The hottest Bias Substack posts right now

And their main takeaways
Category
Top U.S. Politics Topics

The Double-Edged Sword of GPT Models: Navigating Developer Activism and Bias

Rod’s Blog • 39 implied HN points • 28 Feb 24
🕹 Technology Bias
  1. GPT models have revolutionized natural language processing, opening new opportunities in technology and communication.
  2. Developer activists have been exploiting GPT models for various reasons, like gaining unauthorized access to APIs, which raises ethical questions.
  3. The power of GPT models comes with significant responsibility to ensure appropriate use and prevent potential misuse.

Me vs Huemer

David Friedman’s Substack • 341 implied HN points • 13 Feb 24
📖 Philosophy Bias
  1. Consider forming opinions on controversial issues based on evaluating arguments rather than just trusting the experts
  2. Experts may not always have expertise in all aspects of an issue, so it's important to critically evaluate their arguments and not just rely on their authority
  3. It's crucial to judge both arguments and arguers, as bias and incentives can influence the opinions of experts in controversial topics

Google Gemini’s Woke Catechism

From the New World • 301 implied HN points • 23 Feb 24
🕹 Technology Bias
  1. Google's Gemini AI model displays intentional ideological bias towards far-left viewpoints.
  2. The Gemini paper showcases methods used by Google to create ideological biases in the AI, also connecting to Biden's Executive Order on AI.
  3. Companies, like OpenAI with GPT-4, may adjust their AI models based on public feedback and external pressures.

Beware AI in the classroom

CIEO • 78 implied HN points • 11 Jul 23
🚌 Education Bias
  1. AI in the classroom can provide benefits but also comes with costs, such as bias and misinformation.
  2. AI-generated responses may not always be accurate or politically neutral, reflecting biases of developers.
  3. To effectively judge AI responses, individuals need critical thinking skills and knowledge, and teachers play a crucial role in guiding young people.

Quantifying ChatGPT’s gender bias

AI Snake Oil • 523 implied HN points • 26 Apr 23
🕹 Technology Bias
  1. Researchers found strong gender bias in ChatGPT models despite correct benchmark data
  2. Bias examination focused on coreference resolution to identify gender bias
  3. GPT-4 showed slight improvement over GPT-3.5 in gender bias accuracy
Get a weekly roundup of the best Substack posts, by hacker news affinity:

Why You Always Get Punished for Telling The Truth

OK Doomer • 124 implied HN points • 03 Dec 24
🎭️ Culture Bias
  1. People who tell the truth often get punished for it. This happens because society tends to ignore uncomfortable truths.
  2. Cognitive biases make it hard for people to accept negative feedback or truths, leading them to blame those who speak out.
  3. Pretending everything is fine, like the woman who followed Zig Ziglar's advice, doesn't solve the real issues and can create more problems.

Weight Stigma In Practice - The Gynecologist Should Not Be A BYOS Party

Weight and Healthcare • 299 implied HN points • 19 Feb 22
🏥 Health & Wellness Bias
  1. Fat patients often face weight stigma in healthcare, being denied proper equipment and care based on their body size.
  2. Healthcare facilities should provide equitable care for patients of all sizes, including having necessary equipment like speculums and blood pressure cuffs.
  3. Fat patients deserve the same quality of care as thinner patients, without discrimination based on their body size.

The philosophical and empirical cases for ignoring bad or biased pollsters | #210 - February 26, 2023

G. Elliott Morris's Newsletter • 158 implied HN points • 27 Feb 23
🇺🇸 U.S. Politics Bias
  1. Aggregating public opinions may mean slightly less accurate election forecasts, which is acceptable.
  2. Polls dramatically overperformed expectations in the 2022 midterms, despite popular misconceptions.
  3. It's important to be cautious of biased pollsters and consider the methodology, credibility, and track record of pollsters before including their data in models.

