The hottest Machine Learning Substack posts right now

And their main takeaways
Category
Top Business Topics

Has AI Progress Stalled?

The Future of Life • 19 implied HN points • 21 Jul 24
🕹 Technology Machine Learning
  1. AI improvement has slowed down in terms of new abilities since GPT-4 came out, but other factors like cost and speed have gotten much better.
  2. The focus now is on practical changes and making AI more valuable, which will help set the stage for bigger breakthroughs in the future.
  3. Reaching human-level skills in tests doesn't mean AI will be truly intelligent. Future development will need to incorporate more complex abilities like planning and learning from experiences.

OpenAI's Reinforcement Finetuning and RL for the masses

Democratizing Automation • 427 implied HN points • 11 Dec 24
🕹 Technology Machine Learning
  1. Reinforcement Finetuning (RFT) allows developers to fine-tune AI models using their own data, improving performance with just a few training samples. This can help the models learn to give correct answers more effectively.
  2. RFT aims to solve the stability issues that have limited the use of reinforcement learning in AI. With a reliable API, users can now train models without the fear of them crashing or behaving unpredictively.
  3. This new method could change how AI models are trained, making it easier for anyone to use reinforcement learning techniques, not just experts. This means more engineers will need to become familiar with these concepts in their work.

The AI Conference in SF: The Future of AI is Now!

Gradient Flow • 319 implied HN points • 18 May 23
🕹 Technology Machine Learning
  1. The AI Conference in San Francisco aims to bridge the gap between research and real-world applications of AI by providing a vendor-neutral platform for networking and learning.
  2. The conference is seeking speakers with expertise in implementing AI across various industries like healthcare, finance, manufacturing, and more, as well as in model development and deployment.
  3. Cutting-edge developments in AI include advancements such as a benchmarking platform for large language models with Elo ratings, reduced latency in Apache Spark Structured Streaming, and AI systems like Med-PaLM 2 for medical question answering.

Data Science Weekly - Issue 520

Data Science Weekly Newsletter • 239 implied HN points • 10 Nov 23
🕹 Technology Machine Learning
  1. Data scientists share interesting links and news weekly about AI, machine learning, and data visualization. It's a great way to stay updated on trends and tools in the field.
  2. Learning about the basics of deep learning and mathematical foundations is important for anyone starting in machine learning. Understanding key concepts helps you tackle complex problems more effectively.
  3. There are many job opportunities in data science and related fields. Keeping an eye on openings can lead to exciting career advancements and collaborations.

Get the Most Out of Your Custom LLMs

Gradient Flow • 279 implied HN points • 15 Jun 23
🕹 Technology Machine Learning
  1. Custom Large Language Models (LLMs) and Custom Foundation Models can enhance accuracy, data privacy, and security in specialized fields like healthcare, law, and finance.
  2. Training custom models involves crucial stages like Pre-training, Supervised Fine-Tuning, Reward Modeling, and Reinforcement Learning.
  3. WeightWatcher is an open-source tool that helps analyze and improve the performance of deep learning models, aiding in conserving resources, detecting model saturation, and enhancing model quality.
Get a weekly roundup of the best Substack posts, by hacker news affinity:

The Galactic Guide to SHAP Values

Mindful Modeler • 279 implied HN points • 25 Jul 23
🕹 Technology Machine Learning
  1. SHAP values are like forces acting on a planet in a universe analogy, helping explain machine learning model predictions
  2. Each feature in a machine learning model contributes as a force, with SHAP values showing how they impact the prediction
  3. SHAP values aim to maintain the prediction's equilibrium by considering all forces, revealing which features are vital

OpenAI's o1 using "search" was a PSYOP

Democratizing Automation • 435 implied HN points • 04 Dec 24
🕹 Technology Machine Learning
  1. OpenAI's o1 models may not actually use traditional search methods as people think. Instead, they might rely more on reinforcement learning, which is a different way of optimizing their performance.
  2. The success of OpenAI's models seems to come from using clear, measurable outcomes for training. This includes learning from mistakes and refining their approach based on feedback.
  3. OpenAI's approach focuses on scaling up the computation and training process without needing complex external search strategies. This can lead to better results by simply using the model's internal methods effectively.

