Krzysztof Zarzycki

Being #OpenSource has many advantages. For #NVDA has opened the way for community contributions, and has enabled #transparency, #security and #innovation beyond what might have been possible in closed software. Increasingly, governments are also mandating the use of open source. Here is an article on such a step forward in Switzerland: "Switzerland Makes Open Source Software Mandatory For Public Sector" https://lnkd.in/gFEu-k-w #FOSS #NVDAsr #Accessibility #Software #News

11

Auditor app version 86 released: https://lnkd.in/grpWDkqS See the linked release notes for a summary of the improvements over the previous release and a link to the full changelog. Forum discussion thread: https://lnkd.in/gAiPgXrk See https://lnkd.in/dV5s8sb and https://lnkd.in/db9ZhvZ for info about the app and optional monitoring service. #GrapheneOS #privacy #security #android #attestation #VerifiedBoot #MeasuredBoot #HSM #SecureElement #auditor

15

Stellar Repair for Outlook Review : A Reliable Tool for Corrupted PST Files They store everything—emails, calendars, contacts, attachments, journals etc. However, like other files get corrupted, PST files are also prone to corruption due to various factors, which we will be discussing further in this guide. #Stellar #stellarrepair #outlook #PSTFiles #corruptedfiles #communication #communitysupport #datarecovery #userexperience #algorithm #emailrecovery #scanpst #outlooktips #onlinetools https://lnkd.in/d6YRip4e

5

📢 New Blog Alert! Columnar File Readers in Depth: APIs and Fusion Columnar file readers have come a long way, but what if we could take them even further? In this insightful post by Weston Pace, we dive into LanceDB's innovative "read task" API that separates scheduling from decoding, unlocking powerful capabilities: ✅ Flexible batches with low memory ✅ Seamless fusion of decode & query ops ✅ Optimal I/O & CPU parallelism Discover how this groundbreaking approach outperforms traditional APIs and enables true operator fusion. 🔗 Read more: https://lnkd.in/gDT4T4kJ #ColumnarFormats #DataProcessing #PerformanceOptimization #LanceDB

12

Prepare for valuable insights from Matteo Maria Terzuolo at #ftcon24usa 🚨 Dive into #Dart's Pattern Matching – explore how it simplifies complex data structures with clear patterns & practical tips. https://lnkd.in/dFaaGUzi

1

With all the buzz in the industry today regarding #LLMs with their encoders, decoders, multi-headed attention layers, and billions (soon trillions) of parameters, it’s tempting to believe that good #AI is solely a matter of model design. Unfortunately, this is not the case. Good AI requires more than a well-designed model. It also requires properly constructed training and testing #data. In this post, AI/ML SME Keith Pijanowski introduces the concept of data-centric AI, a term first coined by the folks at Snorkel AI, and talks how to use Snorkel Flow with MinIO to create a training pipeline for your AI workload. Check it out. https://hubs.li/Q02HsLnn0

7

1 Kommentar

With all the buzz in the industry today regarding #LLMs with their encoders, decoders, multi-headed attention layers, and billions (soon trillions) of parameters, it’s tempting to believe that good #AI is solely a matter of model design. Unfortunately, this is not the case. Good AI requires more than a well-designed model. It also requires properly constructed training and testing #data. In this post, AI/ML SME Keith Pijanowski introduces the concept of data-centric AI, a term first coined by the folks at Snorkel AI, and talks how to use Snorkel Flow with MinIO to create a training pipeline for your AI workload. Check it out. https://hubs.li/Q02N4D-x0

8

In his blog series, #ZPartner Andreas Reibenbacher discusses various warnings and error messages that may arise when using #CDC Delta, despite the availability of comprehensive documentation, and offers practical workarounds. This series aims to share these experiences to help save your valuable time. In this particular blog, the focus is on the use of WHERE clauses in conjunction with the automatic CDC mechanism. https://lnkd.in/gUajSwzs

9

Missed the last virtual #BevyEngine meetup? The recordings are finally up, listen to Florian sharing his insights into: "Truck Instrument Cluster powered by Bevy" https://lnkd.in/eBiVVk6D #rustlang #bevyengine 🦀

3

Auditor app version 79 released: https://lnkd.in/gFEMkqDA See the linked release notes for a summary of the improvements over the previous release and a link to the full changelog. Forum discussion thread: https://lnkd.in/gB7tF6r6 See https://lnkd.in/dV5s8sb and https://lnkd.in/db9ZhvZ for info about the app and optional monitoring service. #GrapheneOS #privacy #security #android #attestation #VerifiedBoot #MeasuredBoot #HSM #SecureElement #auditor

6

The latest update for #Cigniti includes "98% Accuracy in Leak Detection: How #AI is Transforming HVAC Manufacturing" and "Top 8 Use Cases for #QA Validations in Trade Finance". #Testing #DigitalAssurance https://lnkd.in/eu3qhTi5

1

Autonomous vehicles, medical AI, children exposed to technology — in high-risk AI applications, can better benchmarking help protect consumers? What areas are in critical need of benchmarking? A recent article examines the need for open benchmarking to help consumers gain confidence in AI. It calls for “a new approach to benchmarking—one that is as adaptive and nuanced as the technologies it seeks to evaluate.” We’ve been working on open benchmarks for performance in STEM subjects, university-level math, and detecting AI-generated text. Here are some of our principles:  - Benchmarking datasets aren’t useful if they’re too simple (yes, multiple choice questions are too easy).  - Good benchmarks use complex prompts in niche areas to test the limits of LLMs.  - Subject-matter experts are the best qualified to develop prompts and evaluate answers. Synthetic approaches using LLMs won’t cut it. - The more aspects used for evaluating responses, the better. In our benchmark for natural sciences, we rated each answer in 8 categories. Share your perspective! https://lnkd.in/gN2dTean #AIsafety #ResponsibleAI #OpenBenchmarking #ForbesTech

11

1 Kommentar

[Blog] Redemption: A Prototype for Automated Repair of Static Analysis Alerts David Svoboda https://lnkd.in/eY9MmKyW "In this post, I give an overview of how Redemption uses ACR technology, the kinds of errors Redemption can fix, how the tool works, and what’s next for its development." #staticanalysis #automated

3