Graphcore’s Post

✈️ Want to see how we make LLMs fly? Join me this Wednesday at @GraphcoreGdansk where I'll be diving into transformer optimization tricks and the real cost of inference! We'll try to make sense your favourite LLM's API pricing, and (hopefully) demo some of those features with vLLM. Come geek out with fellow AI enthusiasts! 🤓 Register: https://lnkd.in/dnccPzpK

💻 📖 📑 Research support: #Caribbean #AI #UnitedNations Our office is looking for online volunteers to support research on Caribbean artificial intelligence readiness. 3 weeks, 6-10 hours per week. We are looking for seven Online Volunteers who have experience in research with excellent writing skills. The research conducted will feed into a forthcoming study on the artificial intelligence readiness of the Caribbean. The study will be based on existing research and data from a large variety of sources, and summarize the information along seven categories of readiness. The study is intended to be short and exploratory, and will provide basic information, from a subregional perspective, on artificial intelligence readiness, and outline some possible ways forward. We anchor our work in the overall principle of the 2030 Agenda to “leave no one behind” and SDG 10 “reduce inequality within and among countries”. More details here: https://lnkd.in/eQHM_Fmk

📅 AI and Copyright: Policy, Challenges, and Perspectives 📅 Interested in the future of AI and copyright, the plans of the next Commission, practical tools, and political initiatives? Join us at the DICOPO – Digital Constitutionalism & Policy event on June 20, 2024! 🔍 The #CopyrightDirective foresees a text & data mining exemption for crawling or scraping websites, except when opt-out mechanisms are used. The #AIAct mandates transparency of AI model providers regarding their training data and policies to be put in place to respect IP. How is this all being implemented? What's next? I am honored to be moderating this panel discussion with esteemed colleagues and key players on this critical topic. Key Speakers and panelists: 🎉 Axel Voss - Congratulations to Axel Voss on being re-elected as a Member of the European Parliament for 2024-2029! He has been a pivotal player in both the Copyright Directive and the AI Act! We are curious to hear what is top of his political mind for the next mandate in view of the interrelation on AI and IP. 👨💼 Giuseppe Abbamonte - Director of the Media Policy Directorate at DG CNECT. Giuseppe Abbamonte oversees the implementation of the Copyright Directive and engages with model providers and copyright owners at this intersection. What is the state of play of the opt-out exemption in the Copyright Directive and the code of practice for general-purpose AI models transparency in the AI Act, foreseen to be ready in 9 months? 🛡️ Cullen Miller - Cullen Miller is vice president for policy at Spawning, which created the #DoNotTrain registry, providing a machine-readable opt-out solution for rights holders, partnering with AI labs such as Hugging Face. What are the challenges to practical solutions? 📜 Andrea Renda - leader of CEPS (Centre for European Policy Studies) and influential expert on #AIregulation. What other laws need to be considered and are currently discussed (#ChatGPT taskforce European Data Protection Board), what legal challenges do AI system providers face? Event Introduction and Conclusion: 👋 Oreste Pollicino, President of DICOPO – Digital Constitutionalism & Policy, will introduce the event. 💡 Reflections by Paolo Cesarini and Giovanni De Gregorio will conclude with key takeaways and future outlooks in AI and copyright policy. Mark your calendars for June 20, 2024, and join us in exploring these critical issues! #AI #Copyright #DigitalPolicy #AIAct #Innovation #Transparency #IntellectualProperty

🎉 Exciting News! 🎉 We are thrilled to announce that our paper, "Optimization Framework for Semi-supervised Attributed Graph Coarsening," has been accepted at the Uncertainty in Artificial Intelligence (UAI) 2024 conference. In data-intensive applications, graphs serve as foundational structures across various domains. However, the increasing size of datasets poses significant challenges to performing downstream tasks. To address this problem, techniques such as graph coarsening, condensation, and summarization have been developed to create a coarsened graph while preserving important properties of the original graph by considering both the graph matrix and the feature or attribute matrix of the original graph as inputs. However, existing graph coarsening techniques often neglect the label information during the coarsening process, which can result in a lower-quality coarsened graph and limit its suitability for downstream tasks. To overcome this limitation, we introduce the Label-Aware Graph Coarsening (LAGC) algorithm, a semi-supervised approach that incorporates the graph matrix, feature matrix, and some of the node label information to learn a coarsened graph. Our proposed formulation is a non-convex optimization problem that is efficiently solved using block successive upper bound minimization(BSUM) technique, and it is provably convergent. Our extensive results demonstrate that the LAGC algorithm outperforms the existing state-of-the-art method by a significant margin. A huge thank you to my supervisor Prof. Sandeep Kumar for his invaluable guidance. Additional, thanks to Prof. Ruchir Gupta and my co-authors Subhanu Halder and Archit Kane for their dedication and hard work. https://lnkd.in/giUk7dnp #UAI2024 #GraphTheory #MachineLearning #Research #AI

