Abdi Timer

Delivering personalised recommendations is key to engaging our customers, but traditional models can miss the complexity of user-content interactions. By integrating graph neural networks (GNNs), we’re exploring a cutting-edge approach to better predict clicks and enhance the shopping experience. In her post, Senior Applied Scientist Mariia Bulycheva explains how GNN features were integrated into the models powering the Zalando Homepage: https://bit.ly/4fNnQZC #ZalandoTech #InsideZalando #ZalandoScience #RecSys

44

1 Comment

Only 20% of organisations have a responsible AI program?! In a recent report by BCG and MIT, a staggering 30% of organisations didn't have any form of responsible AI programs. The rest (around 50%) "fell somewhere in between" which I would suggest also means they don't/it is an after thought Is responsible AI the new data governance? For so many years we have seen many organisations fully neglect data governance in the quest for innovation and quick value creation.... Then only to realise a few years later that their data is a complete mess, probably some very uncompliant practices... which in turn completely stunts any further value or innovation! Will we see the same thing with responsible AI? My concern is yes we will however the potential implications of not having it... could be far greater! #ai #ethics #responsibleai

8

🆕🚀🎧 Petros Zerfos and Hima Patel of IBM Research on how the open-source toolkit - Data Prep Kit - can transform how AI teams handle data. From cleaning raw text and processing code to extracting documents and integrating with Ray, this scalable solution simplifies data prep for large language models 🙌 Designed with cloud-native architecture and multimodal capabilities, DPK ensures high-quality input for AI projects at any scale. 🔗 https://lnkd.in/gSKywDde

1

🤖 𝗧𝗵𝗶𝗻𝗸𝗶𝗻𝗴 𝗕𝗲𝘆𝗼𝗻𝗱 𝘁𝗵𝗲 𝗕𝗼𝘁: 𝗪𝗵𝗮𝘁 𝗔𝗜 𝗦𝘁𝗶𝗹𝗹 𝗖𝗮𝗻'𝘁 𝗦𝗼𝗹𝘃𝗲 AI continues to dazzle with its capabilities, but there’s still plenty it can’t compute. Clare Walsh, Director of Education at IoA - Institute of Analytics, explores the gaps in AI’s reasoning and the crucial role of human expertise in bridging those limitations. Learn more about the future of AI-human collaboration and what we need to prioritise for the next wave of advancements. Read the full story, here 👉https://bit.ly/3ZDvh0G #AI #HumanIntelligence #Education #FutureOfWork

7

Our episode on the The Data Exchange Podcast with our fantastic host Ben Lorica 罗瑞卡 , and featuring my great colleague Hima Patel , on data preparation for LLM applications (RAG, fine-tuning, instruct-tuning, and pre-training) using the Data Prep Kit is out! Data Prep Kit is the open-source (Apache 2.0 license) framework and transforms for language and code, which we used in IBM Research to prepare the data for the training of #IBM 's #Granite models ( https://lnkd.in/eUvZVK2t ). Running at any-scale, from a laptop to a datacenter and handling from a few pages of text to hundreds of billions of documents, it is available in pure-#Python , #Ray , #Spark and with #KubeFlow pipelines. Feel free to check it out here: https://lnkd.in/ehGjpbzG #IBMResearch #AI #GenAI #LLM

42

2 Comments

Looking for a steer on your AI investment? My brilliant colleague and partner for our GCP capability Simon Poulton breaks it down in this great read 👇 sharing insights and his commentary on how to stay ahead of the game.

