Issue #313 - The ML Engineer 🤖
Alejandro Saucedo
Tech Executive @ Zalando | Chair/Advisor @ UN, EU, ACM, etc | Join 60k+ ML Newsletter
Thank you for being part of over 60,000+ ML professionals and enthusiasts who receive weekly articles & tutorials on Machine Learning & MLOps 🤖 You can join the newsletter for free at https://ethical.institute/mle.html ⭐
If you like the content please support the newsletter by sharing with your friends via ✉️ Email, 🐦 Twitter, 💼 Linkedin and 📕 Facebook!
This week in Machine Learning:
Ilya Sutskever's NeurIPS keynote is up - an interesting reflection on the decade since the seminal “Sequence to Sequence” work that helped ignite the modern era of large-scale neural NLP: This is a comprehensive session where Ilya recounts how the original Seq2Seq approach established a template for present-day AI: big models plus big data equals breakthroughs. Over time, this scaling principle was validated far beyond translation, culminating in today’s GPT-style models. However, it seems we are exhausting the “fossil fuel” of internet-scale data, and future progress will hinge on new techniques - eg. agents interacting with their environments, generating synthetic data, and improved reasoning capabilities.
Google drops the mic last week releasing Gemini 2.0 with a bunch of new features on their AI studio, doubling down towards "agentic" AI with multimodal input/output: As expected Google is building on the initial Gemini 1.x foundation, extending long-context and multimodality capabilities, improves latency and performance, and introducing features like native image and audio generation. Something that comes across as novel is the integration into Google’s products and ecosystem, however the race continues to move at breakneck speed so we can only expect similar pace from the tech ecosystem.
Tiktok has published the architecture behind their recommendation system called "Monolith", a real-time massive-scale recommendation system designed specifically to address production challenges such as large-scale, sparse, and dynamic feature spaces: The paper provides interesting insights such as collisionless embedding table based on Cuckoo hashing, which enables dynamic inclusion and eviction of new model features. Monolith tightly integrates training and serving which they highlight as one of the reasons they can allow for fast online updates so that the model can adapt to changing user behavior within minutes.
Meta Fundamental AI Research (FAIR) has quietly shared a huge release last week with new open-source AI systems across CLIP, Motivo and Seal: Meta released Motivo, a foundation model that enables embodied humanoid agents to efficiently solve complex tasks without additional training. Meta also released Video Seal, a robust watermarking solution for videos that remains intact through common transformations. They’ve also introduced Flow Matching, a hierarchical byte-level tokenizer-free approach (Dynamic Byte Latent Transformer) for generative modeling. Additionally, they released a new version of Meta CLIP, which improves on previous versions for vision-language alignment. Quite surprising and exciting to see so much movement from META's research arm furthering the research ecosystem across quite a few of these interesting areas.
OpenAI has finally released their Text-to-Video SORA model as a public offering! As per the usual naming convention, this comes with Sora Turbo, focusing on fast generation of higher fidelity videos across multiple aspect ratios, and up to 20-second. These services continue to surprise us with the quality of the video generation, certainly still with quite some limitations (such as many posts showing the limits when rendering scenes from gymnastics, etc). The model is still imperfect, struggling with complex sequences and realistic physics - however it is great to see finally OpenAI is releasing to encourage community input and iteration norm-setting, and responsible use.
Recommended by LinkedIn
Upcoming MLOps Events
The MLOps ecosystem continues to grow at break-neck speeds, making it ever harder for us as practitioners to stay up to date with relevant developments. A fantsatic way to keep on-top of relevant resources is through the great community and events that the MLOps and Production ML ecosystem offers. This is the reason why we have started curating a list of upcoming events in the space, which are outlined below.
Upcoming conferences where we're speaking:
Other upcoming MLOps conferences in 2024:
In case you missed our talks:
Open Source MLOps Tools
Check out the fast-growing ecosystem of production ML tools & frameworks at the github repository which has reached over 10,000 ⭐ github stars. We are currently looking for more libraries to add - if you know of any that are not listed, please let us know or feel free to add a PR. Four featured libraries in the GPU acceleration space are outlined below.
If you know of any open source and open community events that are not listed do give us a heads up so we can add them!
OSS: Policy & Guidelines
As AI systems become more prevalent in society, we face bigger and tougher societal challenges. We have seen a large number of resources that aim to takle these challenges in the form of AI Guidelines, Principles, Ethics Frameworks, etc, however there are so many resources it is hard to navigate. Because of this we started an Open Source initiative that aims to map the ecosystem to make it simpler to navigate. You can find multiple principles in the repo - some examples include the following:
If you know of any guidelines that are not in the "Awesome AI Guidelines" list, please do give us a heads up or feel free to add a pull request!
About us
The Institute for Ethical AI & Machine Learning is a European research centre that carries out world-class research into responsible machine learning.
Kubernetes & Cloud Native Engineer
1wAlejandro Saucedo, the rapid evolution in machine learning is truly remarkable, from Gemini 2.0 to Sora. These innovations are reshaping our technological landscape.