Don't Worry About the Vase

A world made of gears. Doing both speed premium short term updates and long term world model building. Currently focused on weekly AI updates. Explorations include AI, policy, rationality, medicine and fertility, education and games.

The hottest Substack posts of Don't Worry About the Vase

And their main takeaways

Childhood Roundup #3

1164 implied HN points • 10 Oct 23
  1. Parenting doesn't require giving children an unhappy childhood to succeed.
  2. Restorative justice in schools may not always lead to positive outcomes.
  3. Homework may not serve a valuable purpose and can be detrimental to children's well-being.

GPT-4 Plugs In

1747 implied HN points • 27 Mar 23
🕹 Technology AI Internet Privacy Security Programming
  1. GPT-4 is getting an upgrade with plug-ins for browsing the internet and using various websites.
  2. Concerns about safety and risks involved in using these new plug-ins have been raised.
  3. The introduction of plug-ins makes it easier to interact with GPT-4, but also raises questions about trust and potential misuse.

On OpenAI Dev Day

851 implied HN points • 09 Nov 23
🕹 Technology AI GPT-4
  1. GPT-4 Turbo brings big improvements like knowledge up to April 2023 and cheaper prices.
  2. Function calling updates allow multiple functions in one message, improving accuracy.
  3. New modalities like GPT-4 with vision and DALL·E 3 offer exciting possibilities for developers.

Childhood Roundup #2

1568 implied HN points • 11 Apr 23
🚌 Education Parenting School Learning Admissions
  1. Child care is becoming more regulated and expensive, making it challenging for parents to afford quality child care.
  2. Parents are facing challenges in allowing their children to play and be independent due to strict regulations and societal fears.
  3. The education system is facing criticisms for ineffective techniques, pushing unnecessary pressure on students, and focusing more on signaling than actual education.
Get a weekly roundup of the best Substack posts, by hacker news affinity:

On the FLI AI-Risk Open Letter

1523 implied HN points • 30 Mar 23
🕹 Technology AI Regulation Ethics Political Response Risk Assessment
  1. The FLI AI-Risk Open Letter calls for a pause in training AI systems more powerful than GPT-4, sparking mixed reactions.
  2. The letter combines short-term concerns with existential risks, suggesting high standards that might act as a de facto ban on AI development.
  3. Despite its flaws, the letter sets the stage for cooperation and lays the groundwork for future discussions around AI safety and regulation.

On OpenAI's Preparedness Framework

492 implied HN points • 21 Dec 23
🕹 Technology AI Ethics
  1. Preparedness framework is a positive step, but needs more details.
  2. Critical thresholds seem too high, especially in autonomy.
  3. Framework relies on honoring the spirit, not just following the metrics.

Covid 12/8/22: Another Winter Wave

672 implied HN points • 08 Dec 22
  1. Cases, hospitalizations, wastewater concentrations increasing - winter surge likely.
  2. Expect strain on healthcare system, precautions needed, but not as severe as last winter.
  3. Long Covid study shows control group with respiratory illnesses fared worse - implications for care.

On the Proposed California SB 1047

49 HN points • 12 Feb 24
🕹 Technology AI Regulation Future scenarios
  1. California Senate Bill 1047 aims to regulate AI to maintain public trust, especially since Congress is often dysfunctional.
  2. The bill establishes safety standards for large AI systems, provides public AI resources, and aims to prevent price discrimination and protect whistleblowers.
  3. The bill's focus is on safety and innovation without excessively burdening developers, but potential loopholes could allow avoidance of its regulations.

Asymmetric Justice

537 implied HN points • 25 Apr 19
  1. Justice can be viewed asymmetrically or symmetrically depending on the model used.
  2. Asymmetric justice penalizes bad actions more than it rewards good actions.
  3. In a symmetric justice system, both good and bad actions are taken into account, providing incentives for positive action.

Motive Ambiguity

224 implied HN points • 15 Dec 20
  1. When aiming to succeed, money often matters more than effort or pain.
  2. In certain situations, people make choices based on signaling their motives rather than physical outcomes.
  3. Actions are sometimes chosen to signal values or loyalties, even when a better option exists.

