jonstokes.com $5 / month

Jonstokes.com delves into the complexities and societal implications of AI and ML, exploring themes like AI safety, censorship, and power dynamics. It discusses the technical nuances of generative AI, its potential to innovate across fields, regulatory challenges, and ethical considerations surrounding AI's impact on society and individual freedoms.

AI/ML Technology Censorship and Power AI Safety and Ethics Generative AI Applications AI Regulation and Governance Innovation and Societal Change

The hottest Substack posts of jonstokes.com

And their main takeaways

ChatGPT Explained: A Normie's Guide To How It Works

587 implied HN points • 01 Mar 23
🕹 Technology Machine Learning AI Language Models Model Training Natural Language Processing
  1. Understand the basics of generative AI: a generative model produces a structured output from a structured input.
  2. Complex relationships between symbols require more computational power to relate them effectively.
  3. Language models like ChatGPT don't have personal experiences or knowledge; they use a token window to respond based on the conversation context.

AI Safety: A Technical & Ethnographic Overview

391 implied HN points • 30 Mar 23
🕹 Technology AI Ethics Research Models Intelligence
  1. The AI safety debate involves technical details about AI systems like GPT-4 and cultural dynamics around the issue.
  2. The discussion includes concerns about regulating and measuring AI capabilities, as well as the divisions and allegiances within different groups.
  3. Some groups, like the Intelligence Deniers, have strong beliefs about AI being a scam and hold firm against AI progress, leading to potential divisions among AI safety proponents.

I Say This Unironically: Our Society Is Not Prepared For This Much Awesome

319 implied HN points • 21 Feb 23
🕹 Technology AI Education Culture Security Creativity
  1. Generative AI is rapidly changing many aspects of society, affecting everything from artistic creation to education.
  2. Efforts to detect AI-generated content are ineffective, posing challenges for access control and gatekeeping.
  3. AI tools have the potential to enhance educational experiences, improve learning outcomes, but may also disrupt traditional credentialing systems.

Catechizing the Bots, Part 1: Foundation Models and Fine-Tuning

237 implied HN points • 28 May 23
🕹 Technology AI Text Training Models Ethics
  1. Foundation models for large language models go through fine-tuning phases to make them more user-friendly.
  2. Humans play a critical role in shaping the values and behaviors of these models during the fine-tuning process.
  3. Supervised fine-tuning involves exposing the model to smaller sets of carefully selected examples to anchor its output and establish dominant language structures.
Get a weekly roundup of the best Substack posts, by hacker news affinity:

Catechizing the Bots, Part 2: Reinforcement Learning and Fine-Tuning With RLHF

206 implied HN points • 10 Jun 23
🕹 Technology AI Machine Learning Neural Networks Reinforcement Learning Language Models
  1. Reinforcement Learning is a technique that helps models learn from experiencing pleasure and pain in their environment over time.
  2. Human feedback plays a crucial role in fine-tuning language models by providing ratings that indicate how a model's output impacts users' feelings.
  3. To train models effectively, a preference model can be used to emulate human responses and provide feedback without the need for extensive human involvement.

AI Safety: Is AI The Genie Or The Lamp?

237 implied HN points • 01 May 23
🕹 Technology AI Ethics Policy Engineering Regulation
  1. AI safety involves the debate between AI as a tool or an agent, impacting approaches to AI explainability and safety.
  2. There are conflicting folk conceptions of alignment, including individualist and collectivist perspectives centered around control.
  3. The distinction of viewing AI as the genie or the lamp, as an agent with goals or as a software tool, is crucial in shaping AI safety discussions and applications.

The CHAT Stack, GPT-4, And The Near-Term Future Of Software

237 implied HN points • 15 Mar 23
🕹 Technology Artificial Intelligence Software Development Machine Learning Data Storage API
  1. Developers will build apps on top of ChatGPT and similar models to create interactive and knowledgeable AI assistants
  2. The CHAT stack approach involves Context, History, API, and Token window, enabling how software applications will operate in the near future
  3. GPT-4 introduces an enlarged token window, improved control surfaces, and better ability to follow human instructions

How To Regulate AI, If You Must

175 implied HN points • 22 Jun 23
🕹 Technology AI Governance AI Development
  1. AI rules are inevitable, but the initial ones may not be ideal. It's a crucial moment to shape discussions on AI's future.
  2. Different groups are influencing AI governance. It's important to be aware of who is setting the rules.
  3. Product safety approach is preferred in AI regulation. Focus on validating specific AI implementations rather than regulating AI in the abstract.

Here's How Generative AI Can Fix The News

164 implied HN points • 15 Jun 23
🕹 Technology AI Media News Tools Content
  1. Generative AI has the potential to revolutionize the media industry and improve the quality of news stories.
  2. AI can streamline the news reporting process by assisting with drafting, editing, and formatting content.
  3. Creating AI-powered tools for editing, production, art, and promotion can enhance storytelling and make news creation more accessible.

AI Agent Basics: Let’s Think Step By Step

195 implied HN points • 21 Apr 23
🕹 Technology AI Software Machine Learning Programming Cost Analysis
  1. The rise of AI agents is introducing a new software paradigm that allows AI to make plans from text prompts.
  2. LLMs powered agents can generate detailed plans for achieving goals, revolutionizing the way tasks are accomplished.
  3. The agent paradigm offers a more cost-effective, yet higher-cost per run computation model compared to traditional software development, akin to the cloud computing model.

Why I'm A Better Editor Than GPT-4 (& Probably GPT-5)

175 implied HN points • 21 Mar 23
🕹 Technology AI Content creation
  1. A skilled human editor can spot viral potential in stories better than AI models like GPT-4 or GPT-5.
  2. The cost per token for AI models like GPT-4 is high, making human editing more cost-effective for steering content into the viral spotlight.
  3. Context compression and token window optimization are key challenges for AI models to catch up with human editors in understanding and writing content.

GPT-4 Doesn't Have "Gender Bias." It's Just Bad At Language (Still)

154 implied HN points • 18 May 23
🕹 Technology Artificial Intelligence Software Language processing Engineering Data Analysis
  1. Different approaches to evaluating AI performance have practical implications in development, deployment, and regulation.
  2. Language models like GPT-4 struggle with resolving ambiguity in human language due to limitations in understanding context.
  3. Using an engineering approach, providing relevant context, and improving language parsing can help mitigate language model biases and inaccuracies.