How In-Context Learning Emerges

Last Week in AI • 437 implied HN points • 21 Jul 23
🕹 Technology Machine Learning
  1. In-context learning (ICL) allows Large Language Models to learn new tasks without additional training.
  2. ICL is exciting because it enables versatility, generalization, efficiency, and accessibility in AI systems.
  3. Three key factors that enable and enhance ICL abilities in large language models are model architecture, model scale, and data distribution.

There's no such thing as "machine learning."

Top Carbon Chauvinist • 19 implied HN points • 20 Jul 24
🕹 Technology Machine Learning
  1. Machines don't really learn like humans do. They can take in data and improve performance, but they don't understand or experience learning in the same way we do.
  2. The term 'machine learning' can be misleading. It's more about machines mimicking learning processes rather than actually experiencing them.
  3. Understanding how machines operate helps clarify their limitations. They can process large amounts of information but lack conscious experience or true comprehension.

How to AI

The Intersection • 277 implied HN points • 19 Sep 23
🕹 Technology Machine Learning
  1. History often repeats itself in the adoption of new technologies, as seen with the initial skepticism towards digital marketing and now with AI.
  2. Brands are either cautiously experimenting with AI for PR purposes or holding back due to concerns like data security, plagiarism, and unforeseen outcomes.
  3. AI's evolution spans from traditional artificial intelligence to the current era dominated by generative AI, offering operational efficiency, creative enhancements, and transformative possibilities.

Artificial Intelligences in the Guanzi and the Han Feizi

Daoist Methodologies • 275 implied HN points • 20 Mar 23
🕹 Technology Machine Learning
  1. Confucians and Daoists have different approaches to learning: acquiring proven vs. eliminating disproven methods.
  2. Learning can occur without a brain: different entities like dogs, rivers, and markets learn without understanding.
  3. Systems can learn independently without human input, like in the case of artificial intelligence drawing from socio-political systems described in ancient texts.

Prompt Ensembles Make LLMs More Reliable

Deep (Learning) Focus • 275 implied HN points • 15 May 23
🔬 Science Machine Learning
  1. Reliability is crucial when working with large language models, and prompt ensembles offer a straightforward way to make them more accurate and consistent.
  2. Prompt ensembles show generalization across different language models, reducing sensitivity to changing underlying models and prompts.
  3. Aggregation of multiple outputs from prompt ensembles is complex but crucial for improving model performance, requiring sophisticated strategies beyond simple majority voting.

Strategies for an Accelerating Future

One Useful Thing • 1033 implied HN points • 20 Feb 24
🕹 Technology Machine Learning
  1. Advancements in AI, such as larger memory capacity in models like Gemini, are enhancing AI's ability for superhuman recall and performance.
  2. Improvements in speed, like Groq's hardware for quick responses from AI models, are making AI more practical and efficient for various tasks.
  3. Leaders should consider utilizing AI in their organizations by assessing what tasks can be automated, exploring new possibilities made possible by AI, democratizing services, and personalizing offerings for customers.

Genre Grapevine on Machine Learning’s Problem with Bad Stories and Bad Intentions

Genre Grapevine • 176 implied HN points • 29 Dec 23
🎨 Art & Illustration Machine Learning
  1. Bad stories can inspire writers to improve their own writing by learning from the mistakes of others.
  2. Artists and writers have pushed back against AI dominance by engaging in strikes and filing lawsuits to protect their work from being used without permission.
  3. Machine learning programs face challenges in creating truly innovative and original art, as they often get stuck in a cycle of reproducing popular styles and lacking true imagination.