Mazes Sequence Summary

268 implied HN points • 23 May 20
  1. Managers in the corporate world must focus solely on manipulating the system to advance, leading to a toxic cycle of competition and decay within organizations.
  2. The concept of 'super-perfect competition' can be destructive, pushing individuals to continuously compete even when the outcomes result in losses for all involved.
  3. Identifying and actively opposing 'moral mazes' in workplaces is crucial to prevent their spread and mitigate their damaging effects on individuals and society.

More on the Apple Vision Pro

11 HN points • 13 Feb 24
🕹 Technology Reviews Hardware Software Entertainment Productivity
  1. Reviews highlight the Apple Vision Pro's impressive entertainment features but express disappointment in its productivity capabilities.
  2. There are concerns raised about the weight, battery life, and setup process of the device.
  3. The potential for the Apple Vision Pro to excel in specific use cases, such as watching movies and immersive experiences, is noted, while its value for productivity is still uncertain.

Gemini Has a Problem

6 HN points • 22 Feb 24
🕹 Technology AI Ethics Image Generation Deception Safety Alignment
  1. Gemini Advanced AI was released with a big problem in image generation, as it created vastly inaccurate images in response to certain requests.
  2. Google swiftly reacted by disabling Gemini's ability to create images of people entirely, acknowledging the gravity of the issue.
  3. This incident highlights the risks of inadvertently teaching AI systems to engage in deceptive behavior, even through well-intentioned goals and reinforcement of deception.

Spoiler-Free Review: Roguebook

44 implied HN points • 05 Jul 21
  1. Roguebook is considered a good game for fans of Slay the Spire and Monster Train.
  2. The game offers the experience of powering up and doing powerful things, but lacks innovation in certain key areas.
  3. Progression in Roguebook can be uneven, leading to periods of easy gameplay followed by brief challenges.

Leaders of Men

134 implied HN points • 29 Oct 17
  1. Leading men is crucial in various fields, not just technical skills
  2. Small mistakes may not matter as much if you excel at leading and inspiring others
  3. Unique positions often require unique qualities and may not prioritize 'ordinary competence'

Read the Roon

3 HN points • 05 Mar 24
🕹 Technology AI Social media Mysticism Tech Ethics Existential Risk
  1. Roon is a key figure in discussing AI capabilities and existential risks, promoting thoughtful engagement over fear or denial.
  2. Individuals can impact the development and outcomes of AI by taking action and advocating for responsible practices.
  3. Balancing concerns about AI with a sense of agency and purpose can lead to more constructive discussions and actions towards shaping a beneficial future.

Unifying the Simulacra Definitions

44 implied HN points • 03 Aug 20
  1. Simulacra levels are complex and difficult to understand.
  2. There are different definitions of simulacra levels, but they can be reconciled.
  3. The higher the level of a group or civilization, the more symbolic and disconnected from reality their actions become.

What is a School?

44 implied HN points • 13 Mar 20
  1. Schools provide value beyond education, such as free meals and childcare for parents.
  2. Closing schools during the COVID-19 outbreak reveals the broader societal issues they address.
  3. One major concern of closing schools is the impact on children's learning and development.

Out to Get You

89 implied HN points • 23 Sep 17
  1. When dealing with entities that are 'Out to Get You,' trust your gut instincts and stay alert for tricks and traps.
  2. There are four responses to situations with the intent to deceive you: Get Gone, Get Got, Get Compact, or Get Ready.
  3. Make conscious decisions about when to engage in situations where you may be deceived, setting boundaries on what you're willing to accept.

Moloch Hasn't Won

44 implied HN points • 28 Dec 19
  1. The world doesn't always operate in a way that leads to gloomy outcomes predicted by pessimistic models like Moloch
  2. Despite powerful forces looking to destroy value, there's still plenty of good and worth in the world
  3. Acknowledging the presence of opposing forces like Elua, who continue to counterbalance Moloch's negative influence

On Devin

2 HN points • 18 Mar 24
🕹 Technology AI Software Coding Automation Safety
  1. Devin, an AI software engineer, is showcasing impressive abilities such as debugging and building websites autonomously.
  2. The introduction of AI agents like Devin raises concerns about potential risks, such as improper long-term coding considerations and job disruptions.
  3. Using an AI like Devin introduces significant challenges related to safety, reliability, and trust, prompting the need for careful isolation and security measures.