The subject of digital capitalism has troubled me for quite some time. That's why I chose to engage in a course created by https://www.tni.org/en on this matter. A program that enlightens political advocates and aids in comprehending the subject on a more profound level. So, today: How to Fight the Big Tech?! Insights from the Master Class on "Big Tech and the Digital Overlords" 1. Understanding Big Tech's Rise to Power: Big Tech companies have ascended to dominance through various strategies, including monopolization, rapid expansion, and acquiring or eliminating competition. Their stronghold extends across sectors like digital advertising, social media, eCommerce, search engines, and mobile operating systems. 2. Sources of Big Tech's Power: Big Tech's power emanates from its control over data, enabling practices like surveillance capitalism and influencing consumer behavior. Additionally, financial rents are extracted through market domination and imposing high fees on developers and sellers within their ecosystems. 3. Challenges in Regulation: Despite regulations against monopolies, Big Tech circumvents them through extensive lobbying and financial influence. Regulatory attempts have often fallen short, necessitating stronger enforcement and innovative solutions. 4. Intellectual Monopolies and AI: Big Tech's control over intellectual property and AI development further solidifies its dominance. By monopolizing AI technologies and research networks, they stifle innovation and maintain their grip on the industry. 5. Strategies for Change: The session emphasized the importance of coalition-building, public pressure, worker mobilization, policy advocacy, and democratic planning in challenging Big Tech's dominance and fostering a fairer digital landscape. 6. Addressing Environmental Impact and Tax Evasion: Efforts are underway to raise awareness about Big Tech's environmental footprint and hold them accountable for tax evasion practices. 7. Utilizing Government Structures: Advocacy efforts target government policies promoting data privacy, transparency, and accountability, while also addressing concerns related to government surveillance. Big thanks to the speakers: - Borgen de Silva: [Margarida Ramos Borges da Silva](https://lnkd.in/eundazgZ) - Cecialia Rikap: [Author of 'Capitalism, Power, and Innovation: Intellectual Monopoly Capitalism Uncovered'](https://lnkd.in/e-E5F-t9) - James Schneider: [Co-convenor with Uni Global Union of the Make Amazon Pay coalition](https://lnkd.in/et_3kFnK) Let's continue the conversation and work towards a more equitable digital future! #BigTech #DigitalOverlords #TechEthics #ChangeAgents

Our partner in Brazil, InternetLab - human rights and internet policy think tank, conducts interdisciplinary and impactful research on issues involving law and technology, especially internet policy. Their work aims to foster collaboration and dialogue between academics, businesses, civil society organisations, and government officials.   InternetLab's work is crucial to developing public policies that prioritise and safeguard human rights in the digital age, especially given the growing influence of social media platforms in Brazilian society and the rapid advancement of AI.   Learn more about them in the carousel below, and read our interview with InternetLab's Heloisa Massaro on how Brazil prepared for the local elections in 2024: https://lnkd.in/e5tRvw_s

#LatestNews 📢 On the 4th of Novermber, Glenn Micallef, the Commissioner-designate for or Intergenerational Fairness, Youth, Culture, and Sport, will face his confirmation hearing in the European Parliament. While the official questionnaire covers essential issues like Creative Europe's future, working conditions and AI's impact on the arts, we believe it's crucial to also address culture’s role in Europe, asking questions that go beyond specific policy issues. In our latest blogpost, read the three questions we think the cultural sector should be asking: ➡️ https://ow.ly/qyIB50TLBjk

My new article on @botpopuli presents, in simple, non-academic language, a call to debate Latin America beyond a dichotomous vision between the Global North and South in the field of technology: https://lnkd.in/dVxpUD2w Although alliances can be tactically built with national companies in the face of political pressure from large platforms from the Global North to deregulate local markets, it must be analyzed that not all “Big Data from the South” is inherently emancipatory. A policy of national digital champions without criteria beyond the dubious character of “national company”, which disregards the complex relations of globalized capital, ownership and control, is not truly compatible with the needs of Brazilian society and could even deepen its inequalities. I'd like to take this opportunity to spread the word about the excellent work of Botpopuli and datasyn, their newsletter where the article also appeared, one of the few technology portals that frequently gives space to authors from the third world.

Hi folks, I want to share this research using artificial intelligence! Details: 📌We must debug the content, we can´t just copy and paste. 📌Pay attention to the prompt, it's crucial to give context. 📌Explore beyond the initial option.

"Thrilled to have presented the Google PageRank Algorithm in a 5-minute poster competition among experienced researchers! It was an incredible opportunity to simplify and showcase a complex topic in such a dynamic setting. Grateful for the chance to learn from and engage with experts in the field. 🚀📊 #Research #GooglePageRank #DataScience #PosterPresentation #LearningExperience"