When Your Map Doesn't Match Reality

Good Reason • 227 implied HN points • 13 Dec 23
📖 Philosophy Bias
  1. Regardless of how well you know a situation, remember your knowledge is just a map and not reality itself.
  2. Be cautious of projecting your biases onto situations to force them to fit your preconceived notions.
  3. Acknowledging and being aware of your own potential biases can help prevent misunderstandings and misinterpretations.

How Implicit Biases Compound for Aspiring Executives

The Uncommon Executive • 19 implied HN points • 07 Mar 24
💼 Business Bias
  1. Affinity bias, confirmation bias, and negativity bias combined create a self-fulfilling loop that can hinder career progression for minorities.
  2. Implicit biases often show up in actions not taken or opportunities not given, making them challenging to identify and address.
  3. Small biases, left unchecked, can compound over time and lead to lack of diversity at the executive level, making it crucial for both current and aspiring leaders to actively combat biases.

Why Making a Non-Woke AI Is Actually Very Hard

The Future of Life • 19 implied HN points • 26 Feb 24
🕹 Technology Bias
  1. Language models learn from the data they are trained on, which often includes a lot of left-leaning content, making them reflect that bias.
  2. Adjusting a model's political views is complicated because it involves changing an entire worldview, which can mess up the quality of the responses.
  3. Creating a balanced AI requires new training methods, as current models can’t easily switch perspectives without losing their effectiveness.

Biases Against Risky Research

New Things Under the Sun • 224 implied HN points • 31 Mar 23
🔬 Science Bias
  1. Scientific institutions may be risk-averse and favor safe and incremental projects over transformative ones.
  2. Individual reviewers and averaging peer review scores may bias against high-risk, high-reward research proposals.
  3. In grant review processes, negative feedback tends to be more influential than positive feedback, leading to potential bias against novel research.

Gloves Off

The Radar • 19 implied HN points • 27 Dec 23
💼 Business Bias
  1. Old ideas and worn-out concepts in talent management must be identified and discarded to allow for genuine progress.
  2. Binary labels like 'hard skills vs soft skills' and 'introvert vs extrovert' are misleading and can hinder accurate talent assessment.
  3. The concept of 'high potential candidate' often introduces bias and leads to poor decision-making, hindering talent development and organizational growth.

The New Expanded Illustrated Orwelexicon: DSM 666

Unsafe Science • 48 implied HN points • 14 Nov 24
🔬 Science Bias
  1. The New Expanded Illustrated Orwelexicon introduces new words to highlight biases and issues in psychology and academia. It aims to make these concepts clearer and more accessible.
  2. This resource shows how language can shape our understanding of societal problems. By using specific terms, it helps pinpoint dysfunctions within various fields.
  3. Overall, the Orwelexicon encourages critical thinking about psychology and social science. It invites people to reflect on how words influence perceptions and beliefs.

Explaining Anomalous GCSE Results

Cremieux Recueil • 90 implied HN points • 21 Feb 24
🚌 Education Bias
  1. Some Black African students in the UK perform better on GCSE exams than on IQ tests, leading to a puzzle that is not easily explained by sampling differences.
  2. The discrepancy between GCSE performance and general intelligence indicates that GCSEs may be biased indicators of intelligence, favoring certain groups over others.
  3. Psychometric bias might explain why Black Africans in the UK excel in GCSEs despite IQ tests suggesting otherwise, highlighting the presence of biases in assessment methods.

A Case for Moderate IQ Skepticism

Tom Thought • 19 implied HN points • 31 Oct 23
🔬 Science Bias
  1. Performance on cognitive tasks is correlated, so individuals who excel in one area tend to excel in others as well.
  2. IQ tests are useful in predicting various life outcomes, but it's important to recognize that they are not a direct measure of intelligence.
  3. It's crucial to be skeptical of assigning deep meaning to specific IQ scores, especially when comparing across different populations.

Merit First, Diversity Second

Kartick’s Blog • 35 implied HN points • 22 Nov 24
🇺🇸 U.S. Politics Bias
  1. Merit should come first when hiring or building teams. If you focus only on diversity without considering skills, you may end up with less qualified candidates.
  2. It's important to recognize and address our own biases. We should judge people based on their abilities rather than their background.
  3. While diversity and inclusion efforts can be positive, they can also go too far and lead to unfair quotas. It's better to promote a merit-based system instead of rigid percentages.

