3 days agoMember-onlySCENE: Reasoning about Traffic Scenes using Heterogeneous Graph Neural NetworksEncoding the driving scene in an autonomous driving system is a very important task that has not yet been solved completely. This work is one of the most recent ones that I found interesting among the many researchers who are working to encode driving scenes using graph neural networks. …Autonomous Cars9 min readAutonomous Cars9 min read
Published in AIGuys·Jan 27Member-onlyRecipes for building an open-domain chatbot: BlenderBot — v1In the previous blog post, we talked about Meena, a Google AI chatbot. In this post, I will review BlenderBot, the chatbot developed and open-sourced by Meta. I will go through the next versions of the BlenderBot in the next few posts. Let’s get started! Introduction BlenderBot is an open-domain and…Naturallanguageprocessing8 min readNaturallanguageprocessing8 min read
Published in AIGuys·Jan 17Member-onlyMeena — Towards a Human-like Open-Domain ChatbotNote: This blog post is written with the help of ChatGPT! I decided to start a series of blog posts on chatbots and conversational agents, diving into the world of AI-powered language models and their capabilities. In my first post, I review Meena, the conversational agent introduced by Google AI. …Deep Learning6 min readDeep Learning6 min read
Published in AIGuys·Jan 14Member-onlyChinchilla: Training Compute-Optimal Large Language ModelsThe article investigates the optimal model size and the number of tokens for training a transformer language model under a given computation budget. They found that current large language models are undertrained, and by training over 400 language models with varying parameters and the number of tokens, they found that…Deep Learning5 min readDeep Learning5 min read
Published in DevOps.dev·Jan 14Member-onlyMLOps project — part 4b: Machine Learning Model MonitoringWe reviewed Evidently.AI and Seldon ALIBI Detect for model monitoring in our previous blog post. I will go through managed services on Google Cloud and AWS to do the model monitoring task. I like to emphasize the importance of MLOps parts using the following photo from a paper published by…Machine Learning13 min readMachine Learning13 min read
Jan 11Member-onlyShow Your Work — SummaryNote: You can find the book here. “Show Your Work!” is a book by Austin Kleon that aims to help readers learn how to share their creative process, works, and ideas with others. In the first chapter, Kleon argues that in today’s world, it is important for creatives to share…Startup8 min readStartup8 min read
Published in AIGuys·Jan 8Member-onlySpeech Recognition — Review State-of-the-art Papers — Part 1In this series of blog posts, I will review state-of-the-art papers in the field of speech recognition. WAV2VEC: UNSUPERVISED PRE-TRAINING FOR SPEECH RECOGNITION The aim of this paper is to improve supervised speech recognition using unsupervised pre-training. It is therefore possible to exploit unlabeled audio data, which is much easier to collect than labeled data. Wav2vec…Speech Recognition7 min readSpeech Recognition7 min read
Jan 7Member-onlyReinforcement Learning from Human Feedback, InstructGPT, and ChatGPTNote: some parts of this blog post are generated by ChatGPT! :) Welcome to my blog post on ChatGPT! In this post, we will dive into the inner workings of ChatGPT and how it is trained. However, before we get into the specifics of ChatGPT, it’s important to first review…Machine Learning8 min readMachine Learning8 min read
Jan 5Member-onlyWho Not How — SummaryIf you want to go fast, go alone. If you want to go far, go together. “African Proverb” I recently read the book “Who not How” written by Dr. Benjamin Hardy and Dan Sullivan. Dr. Benjamin Hardy is an organizational psychologist and best-selling author of Willpower Doesn’t Work and Personality…Leadership5 min readLeadership5 min read
Published in DevOps.dev·Jan 4Member-onlyMLOps project — part 4a: Machine Learning Model MonitoringNote: The majority of the information in this blog post comes from the documentation and the talks of the tools’ authors and creators. I simply put them together. The details of each company’s solutions depending on the information contained in their documentation and online discussions. After years of research and…Machine Learning12 min readMachine Learning12 min read