longo

On 19 August 1989, some 20,000 people gathered in a field near the Hungarian city of Sopron, close to the border with Austria, for an outdoor party. “A brass band boomed across the field. Goulash cooked in giant pots over open flames; beer and wine were there for the taking. People danced around a bonfire.” The festival, jointly organised by several Hungarian opposition parties, was billed as a pan-European picnic, a celebration of brotherhood between nations. The revellers included a large contingent of refugees from the…

At a time when we have become bleakly accustomed to political capital being made of militarising borders and building walls, it is a timely corrective to read a book devoted to the romance of the alternative. The Picnic re-examines events in Hungary in 1989 that precipitated the collapse of Soviet power in central Europe. In particular, it recreates, through intimate personal histories and eyewitness recollection, the ways in which one idealistic, grassroots protest – the staging of a summer party in a field near the…

Image created by author using Dall-E 2The key to building transparent AI software systemsAn AI algorithm is a software system that has the ability to automate or perform tasks that typically require human intelligence. These systems often include more than just a trained model. They can also include explicit algorithm functionality, such as business rules, that integrate the model’s output into the larger AI system to complete an end-to-end task.To properly implement ethical AI systems, the software used to deploy the…

Image by authorThe fundamentals of feature stores and a few tips on how and why you should monitor themSince Uber first introduced the concept in 2017, the feature store has been steadily gaining popularity as a tool to support data scientists and machine learning engineers with the ability to define, discover, and access high-quality data for their machine learning projects.Diagram by authorIn machine learning projects, raw data is collected, cleaned, formatted and mathematically transformed into data called a “feature.”…

Image by authorHow to move away from retraining at a set cadence (or not at all) in favor of a dynamic approachThis piece was co-authored by Trevor LaVialeWhile the industry has invested a lot in processes and techniques for knowing when to deploy a model into production, there is arguably less collective knowledge on the equally important task of knowing when to retrain a model. In truth, knowing when to retrain a model is hard due to factors like delays in feedback or labels for live predictions. In practice, many…

Image by authorOpinionDon’t throw in the towel on centralized machine learning (ML) teams yet; here’s how to get it rightOver the last decade, the industry has gone from celebrating the rise of the “central ML team” to questioning whether it should exist. I can’t help but feel like I’m watching Rome burn. It doesn’t have to be this way.As the emerging field of machine learning operations (MLOps) continues to grow rapidly and new tools and techniques proliferate, the potential for confusion and discord — even among…

The Australian Securities Exchange said on Wednesday it would delay an overhaul of its main share trading software by up to 20 months, citing problems achieving "scalability and resilience". The delay until late-2024, among several, takes it to nearly a decade of ASX Ltd's project to modernise its equities clearance software Clearing House Electronic Subregister System (Chess). The CHESS software enables ASX's registry to manage settlement of transactions, and the exchange has strived since 2017 to replace with…