24

The UK has officially “gone all in” with AI. Earlier today, the UK Government announced their approval of the “AI Opportunities Action Plan”, written by Matt Clifford. And that is all 50 recommendations approved for implementation (read here: https://lnkd.in/etrBAyxm) The vision is clear: transform the UK from an AI adopter to an AI builder. (Which, in my opinion, is about time…) The UK government estimates this could add £400B to our economy by 2030. Here's how ⇊ ✅ Infrastructure Revolution ↳ 20x increase in AI compute capacity ↳ New AI Growth Zones, starting at Culham ↳ Next-gen nuclear power integration for sustainable AI ✅ Talent & Education Transformation ↳ Major expansion of AI education programs ↳ New flagship AI scholarship initiative ↳ Support for business upskilling programs ✅ Business Support Framework ↳ "Scan > Pilot > Scale" approach for AI implementation ↳ Dedicated AI Sector Champions in key industries ↳ New AI Knowledge Hub for resources and guidance My take, as someone in the AI upskilling space: The ambition is impressive. Training "tens of thousands" of AI professionals by 2030 isn't just about technical education – it's about creating people who can bridge the gap between AI capabilities and business value. What's encouraging? The focus on practical implementation over pure technical knowledge. The plan recognises we need both technical experts AND professionals who can translate AI into business results. The challenge? Building technical infrastructure is one thing; building human infrastructure is another entirely. The businesses that start upskilling their teams now will be best positioned to capitalise on what's coming. And this is something that will continue to drive my work with Edvance AI. 2025 will be a big year for AI - this is just the beginning. #artificialintelligence #upskilling #education

55

6 Comments

It would be great to join AIAI London tomorrow to discuss "Moving from PoC to value - delivering effective, value-driving AI at scale" Personally I am very curious to hear more about how AI is helping to accelerate drug discovery by Nikolay Burlutskiy and obviously learn from other leaders how they operate AI at scale! #ai #genai

29

Learn how Reinforcement Learning from Human Feedback (RLHF) is transforming the landscape of large language models (LLMs) like GPT-3. From feedback loops to best practices, Michał Oleszak's article dives deep into this technique. #AI #LLM

19

1 Comment

Google DeepMind launched Gemma 2, an open large language model available in 9 billion (9B) and 27 billion (27B) parameter versions. It outperforms larger models, providing cost-effective deployment options. Core Innovations Gemma 2 features sliding window attention, soft-capping, and knowledge distillation. Integration and Compatibility Gemma 2 integrates seamlessly with major AI frameworks, supporting Hugging Face Transformers, JAX, PyTorch, and TensorFlow via Keras 3.0. It runs efficiently on various hardware, from gaming laptops to cloud setups. Deployment and Access Developers can access Gemma 2's model weights from Kaggle and Hugging Face. Starting next month, deployment on Vertex AI will be available, with model integration options in Google AI Studio and local environments using Gemma.cpp. Safety and Evaluation Google DeepMind implemented rigorous safety measures, including data filtering and comprehensive testing, to mitigate biases and risks in Gemma 2. Academic Support The Gemma 2 Academic Research Program offers Google Cloud credits for research use, with applications open through August 9.

43

3 Comments

AI Opportunities Action Plan: UK Government Invest £14bn in AI and Data   🌍 What does this mean for the UK?   💡 AI Growth Zones to boost tech infrastructure with a 20x expansion in computing capacity by 2030  💡 Access to diverse and robust data sets consolidated into a “National Data Library” for secure R&D from testing to training  💡 13,000+ new AI professionals by 2030 through education, training programs, and partnerships with universities and industry Read more: https://lnkd.in/eYCk3b8a

53

1 Comment

🤯 Want to know how LLMs think? Gemma Scope is here to peel back the layers! 🧠 Gemma Scope 🔍 is a new tool that helps us understand how large language models (LLMs) work. It's like a microscope for LLMs! 🔬 By examining each layer of the model, we can see how it processes information and generates text. 📝 Key Points: ✅ Transparent LLMs: Gemma Scope makes LLMs more transparent, helping us understand how they work. ✅ Layer by layer: It analyzes each layer of an LLM, revealing how they contribute to the model's overall behavior. ✅ Interactive Demo: You can try Gemma Scope yourself with an interactive demo! 🎮 Curious about how LLMs can transform your business? Book a free call today:https://lnkd.in/dZZYP6yk Link: https://lnkd.in/gYznJgaz #gemmascope #llm #ai #ml #airesearch #aiforrealestate #tech #trending

