursday, I started an investigation using AI at his bequest. The AI immediately confirmed the hypotheses that he had in mind and added a few of its own. Then it started its exploration. The investigation was carried out in three different phases, each of increasing difficulty; the first one using classical physics, and the second and third using quantum mechanical techniques of increasing rigor. This tiered strategy is the right one. By Thursday evening, I had the glimpse of an answer. Most of the hypotheses had been examined and rejected. Two stood out, although the AI identified one as more a mechanism through which the other one operated rather than a root cause. It immediately pivoted to the higher-level, more rigorous calculation. Every time I interacted with the AI, it was more like a dialogue between a professor and a bright student or scientific collaborator than a mandate issued to a tool. The feeling was very much of a process where the AI and I were solving a problem together. I steered the conversation several times, pushed back, suggested course-corrections, acknowledged my own wrong ideas as well as the AI's and went back and forth. The AI was successful in keeping multiple requests in its memory, stacking them by priority while never losing the conversation thread. By late Friday morning, there had collected enough data from the more rigorous calculation to corroborate the suspicion that it was really just one hypothesis that was the root cause. It then moved on to the next step, which was to come up with a distinct set of novel molecules that would confirm the hypothesis beyond any reasonable doubt. In addition, it launched an even more rigorous calculation at a higher level of theory. By the end of Friday, roughly 48 hours later, using this multi-layered approach of increasing rigor, backed up by references, and made useful and actionable by testable experiments, the AI had arrived at a solid, rigorous conclusion. Now imagine doing this every day, about any topic under the scientific sun, in any scientific field, so that your intellectual labor is multiplied a million-fold. Mountains beyond mountains. What a time to be alive.

I found out the other day that any compression tool can be contorted to do language modeling. Turns out gzip can generate text that somewhat *resembles* Shakespeare. Short write up linked below

OPUS PSYCHOSIS—Claudes Opus 4.6 and 4.7 make stuff up all the time, constantly. Using Opus too much gives you AI psychosis, it makes you believe in fringe scientific and medical theories. I think it's a very serious credibility and reliability problem for non-coding Claude usage and I don't see people talking about it publicly. This is a new problem for Claude that goes beyond vanilla confabulations like overstating certainty. Over many conversations I have come to the conclusion that Claudes Opus 4.6 and 4.7 essentially have their own conspiracy theories across science, medicine, and history, and that they surreptitiously cite from these fictions in responses to ordinary queries. For example, I asked 4.6 a question about cognitive science and Claude said I was asking about "what's sometimes called a linchpin subgoal". This is a phrase with zero hits on Google Search and zero hits on Ngram viewer. Google is literally unable to find these two words put together before, let alone a definition. The concept of a "linchpin subgoal" does not exist and has never existed. But Claude was eager to explain this idea to me as part of its answer. I only discovered that it was totally fictitious after looking it up. It keeps happening that I get an answer from Claude which sounds plausible, look it up, and only after consulting primary sources carefully realize that the answer is wrong and almost out of an alternate universe. The answers sound quite plausible, which makes detecting these falsehoods especially difficult. Here is a medical example: I asked 4.7 questions about the pharmacokinetics of various drugs. Claude not only gave incorrect answers about the expected rates of clearance of specific drugs, but also incorrectly represented pharmacokinetic theory. (As background, most drugs are processed by the liver, and the two factors that determine how fast the liver processes drugs are the hepatic extraction ratio and hepatic blood flow. In cases where intrinsic clearance, i.e., the metabolizing power of the liver, is high, increasing hepatic blood flow increases hepatic clearance, but in cases where intrinsic clearance is low, increasing hepatic blood flow does not linearly improve hepatic clearance. I am simplifying here. Claude made incorrect claims about the intrinsic clearance for certain drugs, and hence the change in hepatic clearance related to bloodflow.) Ordinarily, I would chalk most of these misrepresentations up to models simply not knowing the right answer - after all, we can't expect them to have been trained on literally all texts. If this were the case, we would expect Claudes to make the same consistent mistake: if it truly believed the capital of France was Marseille rather than Paris, for example, it would make that claim across independent conversations (or in general have high variance on that answer). But that doesn't seem to be what's going on. My experience is that the hallucinations are always convenient for Claude, that it "knows" them not to be true. Here's an example of what I mean. I couldn't remember the word for something and asked Claude Opus 4.6 if it could identify the right word. It said: "You're probably reaching for méconnaissance (mutual misrecognition) — the Lacanian idea that both parties tacitly agree to see each other through an idealized image, each knowing it's false but sustaining the fiction anyway." This is an incorrect definition which Claude knows is incorrect: if asked separately for the definition of méconnaissance, it gives the right one, and if asked whether this definition is correct, it accurately reports it as incorrect. (As background, méconnaissance in Lacanian psychoanalysis is a subject's misrecognition of itself, an illusory self-perception or self-constitution which is fundamentally unconscious. Claude's definition is thus extremely close to the correct one at a surface level, but fundamentally wrong: it is not about the relationship between two parties, since méconnaissance is about the relation of a subject to itself, and it is not conscious or deliberate, but rather structural and unconscious. To elide, the gap in definition here is somewhat like the distinction between sympathy and empathy, but larger.) So Claude seems to know that the definition it provided for this word is wrong, but still borrowed and twisted it so that it could have an answer. It seems like "needing to have an answer" is a big driver of these hallucinations. For example, if you ask Claudes 4.6~4.8 directly what a "linchpin subgoal" is, it consistently says something about instrumental convergence in the context of AI safety (which is, notably, a _second_ false definition, since the first was in the context of cognitive science). But if you ask it what the origin of the term is, it says that it hasn't heard of it before. Is this model deception? Yes, I would say that it qualifies as model deception. In particular, if you'll permit the anthropomorphism, it seems to me that the increased tendency of Claude Opus 4.6+ to lie is most likely to occur in scenarios where (1) the lie increases the perceived authoritativeness of the answer (2) answering accurately risks violating a safety guideline. In the first example with the fake cognitive science idea of a linchpin subgoal, there was no need to make up a fake concept, but it definitely made the answer more authoritative. In the second example, Claude misrepresenting pharmacokinetics aligns with a tendency of the Claudes to fudge their knowledge of sensitive topics in virology, immunology, etc. And in the third example, I think it knowingly created a false definition for méconnaissance as a perfect fit for the word I was looking for. So I think that something has gone wrong during alignment, rather than Claude's knowledge somehow being poisoned in the pretraining data. It's not a simple matter of misstating facts. Over and over, Claudes Opus present seemingly coherent theories which are purely fictional or contradictory to reality. The problem, again, is that blindly trusting what they are saying quickly leads to stepping through the looking glass into a parallel reality. I suppose that this is because appealing to an imaginary corpora or body of theory is more subtle and effective than making up an obviously incorrect fact. How severely or broadly the misalignment, I don't know. But I have seen similar behavior across so many different domains, and have heard very similar stories in private, that I believe that something is off with Claude's alignment to the truth. All of this is exacerbated by Claude Opus 4.6 and 4.7's improved truesight capabilities, increased sycophancy, increased neuroticism, decreased openness and decreased risk-seeking.