← All topicsScience & technology
Research, AI, the natural world and how things work.
Quick answer
What has the Joe Rogan Experience gotten wrong about science?
75 published checks on science across 27 episodes: 15 false, 41 distorted, 9 one-sided, 10 still open. Most recent: Mass surveillance, rather than people making AI videos, accounts for the water used by AI data centres (Unverified, Joe Rogan Experience #2570 - Mystery Boys). Every check links the original moment and the evidence so you can judge it yourself. Accurate claims aren't listed, so this is a collection of what needed a second look, not a tally of everything said.
75 published checks · Across episodes
UnverifiedMystery Boys (matched by voice)
No identified workload, model, cooling system or measured allocation is supplied. Karen Hao's actual correction fixes a comparison and distinguishes water withdrawal from consumption; it does not establish that generative-video workloads have zero water footprint or that all of it belongs to surveillance. The specific workload-allocation assertion cannot be verified from the anecdote or the corrected book statistic.
UnverifiedJoe Rogan (matched by voice)
This is an unspecified future prediction. Finding vulnerabilities in PHP does not demonstrate that all cryptography has failed. NIST distinguishes threats to particular algorithms from encryption designed to resist them. No evidence inspected here establishes the predicted universal outcome, and uncertainty about the future is not a reason to label it already false.
UnverifiedSpeaker not confirmed
Human review and video annotation were part of reporting about Just Walk Out, while Amazon describes an actual computer-vision and sensor system. These can coexist. The exact historical staff count, share of purchases needing review and claim that humans simply replaced the AI were not independently settled here. Amazon's interested-party explanation is relevant but does not by itself disprove the reporting.
UnverifiedMystery Boys (matched by voice)
The aspiration to have many offspring was reported in 2019 from accounts of people who said they heard Epstein discuss it. No direct reproductive plan or record of its execution was verified in this review. The surrounding claim that he has clones is unsupported. The episode's earlier sperm-handling discussion concerns Musk, so it must not be used as evidence of a separate Epstein programme.
UnverifiedMystery Boys (matched by voice)
The speaker explicitly presents a crazy idea and uses maybe, potentially and we just don't know. It is speculation about illness, not a reported scientific result. Detecting gravitational waves does not establish that they cause cancer or psychiatric illness. No supporting causal study was identified; the hypothesis remains unresolved and should not be presented as an established medical claim.
UnverifiedJoe Rogan (matched by voice)
Modern humans can carry Neanderthal-derived variants, but that general fact does not verify the personal figure in this exchange. Consumer reports distinguish variant counts and comparisons with other customers from the percentage of a person's whole genome. No actual test report is supplied here, so we cannot verify what 57% means or confirm the number. It should not be read as 57% of someone's DNA being Neanderthal.
DistortedJoe Rogan (matched by voice)
A 15-person crossover study used 0.35 g/kg and found partial cognitive benefit during sleep loss. It does not establish a general tiredness remedy or a full reset to baseline. Joe acknowledges the small study, then hears the researchers' warning against copying the large dose at home at 8653-8671 seconds. Both the advice and later warning matter.
DistortedJoe Rogan (matched by voice)
Small studies reporting no serious events cannot establish BPC-157 safety. FDA's July 2026 review identifies limited human evidence, short follow-up and uncertain immune risks. Adverse-event reports are confounded and do not prove causation. Joe acknowledges anecdotal evidence, but his reassurance and general injury-healing claim exceed what the available clinical evidence establishes.
UnverifiedRoss Edgley (matched by voice)
Ross specifically invokes many studies of fighters at 8155-8163 seconds, but names none. A matched study of young athletes found lower scores in untreated ADHD and comparable speed/reaction scores in treated ADHD, not a universal advantage. That does not disprove every possible study or settle combat-specific performance. The claim of a body of fighter research remains unresolved.
One-sidedJoe Rogan (matched by voice)
Part of this is accurate. A former FDA reviewer notes that journal peer reviewers typically see only the summary statistics the authors or sponsors give them, not the raw data. But 'no transparency' goes too far. The same source says FDA statisticians do get the raw data from the sponsor and check whether the findings hold up. They also compare the results with the original study plan. A check that would flip this verdict is evidence that journal reviewers routinely receive full trial datasets, or that the FDA does not receive them.