The Sequence Knowledge # 780: Synthetic Data for Image Models

TheSequence • 21 implied HN points • 30 Dec 25
🕹 Technology Machine Learning
  1. Synthetic image data is now a core tool for vision models and works especially well when real images are scarce, private, or unbalanced by providing labeled pixels and covering rare edge cases.
  2. Modern generative models (diffusion models, GANs) combined with conditional controls like segmentation, depth, keypoints, ControlNet, or LoRA let you steer layout, pose, lighting, and style; typical pipelines script prompts, generate images, and auto-label using the same controls.
  3. Success depends on choosing the right generator and control signals and running a rigorous quality-control loop so synthetic variety actually improves downstream performance, a pattern already used in systems like NVIDIA’s Synthetica for robot training.

GroupBy #34: Hybrid Transactional/Analytical Storage, From Predictive to Generative – How Michelangelo Accelerates Uber’s AI Journey

VuTrinh. • 59 implied HN points • 07 May 24
🕹 Technology Machine Learning
  1. Hybrid transactional/analytical storage combines different types of data processing. This helps companies like Uber manage their data more efficiently.
  2. The shift from predictive to generative AI is changing how companies use machine learning. Uber's Michelangelo platform shows how this new approach can improve AI applications.
  3. Data reliability and observability are important for businesses as their data grows. Companies need tools to quickly find and fix data issues to keep their operations running smoothly.

The Turing Test Is Irrelevant

Top Carbon Chauvinist • 19 implied HN points • 19 Jul 24
🕹 Technology Machine Learning
  1. The Turing Test isn't a good measure of machine intelligence. It's actually more important to see how useful a machine is rather than just how well it imitates human behavior.
  2. People often confuse looking reliable with actually being reliable. A machine can seem smart but still not function correctly in tasks.
  3. We should focus on improving how machines handle calculations and information, rather than just whether they can mimic humans. True effectiveness is more valuable than just good imitation.

The ICML 2025 Outstanding Papers

Gonzo ML • 126 implied HN points • 28 Jul 25
🕹 Technology Machine Learning
  1. The recent ICML 2025 Outstanding Papers show a huge amount of important research in machine learning, but many people feel overwhelmed and can't read everything in-depth.
  2. It's okay to admit that you can't keep up with all the new papers. Using AI tools can help manage the load and ensure you're still getting the important insights you need.
  3. Some of the papers focus on practical issues, like improving predictions and making AI more collaborative, which are vital for real-world applications.

GPT-4o mini

Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots • 19 implied HN points • 18 Jul 24
🕹 Technology Machine Learning
  1. GPT-4o mini is a new language model that's cheaper and faster than older models. It handles text and images and is great for tasks requiring quick responses.
  2. Small Language Models (SLMs) like GPT-4o mini can run efficiently on devices without relying on the cloud. This helps with costs, privacy, and gives users more control over the technology.
  3. SLMs are designed to be flexible and customizable. They can learn from various types of inputs and can adapt more easily to specific needs.

Mondays with the Machine: The Tongue & the Token: Language as Interface in Our Current Age of AI

Brad DeLong's Grasping Reality • 169 implied HN points • 09 Jun 25
🕹 Technology Machine Learning
  1. Natural language interfaces are a big deal because they let us communicate with AI using everyday language. This makes it easier for everyone to use technology without needing to know complex coding or technical skills.
  2. AI systems, like language models, simulate understanding but don't actually think. They can help us find information and assist with tasks, but we should remember that they are not truly intelligent.
  3. Using conversational AI can democratize access to information, making it easier for people to learn and solve problems. However, we must be aware of the risks, like over-reliance on these systems.

Phi-3, your Pocket LLM!

Aziz et al. Paper Summaries • 79 implied HN points • 29 Apr 24
🕹 Technology Machine Learning
  1. Microsoft's Phi-3 is a new AI model that is small enough to run on your phone, yet still performs well. This is a big deal because most AI models are too large for personal devices.
  2. The model uses high-quality, filtered data for training, focusing on reasoning and educational materials. This approach makes Phi-3 better at understanding rather than just memorizing facts.
  3. Even though Phi-3 is powerful, it has some limitations, like not being multilingual. There are also tasks it struggles with, like those needing lots of factual knowledge.

