Karpathy on Software Evolution Software is undergoing its most fundamental shift in decades, and Andrej Karpathy defines how it’s redefining how we build and interact with machines: This is a great new talk from Karpathy on how "Software Is Changing (Again)", covering the new era where LLMs act as programmable computers coded in natural language. He argues that LLMs are still in their "1960s mainframe" phase (which is a funny analogy) drawing parallels to operating systems. Karpathy emphasizes the need for partial autonomy, fast human-AI collaboration loops, and LLM-friendly infrastructure, and it is interesting to see the urgency he pushes for practitioners to design for agents and master the full software stack from code (1.0) to learned weights (2.0) to prompts (3.0) as we rebuild computing from the ground up. |
|
|
|---|
|
Google on Securing AI Agents Google takes on GenAI security principles, concepts and best practices with a new report on how to secure your AI agents: This whitepaper from May warns that LLM-powered agents carry critical vulnerability risks across rogue tool calls and inadvertent data leaks, and argues that teams must embed security from day one via three pillars. 1) Assign a clearly authenticated human controller. 2) Grant only dynamically scoped least-privilege permissions. 3) Log every input, plan, and action for auditability. Google’s own “hybrid defense-in-depth” approach supposedly introduces deterministic policy engines (hard blocks, spending caps, user confirmations) with reasoning-based guard models and adversarial training to catch novel attacks, then keeps both layers sharp through continuous red-teaming, regression testing, and bug-bounty feedback. |
|
|
|---|
|
Anthropic on Building AgenticSys Anthropic worked with dozens of teams building LLM agents and released a set of best practices for building agentic systems with simple composable patterns rather than complex frameworks: This is a useful resource for production ML teams looking to find the fastest path to reliable agentic systems by leveraging LLMs orchestrators (ie base model plus retrieval, tool-calling, and memory). Anthropic proposes to introduce measurable improvements to drive outcomes with prompt chaining for fixed sub-steps, routing for distinct input classes, parallelisation for speed or consensus, orchestrator-worker hierarchies when subtasks emerge on the fly, and evaluator-optimizer loops for iterative refinement. There's also a section on full autonomous agents (LLM + tool loop with environmental feedback) for open-ended problems where step count and order are unpredictable. It is interesting to see how these paradigms evolve throughout, however keeping simplicity at its core, continually A/B-testing each added layer and culling complexity that doesn’t move key metrics. |
|
|---|
|
State of Eng Leadership 2025 The State of Engineering Leadership in 2025; management layers # decreasing; negative sentiment on AI productivity +60%; positive improvements in talent & recruitment; really important insights you don't want to miss out: 617 engineering leaders were surveyed across critical topics in tech, including talent trends, productivity, sentiment, AI and more. There are some important trends such as layoffs and hiring freezes easing yet recession anxiety (65%) and vanished entry-level openings shifting demand to mid-senior talent. Orgs are focused on both flattening and bolstering management, whilst remote work is decreasing, and leaders juggle more teams and more pressure while burnout climbs. Despite intense hype, 60% say GenAI hasn’t boosted output and 51% call its net impact negative, hampered by quality, security, and workflow hurdles, even though there are some narrow wins emerging in automated code generation, refactoring, and documentation. Hiring remains challenging, but practitioners who marry rigorous ML-in-prod skills with strong communication and cost-aware reliability will stay in demand and can unlock real value by applying GenAI to focused, high-leverage tasks. |
|
|
|---|
|
Quantum Computing Lecture Notes Quantum computing is an exciting emerging field intersecting with AI, and this 200+ page lecture notes distil two decades of theory into a comprehensive roadmap: This is quite a comprehensive overview that introduces key concepts in quantum computing starting with qubits, gates and staple algorithms (Deutsch–Jozsa, Simon, Shor, Grover), then detailing ML-relevant subroutines such as the quantum Fourier transform, Hamiltonian simulation and the HHL linear-solver that yield polylog-time speed-ups for sparse, well-conditioned matrices. The more advanced chapters arm you with complexity tools (generalised adversary bounds, QMA, Local-Hamiltonian) to recognise when quantum gains are impossible, and practical sections provide great intuition on variational circuits, PAC learning from quantum data, error-correction and fault-tolerance. Although today’s hardware is still noisy, the theoretical stack for future quantum acceleration of large-scale model training and optimisation is already well mapped out and there's quite interesting potential in the horizon. |
|
|
|---|
|
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
