The data access dilemma
Hi, I'm Marie Roehm from Apheris . Ready to dive into the world of governed, secure, and private ML? Then let's go
What's ahead in this newsletter?
Learning Hub
News from Apheris
Industry Round-up
Enjoy reading
Learning Hub
How to unlock the potential of AI in MedTech: Navigating the data access dilemma
AI is revolutionizing medical technology. But what we’ve learned from working with MedTech organizations is that the full potential of AI remains largely untapped within this industry. Why is this? At the heart of the answer lies the important role of data, which is required to train and customize AI models. Find out why MedTech organizations can’t simply access their machine generated data to improve their AI models in a repeatable and scalable way:
Building for EU AI Act compliance
Everyone knows it’s coming; no one is really prepared. Concerns about privacy and risks associated with the recent AI boom are growing globally. In response, policymakers have voted in favour of the EU AI Act in June 2023. While the exact details of the Act are not yet crystal clear, the underlying structures that organizations will need to put in place to develop and deploy safe and ethical AI, are.
Read our e-book for a structured view of the Act’s risk-based approach, the urgency to prepare, and insights into achieving compliance with Apheris.
For more on this topic, this EU-Startups article shares insights from our CEO, Robin R?hm, ?on how companies can get ahead of the complexities? that lie ahead.
Preventing data leaks with federated learning using NVIDIA Flare
The key goal of federated learning is to carry out data analytics and machine learning on distributed data without sharing or centralizing it. NVFlare, a powerful federated learning framework, built by NVIDIA, allows for just this. Apheris integrates NVFlare so you can easily run NVFlare jobs through Apheris Compute Gateways. The combination of NVFlare and Apheris Compute Gateways offers a powerful federated infrastructure with a focus on empowering data custodians to meet their strict compliance requirements. The Compute Gateway allows the data custodian to control every aspect of computations on sensitive data with computational governance, as well as enterprise-grade security and privacy measures. Visit our product page to learn more about the Apheris Compute Gateway.
领英推荐
News from Apheris
Switzerland is calling – and the Apheris team is there to answer! Our team will be at Basel at this year’s BioTechX.
Join Eric Boernert, from Roche Informatics, and our CEO, Robin R?hm for their talk on how governed, secure, and private computational access to federated multi-model data can be utilized to facilitate connection and multidisciplinary collaboration for a global research community.
Britta Srivas, Customer Solutions Engineer at Apheris, was a speaker at this year’s NHS Data Conference. In her talk, she shared insights into her work, which focusses on federating highly impactful AI models for healthcare in a governed, secure, and private manner. Through federation, those AI models are trained on real-world patient data from our customers around the world to accurately represent the global population. ?
Follow us on LinkedIn and watch out for her session recording.
Industry round-up
Data Protection Impact Assessment for AI – practical considerations
The EU AI Act is due to be finalized by the end of 2023 and will have a massive impact on how AI is developed and deployed in the EU. Fines of up to 6% of the offender’s global turnover or €30 million (depending on whichever is higher) await companies that are not ready to comply from 2027. Read more about the compliance pillars of the EU AI Act
AI in clinical trials
Pharmaceutical companies are increasingly turning to artificial intelligence (AI) to speed up patient recruitment for clinical trials and minimize the number of participants needed. This strategic shift is particularly relevant as human trials are the costliest and most time-consuming facet of drug development. The use of AI can cut clinical trial timelines by up to half. However, the effectiveness of AI depends on the quality and availability of data, and less than 25% of health data is publicly available. Discover more on the transformative impact of AI on the clinical trials landscape.
How to assess qualities of AI products without sharing IP-sensitive information
Before AI products are released to the market, organizations often want (and soon will need) to assess their quality in terms of fairness, safety, truthfulness, and copyright. However, most AI products either enter the market without having undergone such assessments, or evaluators are granted excessive access, leading to privacy, security, and IP risks for AI owners.
In the first of a new series of blogs, OpenMind explores alternatives to traditional external assessments, focusing on external and remote model access for auditing AI systems.
The series will continue to explore issues such as copyright, bias, the alignment of values, illegal content, and privacy-enhancing technologies (PETs).
Wishing you all a happy and cosy autumn!
Marie