Top all-time · #1122
AssemblyAI
Company blog · assemblyai.com/blog · on HN as jamesheaiprogramming
AssemblyAI is the company blog at assemblyai.com, writing about ai and programming. Since 2018, 59 different Hacker News readers have linked it 165 times, most often “How ChatGPT actually works”. No feed was found, so read it on the site itself.
Rarely recommended · 4 of 40 sampled comments recommend it?
Why HN links it
HN readers link AssemblyAI's blog for explainers of machine learning topics, mainly diffusion models, DALL-E 2, Imagen, how ChatGPT works, and emergent LLM abilities, plus speech-to-text guides.
- Recommended for · 4 comments
- Called good explanations of diffusion models and DALL-E 2, a nice introduction to diffusion, and a comprehensive discussion of emergent abilities in LLMs.
Summarised by Claude from a sample of 40 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
With AssemblyAI's industry-leading Speech AI models, transcribe speech to text and extract insights from your voice data.
Most-linked posts?
- How ChatGPT actually works6 people · 6 mentions · first linked 2022-12-26
- AI Generative Models for Audio5 people · 5 mentions · first linked 2023-06-27
- Conformer-1: a robust speech recognition model5 people · 5 mentions · first linked 2023-03-15
- Introduction to Diffusion Models for Machine Learning4 people · 4 mentions · first linked 2022-05-12
- How to run Stable Diffusion locally4 people · 4 mentions · first linked 2022-08-23
- Emergent Abilities of Large Language Models3 people · 5 mentions · first linked 2023-03-07
- Poisson Flow Generative Models – new image generation model inspired by physics3 people · 5 mentions · first linked 2022-10-26
- RLHF vs. RLAIF for language model alignment3 people · 3 mentions · first linked 2023-08-22
- AlphaTensor explained – Motivation, method, and assessment3 people · 3 mentions · first linked 2022-11-22
- /3 people · 3 mentions · first linked 2021-02-24
What Hacker News says?
People have been solving this by using up-scaling AIs to increase the resolution of their images. That's actually built in to some of these systems - Google Imagen for example generates a 64x64 image and then up-scales it to 1024x1024: https://www.assemblyai.com/blog/how-imagen-actually-works/
Comment linking /blog/how-imagen-actually-works/
on “4.2 Gigabytes, Or: How to Draw Anything”Submitted PyTorch vs. TensorFlow in 2022 as a story · 350 points
HN discussionYeah actually, good call. The position of the token matters, since these things use transformers to encode the embeddings. https://www.assemblyai.com/blog/how-imagen-actually-works/
Comment linking /blog/how-imagen-actually-works/
on “DALL·E now available in beta”Submitted Building an end-to-end Speech Recognition model in PyTorch as a story · 288 points
HN discussionHere's a good explanation: https://www.assemblyai.com/blog/diffusion-models-for-machine...
Comment linking /blog/diffusion-models-for-machine-learning-introduction/
on “Stable Diffusion is a big deal”Submitted How to train large deep learning models as a startup as a story · 271 points
HN discussionhttps://alphacephei.com/vosk/install#usage-examples demonstrates the bare-bones vosk-transcriber sample, and there's also https://www.assemblyai.com/blog/getting-started-with-espnet I wasn't able to play with https://github.com/o-oconnell/mp4grep on ARM.
Comment linking /blog/getting-started-with-espnet
on “Tell HN: Otter.ai bot recording meetings without consent”Submitted How DALL-E 2 Works as a story · 252 points
HN discussionThey're currently hosted on Google in a way that you can download them via curl/wget. Here's a guide including the link: https://www.assemblyai.com/blog/how-to-run-stable-diffusion-...
