Latest Corrections
LessWrong May 27, 2026 at 06:13 PM
www.lesswrong.com/posts/qRWRjDQdBWRZFBxTv/standard-deviation...
- This uses the wrong distribution and the wrong factor. Student’s t adjusts confidence intervals for the mean when σ is unknown; for σ itself, the standard small-sample distribution is chi-square, and with two observations the usual unbiased estimate is about 0.886× the distance, not 1.3×.
- The 1.3× rule is not a correct way to estimate standard deviation from two observations. For two values 43 and 47, the usual unbiased estimate of σ is about 0.886×4 = 3.54 litres, not about 5 litres.
LessWrong May 27, 2026 at 06:08 PM
www.lesswrong.com/posts/xJWBofhLQjf3KmRgg/four-ways-learning...
- Acemoglu’s paper does discuss more-transformative future AI, including AGI, and explicitly says its estimates are conditioned on a particular task-level framework. So it is inaccurate to say this assumption is made 'without any discussion.'
LessWrong May 27, 2026 at 06:07 PM
www.lesswrong.com/posts/DiiLDbHxbrHLAyXaq/why-people-like-yo...
- The linked 2015 report was issued by the Blue Ribbon Study Panel on **Biodefense**, not on Biosecurity.
LessWrong May 27, 2026 at 02:47 PM
www.lesswrong.com/posts/4uQxZonCwCZtz39Hw/timeline-of-carl-s...
- The paper’s title is misstated. Contemporary and later sources list it as “How Hard is Artificial Intelligence? …”, not “How Hard is Intelligence? …”.
Substack May 27, 2026 at 01:27 PM
www.interconnects.ai/p/notes-from-inside-chinas-ai-labs
- This blanket claim is inaccurate because OpenAI explicitly advertises internships and had internship/co-op roles posted around the time this article was published.
Wikipedia May 27, 2026 at 07:40 AM
de.wikipedia.org/wiki/Schulthei%C3%9F#Bedeutung_und_Funktion...
- Die Familiennamen Heiß/Heiss gelten in der deutschen Namenforschung nicht als Varianten von Schultheiß. Sie werden vor allem als eigene Namen erklärt, meist als Patronyme zu Mattheis/Matthias.
Wikipedia May 27, 2026 at 07:31 AM
de.wikipedia.org/wiki/Rechtswissenschaft
- Die Zahl ist zu hoch. Nach Angaben der Bundesregierung galten auf Bundesebene zum Stichtag 24. Mai 2024 insgesamt 1.797 Gesetze und 2.866 Rechtsverordnungen, also zusammen 4.663 und nicht mehr als 5.000.
Wikipedia May 27, 2026 at 07:26 AM
en.wikipedia.org/wiki/B%C3%BCrgerliches_Gesetzbuch
- For movable property under § 929 BGB, ownership normally passes only with both agreement and delivery; the transfer agreement alone is not generally enough.
Wikipedia May 27, 2026 at 02:15 AM
en.wikipedia.org/wiki/A_Monument_to_Something_That_Never_Hap...
- The monument was unveiled on 29 August 2010, not 28 August 2010.
- Sources say the date 28.03.1983 was intentionally arbitrary, not the specific moment of a near-disaster.
Wikipedia May 27, 2026 at 01:41 AM
en.wikipedia.org/wiki/Maggi_Rubenstein
- This sentence wrongly places SFSI’s founding in 1976. Multiple sources, including SFSI itself, say the organization was founded in 1972, several years earlier.
Wikipedia May 27, 2026 at 01:40 AM
en.wikipedia.org/wiki/Mainz
- This is outdated. Mainz’s most recent city council election was held in 2024, not 2019.
- This population figure is incorrect for December 31, 2024. Official sources put Mainz above 224,000 residents on that date.
Wikipedia May 27, 2026 at 01:38 AM
en.wikipedia.org/wiki/Yakub_(Nation_of_Islam)
- This reverses the Genesis account. In Genesis 30, Jacob’s wages are the dark lambs/sheep and the speckled or spotted goats, not Laban’s.
Wikipedia May 27, 2026 at 01:37 AM
en.wikipedia.org/wiki/A_Monument_to_Something_That_Never_Hap...
- The unveiling date is off by one day. Multiple Tarnów sources place the unveiling/opening on 29 August 2010, not 28 August 2010.
- The monument's opening/unveiling date was 29 August 2010, not 28 August 2010.
LessWrong May 27, 2026 at 01:25 AM
www.lesswrong.com/posts/ytzrakjgcvCfLCCZp/contra-wentworth-o...
- This is incorrect because there are known medical interventions that can increase adult height, most notably limb/stature-lengthening surgery.
LessWrong May 27, 2026 at 12:27 AM
www.lesswrong.com/posts/Lid8cv9qF7nac53va/your-left-brain-do...
- This overstates how frontier models are trained. Official documentation describes curated mixtures of filtered web data plus other datasets, not 'the whole of the internet.'
- Specialized AI models plainly do exist. AlphaFold is built for protein structure prediction, and GraphCast is built for weather forecasting; both are domain-specific systems rather than general frontier LLMs.
LessWrong May 27, 2026 at 12:18 AM
www.lesswrong.com/posts/xsDWd7e2yrPdtXMSu/what-am-i-if-not-a...
- The public rollout files linked by the post do not show the same person across every Mistral run. They contain multiple different names, genders, countries, and biographies.
LessWrong May 27, 2026 at 12:13 AM
www.lesswrong.com/posts/FDxvpznjMzGNHjDRS/brackets-are-a-bad...
- Pickpocketing could carry a death sentence, but execution was not automatic. In practice, many capital convicts were pardoned, and Old Bailey sources say pickpocketing had ceased to lead to executions there by the later 18th century.
LessWrong May 26, 2026 at 09:15 PM
www.lesswrong.com/posts/GbWwesBnetyiomxEH/many-portions-of-m...
- This misstates the paper’s metrics. Jabarian and Imas report Pangram had near-zero false positive and false negative rates; a near-zero false positive rate implies a near-100% true negative rate, not under 5%.
- This misreads the reported metric. In the cited paper, Pangram’s false positive rate is 0.0000 and its true positive rate is 0.7791; a 0% false positive rate implies a true negative rate of 100%, not 28%.
LessWrong May 26, 2026 at 09:08 PM
www.lesswrong.com/posts/pGvyqAQw6yqTjpKf4/the-gift-we-give-t...
- This overgeneralizes pollination. Many flowering plants are not pollinated by bees at all: some are pollinated by wind or water, and many others by animals such as birds, butterflies, moths, flies, beetles, wasps, and bats.
X May 26, 2026 at 02:44 PM
x.com/ednewtonrex/status/2059174131665391802
- The encyclical does mention health and personal data, but most of its references to "data" are broader discussions of digital power, governance, access, and ownership.