LLM Links, 2/1/2025

In My Tribe • 318 implied HN points • 01 Feb 25
🕹 Technology Machine Learning
  1. OpenAI's new AI agent, ChatGPT Operator, can take actions online for users, like booking services. However, some feel it doesn't yet handle more complex tasks very well.
  2. Different users highlight various ways they use AI, showing that it can be useful for specific inquiries, but many still feel they are stuck in old routines.
  3. AI technology is advancing fast, leading to concerns about job loss and social changes. People think the impacts of AI will evolve slowly, despite rapid progress in the tech itself.

Agentic AI: Creating An AI Agent Which Can Navigate The Internet

Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots • 19 implied HN points • 17 Jul 24
🕹 Technology Machine Learning
  1. WebVoyager is an AI agent that can browse the web by analyzing screenshots and deciding what to do next. It works like a human browsing the internet, using both visual and text information.
  2. The agent interacts with webpages by performing actions like clicking, scrolling, and typing. This allows it to complete tasks on websites without needing help from humans.
  3. WebVoyager's ability to handle complex web navigation shows the potential of AI agents to perform useful tasks autonomously. It learns to navigate better by using real-world websites rather than just simplified models.

Two reflections on AI

De Pony Sum • 255 implied HN points • 16 Oct 23
🕹 Technology Machine Learning
  1. Recent developments in AI, like language models, have surprised many with their capabilities and impact.
  2. There is a need for curiosity and humility when engaging with new AI technologies.
  3. Advancements in language models, such as using LATS, show promising improvements and future potentials.

The Sequence Radar #696: Google AI Ultra’s New Gold-Medal Reasoning Model is Available

TheSequence • 119 implied HN points • 03 Aug 25
🕹 Technology Machine Learning
  1. Google released a new AI model called Gemini 2.5 Deep Think that can solve complex math problems like a human. It performed so well that it won a gold medal at the International Math Olympiad.
  2. This model uses advanced strategies to explore many possible solutions at once, making it faster and more creative than previous AIs.
  3. The emergence of such powerful AI means we need to discuss how to use these systems responsibly, ensuring they benefit everyone and maintain fair access.

For Pete's sake, learn what a machine is.

Top Carbon Chauvinist • 19 implied HN points • 17 Jul 24
🕹 Technology Machine Learning
  1. A machine is made up of parts that do work by handling loads, like electricity or mechanics. It does not actually understand or think about what it does.
  2. When programming a machine, like a catapult, you're just adjusting physical elements, not teaching it to know or understand concepts like 'rock' or 'lever'.
  3. Living things are not machines because they aren't made of manufactured parts. They grow and evolve in ways that machines cannot.

The Sequence Research #673: Infinite Self-Improvement: Unpacking Sakana's Darwin Gödel Machine

TheSequence • 154 implied HN points • 27 Jun 25
🕹 Technology Machine Learning
  1. The Darwin Gödel Machine (DGM) is a new kind of AI that can change its own code to improve. It combines two ideas: self-modifying machines and evolving through trial and error.
  2. Instead of needing complicated proofs for changes, DGM tests its code edits under real-world conditions. This helps it learn quickly and safely from what works.
  3. DGM has shown significant improvement in coding benchmarks, outperforming humans and traditional methods. This means it can continually get better at coding and solving problems.

Holy Grails of Data: Self-Service, Single Truths, and the Role of AI

SeattleDataGuy’s Newsletter • 365 implied HN points • 27 Dec 24
🕹 Technology Machine Learning
  1. Self-service analytics is still a goal for many companies, but it often falls short. Users might struggle with the tools or want different formats for the data, leading to more questions instead of fewer.
  2. Becoming truly data-driven is a challenge for many organizations. Trust issues with data, preference for gut feelings, and poor communication often get in the way of making informed decisions.
  3. People need to be data literate for businesses to succeed with data. The data team must present insights clearly, while business teams should understand and trust the data they work with.