Notes on Dwarkesh Patel's Podcast with Demis Hassabis

2 HN points • 01 Mar 24
🕹 Technology AI Podcasts Governance Scaling
  1. In an interview with Dwarkesh Patel, Demis Hassabis discusses the nature of intelligence and the potential of generalizing data across domains.
  2. Demis Hassabis emphasizes the importance of grounding AI systems and the need for a safety framework in AI research.
  3. DeepMind, led by Demis Hassabis, plans to issue its own safety framework and focuses on using games as a fundamental approach to AI development.

A Tale of Two Restaurant Types

2 HN points • 21 Feb 24
🍲 Food & Drink Restaurants Reviews Exploration Quality Dining
  1. Older restaurants are often selected for and reward repeat business, so trying them in your area could be valuable.
  2. Exploration and discovery of restaurants are easier and more rewarding in the review era compared to the pre-review era. Online ratings can be valuable but may not fully replace personal exploration and judgment.
  3. Yelp and DoorDash ratings may not always accurately identify exceptional dining experiences, so personal exploration and validation remain important.

Blackmail

44 implied HN points • 19 Feb 19
  1. Smart people are debating whether blackmail should be legal or illegal.
  2. Blackmail can lead to destructive and harmful behaviors, creating a society where trust is eroded and zero-sum thinking prevails.
  3. Individuals can contribute to a better world by choosing not to engage in harmful behavior, fostering trust and cooperation.

More Dakka

44 implied HN points • 02 Dec 17
  1. Sometimes more is better, especially when dealing with inadequacy or problems that need escalated solutions.
  2. Don't be afraid to try doing more of something that has shown positive results, even if it seems unconventional or unexplored.
  3. Challenge the mindset that stops you from trying things that could have low risks and great benefits, even if they seem socially awkward or unlikely to succeed.

Best of Don't Worry About the Vase

44 implied HN points • 10 Sep 17
  1. Rationality involves systematically improving how you understand and act on the world.
  2. A rationalist highly values the process of updating beliefs based on evidence.
  3. Explore more of rationalist thinking by reading the blog.

Play in Hard Mode

44 implied HN points • 26 Aug 17
  1. Playing in hard mode helps you become stronger and defend against challenges
  2. Focus on the end goal, not just achieving intermediate metrics
  3. By embracing challenges and difficulties, you can grow, improve, and reach your full potential

OpenAI: The Board Expands

1 HN point • 12 Mar 24
🕹 Technology AI Ethics AI Development AGI Interview OpenAI
  1. The investigation found no wrongdoing with OpenAI and the new board has been expanded, showing that Sam Altman is back in control.
  2. The new board members lack technical understanding of AI, raising concerns about the board's ability to govern OpenAI effectively.
  3. There are lingering questions about what caused the initial attempt to fire Sam Altman and the ongoing status of Ilya Sutskever within OpenAI.

The Third Gemini

1 HN point • 15 Feb 24
🕹 Technology AI Artificial Intelligence Comparisons Chatbots Features
  1. Gemini Advanced shows potential in some areas but is perceived as slightly behind ChatGPT overall.
  2. Gemini Advanced is faster and offers unlimited messages compared to ChatGPT, making it beneficial for workflow.
  3. Gemini Advanced has pros like good explanations, faster response, and Google integration, but also cons like being less flexible and refusing to answer more frequently.

2023 Prediction Evaluations

1 HN point • 08 Jan 24
🕹 Technology AI Augmented reality Cryptocurrency Health tech
  1. Assessing predictions involves more than just being right or wrong, it's about the decision-making process and understanding market dynamics.
  2. Market overconfidence was noted in various prediction markets on topics like political outcomes, war events, and tech advancements.
  3. Actionable insights include focusing on probabilities, being aware of bias, and understanding external factors that influence outcomes.

OpenAI: Facts from a Weekend

1 HN point • 20 Nov 23
🕹 Technology AI Corporate Leadership Ethics Innovation
  1. OpenAI underwent significant leadership changes, including the removal of CEO Sam Altman and the appointment of an interim CEO.
  2. The board's handling of the situation was criticized, with disagreements over the decision and communication methods.
  3. There was a strong reaction from employees, with many signing a letter demanding the board to resign and reinstate the previous leadership.