Does the kids COVID vaccine work? Does the XBB1.5 booster work? Test negative case control studies are prone to bias

Vinay Prasad's Observations and Thoughts • 77 implied HN points • 06 Feb 24
🏥 Health & Wellness Bias
  1. Test negative case control studies may not be reliable for assessing vaccine effectiveness
  2. Consider the biases present in test negative case control studies
  3. To assess vaccine efficacy, randomized trials are crucial and should be prioritized over test negative case control studies

Pawel about Pawel

Pav writes • 19 implied HN points • 03 Jun 23
🚌 Education Bias
  1. Paweł values distilling knowledge and brevity in writing
  2. He aims to overcome cognitive biases and prioritize precision in arguments
  3. Paweł's writing focuses on simplicity and communicating complex ideas in an understandable way

My Cave of Echoes

DruGroup • 39 implied HN points • 14 Dec 21
📖 Philosophy Bias
  1. It's important to be friends with people who are different from you. Learning from their experiences can help expand your thinking.
  2. Listen more and ask questions when talking to others. Understanding their views can help you see where they are coming from.
  3. Try to consider other viewpoints even if you don't agree. It helps avoid misunderstanding and can make your own beliefs stronger.

How AI becomes biased, visually explained 🐧

Year 2049 • 8 implied HN points • 17 Jan 25
🕹 Technology Bias
  1. AI can show bias based on how it learns from the data given to it. If the data contains biases, the AI will likely reflect those biases in its decisions.
  2. Using simple examples, like a penguin metaphor, helps explain complex AI concepts. It's easier to understand difficult ideas with relatable stories.
  3. It's important to be aware of AI bias as it affects how AI technologies interact with people. Being educated about these biases can lead to better, fairer AI development.

Reassessing the Confirmation Bias

Optimally Irrational • 11 implied HN points • 14 Jun 23
🚌 Education Bias
  1. Confirmation bias is a widely acknowledged cognitive bias where we tend to seek information that supports our existing beliefs.
  2. In today's world of political polarization, confirmation bias contributes to escalating tensions as people isolate themselves in echo chambers.
  3. Seeking confirmatory information aligns with an optimal information acquisition strategy, especially when considering costs and efficiency.

Decoupling decoupling from rationality

Unconfusion • 1 HN point • 07 Jan 24
📖 Philosophy Bias
  1. Some people think differently about ideas, separating them from context, while others mix context with their thoughts. This difference can help explain disagreements between people.
  2. Rational thinking involves a careful process where we slow down our automatic thoughts and consider ideas more deeply. This is important to avoid mistakes and biases in our reasoning.
  3. The term 'decoupling' has changed over time and now it sometimes refers to a personality trait rather than a careful thinking process. It's important to understand this shift to better discuss rationality and disagreements.

Artificial Intelligence: What to Worry About

Prompt Engineering • 1 HN point • 11 May 23
🕹 Technology Bias
  1. AI advancements are fast and significant, leading to uncertainties about the future models of society.
  2. AI may result in mass unemployment, but historically, technology revolutions have not led to catastrophic unemployment levels.
  3. Challenges with AI include the misuse of data leading to bias, the potential for AI to outsmart humans, and the widening class divide due to unequal access to AI tools.

How a Study Can Show Something to Be True When It’s Completely False — Regression to the Mean

Harnessing the Power of Nutrients • 59 implied HN points • 01 Apr 11
🔬 Science Bias
  1. Studies can mistakenly show false results as true due to regression to the mean, a common phenomenon in research.
  2. Research results can be influenced by statistical artifacts like regression to the mean, highlighting the importance of critical evaluation of study data.
  3. Proper randomization is crucial in research to avoid misleading results caused by regression to the mean.