The Importance Of Granular Data Design For Fine-Tuning

Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots • 59 implied HN points • 02 May 24
🕹 Technology Machine Learning
  1. Granular data design helps improve the behavior and abilities of language models. This means making training data more specific so the models can reason better.
  2. New methods like Partial Answer Masking allow models to learn self-correction. This helps them improve their responses without needing perfect answers in the training data.
  3. Training models with a focus on long context helps them retrieve information more effectively. This approach tackles issues where models can lose important information in lengthy input.

Mondays with the Machine: Thinking Out Loud About the Current "AI" Boom-Bubble

Brad DeLong's Grasping Reality • 169 implied HN points • 02 Jun 25
🕹 Technology Machine Learning
  1. New technologies like AI often cause panic as people worry about their impact, similar to how calculators were once banned in schools. Over time, we learn to use these tools responsibly.
  2. AI chatbots can seem human-like, but they are actually complex tools for finding information. Instead of treating them like people, we should learn how to use them effectively for our needs.
  3. While AI can generate a lot of ideas quickly, it lacks the depth and truthfulness that history provides. History gives us valuable lessons, but AI can still help spark new thoughts and start conversations.

SAI #19: The Data Value Chain.

SwirlAI Newsletter • 255 implied HN points • 25 Feb 23
🕹 Technology Machine Learning
  1. Understanding the Data Value Chain is essential for building successful Data Products.
  2. Implementing Data Contracts in the Data Pipeline ensures data quality and prevents unexpected outages.
  3. Knowing the 4 types of ML Model Deployment helps in deploying machine learning models effectively.

Deep Learning Weekly: Issue #308

Deep Learning Weekly • 255 implied HN points • 05 Jul 23
🕹 Technology Machine Learning
  1. This week's issue of Deep Learning Weekly covers Meta's AI system cards, real-time machine learning foundations at Lyft, and a local code generator tool using Microsoft's guidance library.
  2. Industry news includes Inflection AI's $1.3B investment, Meta AI sharing 22 system cards on AI experiences, and Unity launching new AI platforms for real-time 3D creation.
  3. In the MLOps and Learning sections, topics range from dealing with train-serve skew in ML models to using LLMs for data extraction and building local code generators.

Data Science Weekly - Issue 510

Data Science Weekly Newsletter • 279 implied HN points • 31 Aug 23
🕹 Technology Machine Learning
  1. Autonomous drones can now race at human champion levels using deep reinforcement learning. This shows how advanced technology can mimic skilled human behavior in competitive sports.
  2. Google is rapidly developing its AI capabilities and plans to surpass GPT-4 by a significant margin soon. This could lead to more powerful AI tools for various applications.
  3. Reinforced Self-Training (ReST) is a new method for improving language models by aligning their outputs with human preferences. It offers better translation quality and can be done efficiently with less data.

Diffusion Models are Evolutionary Algorithms

Gonzo ML • 441 implied HN points • 09 Nov 24
🕹 Technology Machine Learning
  1. Diffusion models and evolutionary algorithms both involve changing data over time through processes like selection and mutation, which can lead to new and improved results.
  2. The new algorithm called Diffusion Evolution can find multiple good solutions at once, unlike traditional methods that often focus on one single best solution.
  3. There are exciting connections between learning and evolution, hinting that they may fundamentally operate in similar ways, which opens up many questions about future AI developments.

Captain's log: the irreducible weirdness of prompting AIs

One Useful Thing • 902 implied HN points • 04 Mar 24
🕹 Technology Machine Learning
  1. Stop trying to use incantations: There is no single magic word that works all the time with AIs. Promising rewards or being polite may help occasionally, but not always.
  2. There are prompting techniques that work consistently: Techniques like adding context to prompts, providing a few examples, and using Chain of Thought can help in crafting better prompts for AIs.
  3. Prompting matters significantly: The way you prompt AIs can have a huge impact on the outcomes. Good prompts can turn a difficult task into an easy one for AI.