The RLHF Book Really happy to see a brand new free online book on RLHF! It is not every day we can find comprehensive information about an emerging topic like Reinforcement Learning from Human Feedback: This little online free-book on RLHF offers a concise overview of all-things Reinforcement Learning from Human Feedback (RLHF) for production ML use-cases on its role in post-training language models. It provides quite a comprehensive overview across the multi-step process of training a base model to follow instructions, collecting human preference data to build a reward model, and applying reinforcement learning techniques to fine-tune the model’s behavior. The RLHF book contains a few chunky chapters on 1) framing the problem space, 2) optimization of LLMs (policy gradients, rejection sampling, instruction training, etc), 3) Advanced topics, and 4) Open questions. Check it out! |
|
|
|---|
|
Music AI Generation & Lawsuits This weekend I was playing with GenAI Music Generation models, and I was blown away that I was able to create songs that are not only relalistic but that were actually not terrible! In the background of this, there has been some mind-blowing developments in this space, not only with the technology but also with the respective lawsuits. It is worth doing a revisit of the cutting‐edge research that enables music generation such as Google's MusicLM, Meta's MusicGen, and many others highlighted in this research survey. Similarly it is interesting to see the copyright lawsuits that are appearing towards music generation services such as Riffussion, Suno and Udio. This is a really exciting area of research, but at the same time it's a critical phase where industry is fast evolving to fit into the new-world of generative AI systems with copyright and IP, but also with creativity and innovation at high stakes. |
|
|
|---|
|
Programming Changing as We Know It The one and only Tim O'Reilly weights in on how software engineering will disappear as we know it - putting into words what we know already: software engineering won't go anywhere, but it will evolve as it has throughout the last few decades: This is quite a nice reflective piece which revisits a decade of evolving software development practices, and how we can extrapolate towards the impact of LLMs in software development, and more importantly how productionisation of ML will also drive a lot of the changes due to the new challenges that arise different to traditional code. Especially for practitioners like all of us it will be important to stay ahead to future-proof our careers as we are going through what seems to be a major shift and evolution into the field. |
|
|---|
|
New FC Foundation Model from China China strikes again with another Foundation model that shakes the world with a surprise! This time it is a time-series forecasting Foundation model which challenges the tech giants and their previous large model releases. Often in foundation models for time-series forecasting, the approach taken on their tokenization approach tends to have a significant impact on the performance (for better or for worse), and this model introduces a native continuous patch tokenization and a novel TimeFlow Loss based on flow-matching to generate flexible, probabilistic forecasts without relying on discrete tokenization, which is really interesting compared to other competitors. Of course, we also would not be able to talk about a foundation model, without a massive dataset, and this model carries a 1 trillion+ time-series points of training data, which enables for pretty impressive zero-shot performance on both point and probabilistic forecasting benchmarks vs other models. This is an exciting space to keep an eye for! |
|
|
|---|
|
Opinions from a Decade in Software This reflective piece revisits a decade of evolving software development practices. A lot of the lessons really resonate with me, so it is definitely a worth read as it tackles quite a lot of observations that reflect the state of development as of today. Some of the key points challenge the allure of over-engineered solutions and trendy tools, advocates for good-old typed languages, and reminds us that soft skills are as important as hard skills in the programming world. It is quite a short read but definitely worth skimming through, as these are at the very least good reminders of some useful lessons from years of developmnet. |
|
|
|---|
|
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 2025:
In case you missed our talks:
|
|
|---|
| | |
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. - Kompute - Blazing fast, lightweight and mobile phone-enabled GPU compute framework optimized for advanced data processing usecases.
- CuPy - An implementation of NumPy-compatible multi-dimensional array on CUDA. CuPy consists of the core multi-dimensional array class, cupy.ndarray, and many functions on it.
- Jax - Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more
- CuDF - Built based on the Apache Arrow columnar memory format, cuDF is a GPU DataFrame library for loading, joining, aggregating, filtering, and otherwise manipulating data.
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! |
|
|---|
| | |
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: - MLSecOps Top 10 Vulnerabilities - This is an initiative that aims to further the field of machine learning security by identifying the top 10 most common vulnerabiliites in the machine learning lifecycle as well as best practices.
- AI & Machine Learning 8 principles for Responsible ML - The Institute for Ethical AI & Machine Learning has put together 8 principles for responsible machine learning that are to be adopted by individuals and delivery teams designing, building and operating machine learning systems.
- An Evaluation of Guidelines - The Ethics of Ethics; A research paper that analyses multiple Ethics principles.
- ACM's Code of Ethics and Professional Conduct - This is the code of ethics that has been put together in 1992 by the Association for Computer Machinery and updated in 2018.
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!
|
|
|---|
| | |
| | | | The Institute for Ethical AI & Machine Learning is a European research centre that carries out world-class research into responsible machine learning. | | | | |
|
|
|---|
|
|
This email was sent to You received this email because you are registered with The Institute for Ethical AI & Machine Learning's newsletter "The Machine Learning Engineer"
|
| | | | |
|
|
|---|
|
© 2023 The Institute for Ethical AI & Machine Learning |
|
|---|
|
|
|
The Institute for Ethical AI & Machine Learning
