The hottest ML Substack posts right now

And their main takeaways
Category
Top Technology Topics

Let’s be honest, Generative AI isn’t going all that well

Marcus on AI • 15532 implied HN points • 12 Jan 26
🕹 Technology ML
  1. Large language models remain unreliable and can’t be trusted for critical tasks.
  2. Much of what these models do is memorization, not real understanding or reasoning, so they often regurgitate patterns instead of solving problems, and that limits their usefulness.
  3. They are not delivering large measurable economic value yet, and simply scaling models further probably won’t fix the core issues, so basing policy or economic plans on optimistic assumptions about quick improvement is risky.

The Sequence 802: The Thinking Machine: A Deep Dive into Test-Time Compute and the New Scaling Paradigm

TheSequence • 21 implied HN points • 05 Feb 26
🕹 Technology ML
  1. For years AI advanced by scaling up pre-training—more data, bigger models, and huge GPU time to bake capabilities into fixed weights.
  2. Test-time compute flips that idea by letting models use extra computation during inference to reason, plan, backtrack, and self-correct—basically "letting the model think."
  3. The big implication is that model performance depends not just on training compute but also on how much compute is allowed at inference, changing tradeoffs for how we build and deploy AI.

7 Must-Have Features for Crafting Custom LLMs

Gradient Flow • 299 implied HN points • 21 Sep 23
🕹 Technology ML
  1. Crafting custom large language models (LLMs) is essential for addressing concerns about intellectual property, data security, and privacy.
  2. Tools for building custom LLMs must include versatile tuning techniques, human-integrated customization, and data augmentation capabilities.
  3. Developing multiple custom LLMs requires features like experimentation facilitation with tools such as MLflow, the use of distributed computing accelerators, and documentation excellence for alignment, accuracy, and reliability.

The New Era of Efficient LLM Deployment

Gradient Flow • 299 implied HN points • 13 Jul 23
🕹 Technology ML
  1. AI tools are becoming pervasive in tech with potential to increase productivity and contribute trillions annually to global productivity
  2. Efficient deployment of large language models (LLMs) is crucial for businesses to scale their AI initiatives and drive digital innovation
  3. Rethinking MLOps infrastructure is essential to accommodate the scale and complexity of LLMs, with a need for solutions addressing challenges in inference, serving, and deployment
Get a weekly roundup of the best Substack posts, by hacker news affinity:

Deep Learning Weekly : Issue #309

Deep Learning Weekly • 216 implied HN points • 12 Jul 23
🕹 Technology ML
  1. Deep Learning Weekly Issue #309 covers topics like Code Interpreter on ChatGPT Plus and ML system design with 200 case studies.
  2. Industry innovations include AI-generated chart captions and Nvidia's AI approach to carbon capture.
  3. Learning section highlights topics like Tiny Audio Diffusion and Swin Transformer for object recognition.

About my own interest in the current generative AI craze

Bojan’s Newsletter • 176 implied HN points • 27 Feb 23
🕹 Technology ML
  1. The author has not done fundamental work in generative AI, but has potential projects that may go in that direction.
  2. The author's interest in generative AI is linked to their long-term interest in the future of work, which directly affects their professional life.
  3. Generative AI tools have the potential to transform work dynamics significantly, especially in creative fields.

Could Decision Trees Help With GenAI?

Bojan’s Newsletter • 137 implied HN points • 13 Mar 23
🕹 Technology ML
  1. Decision Trees are known for being accurate and robust in tabular data modeling.
  2. Generative AI systems can sometimes create inaccurate content, especially in domains where accuracy is crucial.
  3. Using tree-based ML models could potentially address issues of hallucination in Generative AI.

Why Job Displacement Predictions are Wrong: Explanations of Cognitive Automation

Scaling Knowledge • 117 implied HN points • 30 May 23
🕹 Technology ML
  1. Predictions about job displacement due to large language models are often wrong because they lack explanations of how LLMs and human intelligence differ.
  2. Jobs are more likely to be augmented than automated by technologies like LLMs, as human creativity and autonomy are essential in many fields like software engineering, medicine, law, and media production.
  3. Regulations on AI and cognitive automation may hinder progress and knowledge creation, leading to unforeseen consequences and limiting the potential benefits of such technologies.

Edge 373: Computationally Efficient LLM Reasoning with ReWOO

TheSequence • 413 implied HN points • 27 Feb 24
🕹 Technology ML
  1. ReWOO is a new reasoning technique optimized for information augmented LLMs, focusing on step-wise reasoning, tool-calls, and summarization as separate modules.
  2. RAG techniques impact the reasoning abilities of LLMs in generative AI applications, often requiring coordination between LLMs and external tools, which can increase computational demands.
  3. LLMFlows is introduced as a framework for building LLM applications, showcasing the importance of augmenting LLMs with external data like RAG to enhance their capabilities.