2

"We surveyed UK respondents, asking them to list (without prompting) the three factors they thought most likely to cause human extinction in the next 100 years. ● We estimate AI risk to be reported by 9% of British adults, at a similar level to Natural disasters (10%). ● The relative ordering of risks aligns well with previous polling from YouGov in 2023, using an alternative methodology. Climate change (58%) and Nuclear weapons/WMDs (42%) were top concerns, followed by Pandemics (37%). ● Some demographic effects were quite large, such as that men were substantially more likely than women to raise the issue of nuclear weapons. ● Polling suggests AI risk is not a wholly niche issue, but is substantially less of a concern to the public than cases such as nuclear weapons or the climate." Rigorous and relevant work from Rethink Priorities, David Moss and Jamie Elsey

30

3 Comments

👀 Discover how GPU acceleration is transforming causal inference. Using RAPIDS cuML with DoubleML, you can tackle large datasets up to 12x faster. ➡️ https://nvda.ws/3YRM4eQ

160

1 Comment

We recently held the BCI and WBE for Safe AI Workshop. Diogo Di Lucena's (AE Studio) talk reviewed 'Self Modeling in Neural Networks.' Di Lucena discussed how he and his collaborators implement self-modeling in neural networks, inspired by Graziano's attention schema theory, to reduce model complexity and foster prosocial behavior in multi-agent systems. By creating simplified models of one's attention and the exact mechanisms used to model others' attention, the team demonstrates that self-modeling reduces complexity, making it easier for agents to model the system and potentially enhancing prosocial behavior. Evaluating their method on various tasks and models, the authors find consistent results in complexity reduction, suggesting that their approach, grounded in neuroscience, is distinct from other regularization methods and applicable across diverse contexts to improve model complexity and prosocial behavior in multi-agent settings. Learn more: https://lnkd.in/gFtdec-m

2

🎉 AISI's Systemic AI Safety Grant Call is Now Open for Applications! 🎉 The UK’s AI Safety Institute are seeking applications focused on a range of safety-related problems: this could involve monitoring and anticipating AI use and misuse in society, or the risks they expose in certain sectors. We want to see applications that could enhance understanding of how government could intervene where needed - with new infrastructure and technical innovations - to make society more resilient to AI-related risks. We expect to offer around 20 grants up to £200,000 each in this phase, and particularly welcome projects bringing together researchers, industry and civil society. Find out more information and full guidance on how to apply: https://lnkd.in/euCYmY8S Working closely with UK Research and Innovation and The Alan Turing Institute, we’re supporting projects that advance this new area of research.

13

1 Comment

Missed our recent webinar on building with Large Language Models? Don’t worry—you can catch up anytime! 👋 Meet Tim Thompson, Senior Full Stack Engineer at Cambridge Spark, alongside our panel of AI experts, as they dive into: ✨Practical insights on working with LLMs and how to apply them ✨Overcoming real-world challenges with LLMs ✨Expert tips for professionals and beginners working with these cutting-edge models. Whether you’re already integrating LLMs into your data projects or just exploring the possibilities, this is a must-watch! 🤖 🔗 Watch it here on demand: https://lnkd.in/ehSvDZ8G #AI #LLMs #LargeLanguageModels #LLM #ArtificialIntelligence #MachineLearning #Data #DataAnalysis #CambridgeSpark

16

Researchers at Google released Gemma Scope, a tool that reveals how each layer in Gemma 2-family large language models responds to input. By using over 400 sparse autoencoders (SAEs) to interpret embeddings from different layers, Gemma Scope probes how distinct concepts are represented within the model. This approach could help answer fundamental questions about how language models process information, adapt through fine-tuning, and handle different types of prompting. Read our summary of the paper in #TheBatch: https://hubs.la/Q02PHtlZ0

154

3 Comments