The hottest Accuracy Substack posts right now

History doesn't always self-correct quickly, with myths and false facts persisting for decades.
Retractions do happen in history, showing some correction in the field.
Openness in history, like 'Open History' initiatives, can help correct the record and benefit historians.

Prediction markets offer straightforward incentives for accuracy and can untangle true beliefs from stated beliefs.
Success in prediction markets is based on skill in making accurate probabilistic forecasts, not luck.
Prediction markets can aggregate useful information, potentially outperforming polls in predicting outcomes.

Most wrist blood pressure monitors lack validation for accuracy.
Omron's HeartGuide wrist BP monitor was found to be inaccurate.
Cuffless wearable devices may hold promise for future accurate blood pressure monitoring.

Normalization of past disaster losses involves estimating damage from past events based on current societal conditions.
Using incorrect data sources can lead to significant errors in research and incorrect conclusions.
Maintaining scientific integrity and ensuring data accuracy are crucial in climate science research.

PacBio's raw accuracy has only slightly improved over 4 generations of instruments.
Fundamental issues, like polymerase kinetics, may be limiting the accuracy of PacBio sequencing.
PacBio has prioritized longer read lengths over accuracy, using cyclic consensus sequencing for improved results.

Get a weekly roundup of the best Substack posts, by hacker news affinity:

Artificial Intelligence advancements are causing concern about potential job losses
AI technology like ChatGPT can generate content quickly and easily, threatening certain professions
Accuracy and credibility of generated content by AI bots still remain a challenge to overcome

We pay for entertainment in news, not always accuracy.
Trust in media is often shaped by how they cover big, impactful events.
Reader trust is not based on daily news but on accuracy during significant moments.

When testing AI chatbots with questions on Finnish history, the farther the topic is from well-known subjects, the worse the AI performs.
Comparing OpenAI's ChatGPT with Microsoft's Bing, Bing may require more specific prompts to provide detailed answers.
AI chatbots like ChatGPT and Bing may offer inaccurate or hallucinated information, highlighting the importance of fact-checking and verifying information when using AI for historical research.

Drive time isochrones can give a false sense of precision because they are based on average traffic conditions, which can vary greatly.
Improving isochrone accuracy can be done by increasing the number of trips, testing different conditions, and varied departure angles.
To better communicate uncertainty, consider simplifying isochrone shapes and creating bands to show the range of possible outcomes.

The quote 'Information wants to be free' originated in a conversation between Stewart Brand and Steve Wozniak in 1984, highlighting the conflicting views on the value and accessibility of information.
The idea that information should be free, but the time spent on it should not, underscores the balance between information's economic worth and the scarcity of attention.
The original transcription of the conversation revealed that the term 'information wants to be free' was not accurately spoken by Brand, with the qualifier 'almost' altering the context of the statement.