Comment linking /blog/how-to-run-stable-diffusion-locally-to-generate-images/
on “DALL·E: Introducing Outpainting”Submitted Differentiable Programming – A Simple Introduction as a story · 159 points
HN discussionThese are good explanations: - https://www.assemblyai.com/blog/how-dall-e-2-actually-works/ - https://www.youtube.com/watch?v=F1X4fHzF4mQ
Comment linking /blog/how-dall-e-2-actually-works/
on “DALL-E 2 generates images of Kermit The Frog in various films”Submitted How Imagen Works as a story · 142 points
HN discussionAssemblyAI (YC S17) | Senior Backend Engineer | Full-time | Remote We're building the Top Rated Speech-to-Text API for Developers. Customers use our API to transcribe phone calls, meetings, videos, podcasts, and other types of media. Our accurate transcripts are used to power features like visual voicemail, call analytics, closed captioning, meeting summaries, and a slew of other features. We're growing fast, and pr…
Comment linking /
on “Ask HN: Who is hiring? (June 2021)”Submitted Build Your Own Imagen Text-to-Image Model as a story · 111 points
HN discussion1/2 ASSEMBLYAI (YC 17) | REMOTE | WORLDWIDE | FULL-TIME | https://assemblyai.com/careers We're a remote-first Deep Learning company focused on researching and applying advanced neural networks and machine learning technology. Our team is made up of problem solvers, innovators and top AI researchers with over 20+ years of experience in Machine Learning, NLP, and Speech Recognition from companies like DeepMind, Google…
Comment linking /careers
on “Ask HN: Who is hiring? (September 2022)”Submitted AssemblyAI: speech-to-text API as a story · 107 points
HN discussionI'm somewhat surprised neither this article nor the previous one mention anything about the Florence-2 model series. I had thought that Florence-2 was not just surprisingly capable for this kind of work, but also easily fine-tunable for a particular kind of document, when you expect to process a lot of instances of that document and want to further optimize accuracy. It's extremely small (0.23B and 0.77B parameters)…
Comment linking /blog/florence-2-how-it-works-how-to-use/
on “Why LLMs still have problems with OCR”Submitted Introduction to Diffusion Models for Machine Learning as a story · 98 points
HN discussionIt might be worth checking out https://www.assemblyai.com they let you build a custom audio model. One challenge with the audio from these radios is that it goes through some heavy compression. Traditional models will have a lot of challenges. Give a system that uses analog audio a try. The quality of the audio is a lot better.
Comment linking /
on “Transcribed police scanners in real-time”Submitted Fine-Tuning Transformers for NLP as a story · 67 points
HN discussionGreat question! There are three main parts to note: designing the videos, the backend for handling media, and the download renderer. The video designing process is a React app that leverages a Ruby on Rails based backend API. The React app handles the views, and if you look at the UI, you will see how the app steps are persisted in the URL. Our React app is based on Redux Toolkit, which is phenomenal. The Rails appl…
Comment linking /
on “Show HN: Podcast Audiograms – Make clips with captions, title and audio waves”Submitted Using JAX in 2022 as a story · 66 points
HN discussionThere's a comparison of JAX with PyTorch for Hessian calculation here! https://www.assemblyai.com/blog/why-you-should-or-shouldnt-b... Would definitely be interested in an article like that if you decide to write it
Comment linking /blog/why-you-should-or-shouldnt-be-using-jax-in-2022/
on “Differentiable Programming – A Simple Introduction”Submitted PyTorch Lightning for Dummies – A Tutorial and Overview as a story · 34 points
HN discussionAnd in terms of relative popularity of TensorFlow and PyTorch: https://www.assemblyai.com/blog/pytorch-vs-tensorflow-in-202...
Comment linking /blog/pytorch-vs-tensorflow-in-2022/
on “Meta's A.I. exodus: Top talent quits as lab tries to keep pace with rivals”Submitted Is Word Error Rate a Good Metric for Speech Recognition Models? as a story · 29 points
HN discussionRegarding general infrastructure, we are running on Heroku, AWS, and GCP for various things. I touched on the video APIs above, but regarding transcription - I have a lot of thoughts. We chose to work with AssemblyAI ( https://assemblyai.com ) after trying AWS's transcript service, RevAI and Google's Speech-to-text API. First, we started with doing manual transcripts through Rev, but the cost was unmanageable at sca…
Comment linking /
on “Launch HN: Milk Video (YC W21) – Edit online event recordings quickly”Submitted A Beginner's Guide to TorchStudio, the PyTorch IDE as a story · 16 points
HN discussionI could go pretty deep here, so let me know if I should elaborate on anything. The backend is a Ruby on Rails application that serves the frontend app's API. This interfaces with the user tables, database, and handles all the "state" of the app. The serverless stuff has changed over the months, but primarily it handles the stuff I don't want Rails to handle: file uploads, video processing and transcription. First, h…
Comment linking /
on “Launch HN: Milk Video (YC W21) – Edit online event recordings quickly”Submitted Universal-1: Robust and accurate multilingual speech-to-text as a story · 11 points
HN discussion