A minimal viable product for alignment

Musings on the Alignment Problem • 399 implied HN points • 29 Mar 22
🕹 Technology ML
  1. Progress in AI can expand the range of problems humanity can solve, addressing the limitation of human capabilities.
  2. Automating alignment research using AI systems can accelerate progress by overcoming talent bottlenecks and enabling faster evaluation and generation of solutions.
  3. An alignment MVP approach is less ambitious than solving all alignment problems but can still lead to solutions by leveraging automation and AI capabilities.

Pinterest improves their Closeup Recommendation System through foundational changes

MLOps Newsletter • 98 implied HN points • 07 Oct 23
🕹 Technology ML
  1. Pinterest improved their Closeup Recommendation System with foundational changes like hybrid data logging and sampling.
  2. Pinterest uses a model refreshing framework to keep their Closeup Recommendation model up-to-date and adaptable.
  3. Distilling step-by-step can help train smaller, more efficient, and interpretable language models like LLMs.

Keep Your Dreams with the Help of AI

Addition • 78 implied HN points • 02 May 23
🕹 Technology ML
  1. A partnership created The Dreamkeeper, an AI tool to preserve dreams
  2. The Dreamkeeper uses AI models to remember, visualize, and organize dreams
  3. AI can potentially record dreams through fMRI scans in the future

Market Map & Analysis: AI Synthetic Data Companies

The Strategy Deck • 78 implied HN points • 06 Jul 23
🕹 Technology ML
  1. Synthetic data is crucial for ML by replacing real-world data, protecting sensitive information, and validating AI applications.
  2. Synthetic data is used in computer vision for autonomous vehicles and is expanding to other data types like text and tabular data.
  3. There are specialized and general-purpose synthetic data platforms developing innovative solutions for various industries and use cases.

Dimension Hopper Part 1

General Robots • 395 HN points • 12 Jun 23
🕹 Technology ML
  1. The project involves creating a 2D platformer where players design levels and AI generates visual representations.
  2. The journey to achieve this project involved experimenting with different techniques and models, such as adjusting depth images and adding more detail to improve visual outcomes.
  3. Using the right control images, supporting structures, and techniques like adding adjustable roughness, greatly improved the quality of the generated images.

OpenAI’s Sora for video, Gemini 1.5's infinite context, and a secret Mistral model

Democratizing Automation • 221 implied HN points • 16 Feb 24
🕹 Technology ML
  1. OpenAI introduced Sora, an impressive video generation model blending Vision Transformer and diffusion model techniques
  2. Google unveiled Gemini 1.5 Pro with nearly infinite context length, advancing the performance and efficiency using the Mixture of Expert as the base architecture
  3. The emergence of Mistral-Next model in the ChatBot Arena hints at an upcoming release, showing promising test results and setting expectations as a potential competitor to GPT4

How to cultivate a high-signal AI feed

Democratizing Automation • 166 implied HN points • 28 Feb 24
🕹 Technology ML
  1. Be intentional about your media diet in the ML space, curate and focus your energy to save time and avoid misleading content.
  2. When evaluating ML content, focus on model access, credibility, and demos; choosing between depth or breadth in your feed; and checking for reproducibility and verifiability.
  3. Ensure to socialize your information, build relationships in the community, and consider different sources and content types for a well-rounded perspective.

The Good A.I. We’re Not Talking About

The Digital Anthropologist • 19 implied HN points • 04 Jan 24
🕹 Technology ML
  1. Artificial Intelligence (AI) is not just about Generative AI (GAI) like ChatGPT. There are various other proven AI tools like Machine Learning (ML), Deep Learning, Natural Language Processing (NLP), and Expert Systems being successfully used in industries such as healthcare, manufacturing, and more.
  2. AI tools have been around for decades and have shown significant positive impacts on society. Despite the hype around GAI, it remains a small part of the broader AI landscape.
  3. Beyond the flashy headlines, many AI applications are working behind the scenes in specialized industries, quietly making a positive difference. While GAI is getting attention, the real-world impact of other AI tools continues to be substantial.

LLM agents and integration dead-ends

Democratizing Automation • 146 implied HN points • 12 Jul 23
🕹 Technology ML
  1. The biggest immediate roadblock in generative AI unlocking economic value is the barrier of enabling direct integration of language models
  2. Many are exploring the use of large language models (LLMs) for various business tasks through LLM agents, which are facing challenges of integration and broad scope
  3. The successful commercial viability of LLM agents depends on trust, reliability, management of failure modes, and understanding of feedback dynamics

📌 ML Engineering Event: Mastering AI and ML at Production Scale at apply()

TheSequence • 84 implied HN points • 19 Feb 24
🕹 Technology ML
  1. The event offers real-world insights from engineering leaders on ML model deployment and best practices.
  2. Participants can engage in sponsor-free knowledge sharing sessions with peers, focusing on in-depth discussions.
  3. Attendees have the opportunity to network with a diverse group of AI and ML engineers, including industry veterans and emerging leaders.

