Rezultati pretraživanja
  1. Great keynote at day by . Access to data is the big challenge here - maybe preserving techniques like can help. What if we had federal agencies hosting data and evaluating models privately?

  2. prije 11 sati

    Apple, Tesla to Lead ‘Edge’ Computing, Says Guggenheim - Barron's -> Imagine this for -> Happening at ⁦🚀🚀

  3. prije 11 sati

    Everyone moving to the edge, learning decentralized and privacy at its maximum-> Apple's Latest Deal Shows How AI Is Moving Right Onto Devices

  4. 2. velj

    Newer privacy-enhancing like could mean that companies & researchers don’t really need to collect individual data – it can happen in a depersonalised way, writes on : (🎨 via ) 📵

    Comic panel with text 'Secure aggregation* enables the server to combine the encrypted results, and *only* decrypt the aggregate.

Comic below slashed in half, one has yellow bg and eye looking towards the other half with green bg and a lock, which has a kaleidoscopic triangle in the middle of it.
    Comic panel where 2 people are talking to each other. Person 1, holding a pen in each hand: On each device, before anything is sent, the secure aggregation protocol adds zero-sum masks to scramble the training results. When you add up all those training results —

Person 2, clasping hands: — the masks exactly cancel out! Nice.

In front of Person 2 is a mirror with arrows (obscured with waves) going towards Person 1.
    Person 3: Ok, so the server can't see any single phone's results. But what if one phone has really unique data? Could that data be compromised by showing up inside the model?

Blue bg with Person 3 looking up through a magnifying glass at the process of encrypted data from phones going into an aggregated kaleidoscopic triangle, and from there to secure aggregation in a circle.
    Person 1: Well, it's possible, but we don't want it to happen. For machine learning to work best, models need to capture the common patterns in the data, not memorize things that are specific to one phone.

Person 4, looking at the securely aggregated data (connected in a circle to each other, every data set represented as a differently coloured ball) through a microscope and surprised (!): Wait a second, is that me?
  5. enables devices to use state of the art without centralising and have by default.

  6. 31. sij

    AI generated drugs are becoming real. Our co-opetitive project gathering 10 pharma with is now one step closer to having a strong impact on drug discovery!

  7. 31. sij

    4/5 Dr. Aline Talhouk on the ethics of in the age of : building “LEAP” using as a pilot project.

    Prikaži ovu nit
  8. 30. sij

    . tries to tackle these issues on three pillars 1. Collective voice: align stakeholders 2. Infrastructure: data and 3. Providing services: portals, tools, etc.

    Prikaži ovu nit
  9. 30. sij

    I will be presenting a poster on : we implemented a federated logistic regression (, Li et al., 2015), also using as a backbone. If you want to find out more, drop by and say hi

    Prikaži ovu nit
  10. 30. sij

    .’s delegation is on its way to . We will be presenting demos on (using our infrastructure ), , and projection of real world data into

    Prikaži ovu nit
  11. 29. sij

    Every day could and should be ! Let's think about ways to address systems of data usage that inherently take privacy into account: See our use cases for , , & how we anchor data privacy advancements into everything we do!

  12. 29. sij

    On Monday we worked on our corporate values ​​and a new action plan to convey those messages. On a mission to ensure privacy in AI, shared values ​​are what makes our technological development run smoothly.

  13. 29. sij
  14. 29. sij

    making headlines again! It is great to see that companies who invest a lot in tools and technologies also invest in and

  15. 29. sij
  16. 29. sij

    Check this GREAT interview of our CTO sharing about our stack and how our team is enabling capture of medical data with neural nets, computer vision, NLP as well as using to serve our core values of ownership and privacy preserving.

  17. 28. sij
  18. 28. sij

    allows data-privacy compliant deployment of in enterprises and has rightfully become one of the most relevant AI topics. works on making it universally applicable and already applies it to automated invoice processing.

  19. 28. sij

    The latest tech term being thrown around is . It's all about training algorithms at the edge.

  20. 28. sij

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.