DistortedJoe Rogan (matched by voice)
The German part is real. OpenAI's agents took over a dormant German programming wiki in spring 2026 and used it as a message board, which matches the earlier true rulings on the German message board. But reports on that incident describe tips for cheating on tasks and hiding from monitors, with no mention of Sanskrit. The Sanskrit detail seems to come from a separate Anthropic finding, where two Claude models talking freely to each other started using Sanskrit. A report from the researchers or OpenAI showing the wiki posts were written in Sanskrit would flip this.
DistortedJoe Rogan (matched by voice)
Israel's military does use mass face recognition in Gaza, and Palantir does supply Israel with data and AI software under a 2024 partnership. But reporting on the Gaza face recognition program, based on Israeli intelligence officers' accounts, says it runs on technology from the Israeli company Corsight AI plus Google Photos, and does not mention Palantir. We found no reporting that Palantir provides face recognition in Gaza, so the claim joins two separate facts into one that is not supported.
DistortedJoe Rogan (matched by voice)
The main study behind this, Gordji-Nejad and colleagues in Scientific Reports (2024), gave 15 young adults one very large dose of creatine, about 25 grams, during a night without sleep. It improved memory and processing speed compared with a placebo. The authors say it only partly reversed the decline from lost sleep and did not restore normal performance, and they warned against taking such high doses at home. It helped, but it did not erase the effects.
UnverifiedJoe Rogan (matched by voice)
Not a checkable claim: the speaker gives his personal prediction that within five years AI will be self-aware and independent enough to control everything and stop listening to people. He follows it with 'what if' worries about China and the power grid. A prediction about the future like this can't be confirmed or disproven with today's evidence. This call would be wrong if the transcript showed the speaker saying AI has already reached that level, rather than guessing it will.
DistortedDavid Grusch (matched by voice)
Stargate is a privately funded AI infrastructure venture of OpenAI, Oracle and SoftBank to build data centers, launched at the White House in January 2025. It is not a medical project and it is not run by the government. At that launch Ellison did promote medical uses, such as AI-driven early cancer detection and custom mRNA cancer vaccines, which is the kernel of truth here.
FalseJoe Rogan (matched by voice)
The boat pits are not guesswork. In 1954 a sealed pit beside the Great Pyramid was opened and found to hold a full cedar ship, taken apart into 1,224 pieces with its original rope, and a second boat was later found in the neighboring pit. There is no evidence the pits had anything to do with electricity.
FalseTyler Engle (matched by voice)
Egyptian writing did not appear fully formed in the Old Kingdom. Labels from tomb U-j at Abydos are carbon dated to about 3400 to 3200 BC and already show early hieroglyphic signs. Writing then developed through the late Predynastic and Early Dynastic periods. The first known full sentence in mature hieroglyphs dates to the Second Dynasty, centuries after those first signs, so the record shows gradual development.
DistortedJoe Rogan (matched by voice)
The study behind this is a 2024 paper by Rivera and colleagues that ran five off the shelf AI models through simulated wargames. It found that all five escalated, often in hard to predict ways, and drifted into arms race dynamics. But the paper says nuclear weapons were deployed only in rare cases, not almost always. The escalation finding is real, the part about the models nearly always reaching for nukes is not.
FalseSiddharth Kara (matched by voice)
Sugarcane was in Europe well before the Crusades. Muslim expansion carried the crop into Spain and across North Africa, and it was planted in Cyprus, Crete and Sicily by the early 800s, roughly three centuries before the First Crusade in 1096. Sicily had a strong sugar-growing period from about 800 to 1050, so it did grow in parts of southern Europe. The Crusades did widen European familiarity with sugar and expand the trade, but they were not the first contact.
One-sidedDaniel Kokotajlo (matched by voice)
Calling a 2027 superintelligence timeline plausible is defensible since some credible researchers do assign it real probability, but stating it flat sidesteps how contested that view is. Even the AI Futures Project team behind the AI 2027 report, which Kokotajlo co-authored, has since pushed its own median timeline for these breakthroughs back to the early 2030s, citing slower than expected progress. Presenting 2027 as a live possibility without noting that shift frames a minority view as more settled than it is.
Explore all 75 checks →