GPT is Rather Good at Feed Ranking

Messy Progress • 3 HN points • 07 Mar 23
🕹 Technology Bias
  1. GPT makes content-based feed ranking easy and has the potential to shift ranking power to users and groups.
  2. The ChatGPT API simplifies the process of creating content-based ranking models, making it more accessible and efficient.
  3. Using large language models like GPT to generate labels for training small models can lead to practical and cost-effective approaches in content-based ranking.

Confirmation Bias: A Bayesian Interpretation

In My Tribe • 1 HN point • 28 Feb 24
🚌 Education Bias
  1. Having a strong prior belief is fine, but bias comes in when one refuses to consider evidence against that belief.
  2. Using Bayesian reasoning means weighing new evidence against what you believed before, termed your 'prior.'
  3. Bias occurs when someone puts a negative weight on new information, ignoring evidence that contradicts their prior beliefs.

AI: The Bias Paradox

Mind Prison • 2 HN points • 15 Feb 23
🕹 Technology Bias
  1. AI reflects human flaws and biases, it doesn't fix them
  2. Trying to create unbiased AI is a paradox, as humans introduce bias during refinement
  3. Accepting bias in AI is crucial for understanding its limitations and ensuring transparency

Platform Power and Platform Bias in the Music Industry

Platform Papers • 0 implied HN points • 16 Mar 22
🎶 Music Bias
  1. Large digital platforms like Spotify can influence the success of songs and artists by controlling playlists and exposure.
  2. Appearing on Spotify's popular playlists can lead to significant increases in streams and commercial success for artists.
  3. While Spotify has the power to impact success, there are indications of bias favoring independent-label music and music by women, raising questions about fair treatment in the music industry.

Lies, Damn Lies, and Susie Neilson

Fava’s Substack • 0 implied HN points • 16 Apr 23
🇺🇸 U.S. Politics Bias
  1. Journalistic objectivity is being questioned and data is emphasized over feelings.
  2. Susie Neilson's reporting is criticized for alleged biases and questionable data manipulation.
  3. Accuracy in reporting data is essential, as misinterpretations can impact conclusions and public perception.

We'll never let go, until we do

Links I Would Gchat You If We Were Friends • 0 implied HN points • 29 Feb 16
🕹 Technology Bias
  1. The FBI wants access to our minds, not just our phones, according to the 'extended mind hypothesis.'
  2. Netflix recommendations show racial biases - defaulting to 'white' movies until you indicate preference for diversity.
  3. Mark Zuckerberg's grey T-shirt uniform is an unexpected lifestyle choice, suggesting simplicity and consistency.

GPT Bias: A Challenge for Responsible AI

Rod’s Blog • 0 implied HN points • 27 Feb 24
🕹 Technology Bias
  1. GPT models can inherit and amplify biases from the data they are trained on, leading to negative impacts like misinformation and discrimination.
  2. GPT bias stems from both data bias (issues with the training data) and model bias (issues with the model design and architecture).
  3. There have been advancements in GPT models over the years, with newer versions like GPT-4 implementing techniques to reduce biases compared to earlier versions.

RDEL #38: Does AI help or hurt gender diversity in tech hiring?

Research-Driven Engineering Leadership • 0 implied HN points • 15 Apr 24
🕹 Technology Bias
  1. AI in recruitment can increase gender diversity in tech hiring by improving the completion rate of job applications among women.
  2. Providing AI-generated scores alongside applications can reduce gender bias in evaluations and lead to more equal assessment of male and female candidates in tech hiring.
  3. Although concerns exist about bias in AI tools, research shows that AI can outperform human bias in the hiring process, offering a promising solution to create a more diverse and qualified tech workforce.

Position bias in features

Simplicity is SOTA • 0 implied HN points • 12 Feb 24
🔬 Science Bias
  1. Position bias can affect the inputs of machine learning models when features reflect prior user behavior, leading to biased estimations of relevance.
  2. Using inverse propensity weighting (IPW) like IPW-CTR can help mitigate position bias in features, but it can result in high variance due to dividing by small numbers.
  3. The choice of weights to measure position bias is crucial, as observed click propensities may overestimate the bias, impacting the performance of features designed to address bias-variance trade-offs.