S. Somasegar on the Present and Future of Generative AI

TheSequence • 161 implied HN points • 15 Mar 23
🕹 Technology ML
  1. Generative AI is a subsegment of intelligent applications with potential in enterprise and consumer use cases.
  2. Developer tools will be reimagined with foundation models, enhancing productivity and code quality.
  3. New capabilities in generative AI models include the use of 'agents' for natural language interpretation and actions.

Hot Topics #21 (Mar. 17, 2023)

The Merge • 19 implied HN points • 17 Mar 23
🕹 Technology ML
  1. GPT-4 is a new large-scale model by OpenAI that can accept image and text inputs to produce text outputs.
  2. PaLM-E is an embodied multimodal language model that incorporates real-world sensor data into language tasks.
  3. Meta-black-box optimization can discover effective update rules for evolution strategies through meta-learning.

Gig Economy, AI, and LLM: Friends or Foes?

Gad’s Newsletter • 47 implied HN points • 05 Feb 24
🕹 Technology ML
  1. The gig economy connects freelancers with businesses through digital platforms for flexible, temporary work.
  2. Advancements in AI, particularly LLM and ML, are empowering gig workers by automating tasks, providing data-driven insights, and improving service quality.
  3. Challenges in the gig economy arise from the potential job displacement due to automation and AI advancements, along with ethical concerns about bias and privacy.

📌 ML Engineering Event: Lineup for apply() 2024 is Now Live!

TheSequence • 42 implied HN points • 08 Mar 24
🕹 Technology ML
  1. The lineup for the apply() 2024 ML Engineering Event, featuring industry leaders from LangChain, Meta, and Visa, is now live.
  2. The agenda includes keynote sessions on LangChain, semi-supervised learning, and uplift modeling by experts from the respective fields.
  3. Attendees can look forward to gaining insights and actionable tips for mastering AI and ML at the event.

Standing on the brains of giants

Startup Strategies • 71 implied HN points • 05 May 23
🕹 Technology ML
  1. AI is often using the intelligence of others, not truly artificial intelligence.
  2. Machines are successful because they combine the thoughts and ideas of many people.
  3. These AI systems can blur the lines between human and machine-generated ideas.

Meta's Inside the lab & Nvidia's GTC Spring

Sector 6 | The Newsletter of AIM • 39 implied HN points • 27 Feb 22
🕹 Technology ML
  1. Meta hosted a virtual event called 'Inside the Lab', focusing on their advancements in the metaverse. It aimed to share updates after their rebranding from Facebook.
  2. Nvidia's GTC Spring also featured important news in AI and machine learning. This event is known for showcasing the latest technology developments.
  3. These events highlight the growing interest and progress in virtual realities and AI technologies in the industry. People are excited about the future possibilities.

🙄 The AI is coming for my UX job

Counting Stuff • 54 implied HN points • 13 Apr 23
🕹 Technology ML
  1. A startup is using AI to create fake personas for product testing, but it misses the point of user testing.
  2. Usability studies run by project managers may be biased without proper training, focusing on understanding user motivations rather than specific actions.
  3. Like machine translation disrupted the translation market, AI in UX may provide some value for simple tasks but human experts are still needed for complex nuances.

📌 Exciting news! The speaker lineup for apply() 2024 is now live

TheSequence • 21 implied HN points • 15 Mar 24
🕹 Technology ML
  1. The speaker lineup for apply() 2024 event is now live, featuring industry leaders from companies like LangChain, Meta, Visa, and more.
  2. The event offers actionable insights to master AI and ML in production, with sessions on topics like LangChain Keynote, Semi-Supervised Learning, and Uplift Modeling.
  3. Attendees can register for free to join the event live on April 3rd, with the option to receive on-demand videos as well.

Local Llama and much more!

ScaleDown • 11 implied HN points • 15 Aug 23
🕹 Technology ML
  1. The newsletter focuses on deploying LLMs locally, offering tips and expert answers.
  2. It includes a comprehensive guide on local deployment of LLMs, combining reliable methods with innovation.
  3. The newsletter addresses top LLM questions, covering topics like overfitting, customization, and linguistic diversity.

LLM Chronicles #5: GPT For Ecommerce Search Engine With Pinecone

Pratik’s Pakodas 🍿 • 8 implied HN points • 09 May 23
🕹 Technology ML
  1. In certain scenarios, companies use 2 types of hybrid search: weighted scoring and filter and rerank, especially prevalent in e-commerce.
  2. GPT can be leveraged for query understanding to parse out complex queries and populate Elasticsearch/Solr with detected entities.
  3. Although using GPT-4 for this purpose may be costly and slow, training an open-source model like MPT-7B can be a more viable option.