RSS Real World Data Science

Nowcasting, an amalgamation of the term now and forecasting, has been used by all kinds of organisations from banks attempting to track the impact of fiscal policies to humanitarian agencies gauging the number of people displaced by a disaster. Now researchers at the European Central Bank and Banque de France demonstrate a 3 step process that can make nowcasting estimates adaptable and more accurate, says Atmajitsinh Gohil

Have you read our series on AI? With articles from Anna Demming, Diego Miranda-Saavedra PhD, Fatemeh Torabi, Lewis Hotchkiss, Emma Squires, Olivia Varley-Winter, Isabel Sassoon, PhD, Julia Lane, Lesley Hirsch, Adam Leonard On everything from optimising AI performance to the challenges AI poses to the labour market 📖 https://lnkd.in/ei7SSusg

Optimising AI for direct users and applications can be knotty enough but how do you optimise AI developments for society as a whole? In the last instalment of this article series on AI we look at the governance strategies that can tackle the problems emerging now - not just the potential workforce impact, but the carbon cost of training large language models, like non-consensual gen-AI porn aggravating online gender inequality, and a widening digital divide disadvantaging pupils, workers and citizens who cannot afford all the latest AI tools, among others – to ensure progress in AI technologies brings desirable progress for everyone.

Throughout our series on AI we’ve looked at how to optimize AI models by dealing correctly with the data requirements, evaluation and ethical considerations. All good on paper but real world constraints can get in the way and currently most AI projects fail. Here a panel of experts in AI talk about how best to work with it within real world conditions and swell the minority of projects that succeed.

Expanding the deployment of Artificial Intelligence across various sectors has been widely touted as a significant threat to jobs but so far data to justify these fears or guide a response to them has been lacking. An “Industry of ideas” that stems from a concept for which Paul Romer share the 2018 Nobel Prize in Economics may offer some insights. https://lnkd.in/g8DauTyW Julia Lane, Lesley Hirsch, and Adam Leonard look at an approach piloted in the US based on the industry of ideas that may offer useful reclassification of workforce data, link funding to workforce outcomes and provide the local level detail needed to navigate potential AI disruption in the workforce

It took just sixteen hours for Microsoft’s shiny new chatbot Tay to be shut down for profanity. The real world can be a dangerous place for an AI algorithm, full of unforeseen data, evolving behaviours and significant human consequences. Isabel Sassoon, PhD highlights some of the hurdles to navigate when testing and assessing whether a model is fit for purpose.

We all know AI development and deployment can pose an ethical minefield, from criminal sentencing biased by race in the US, to students systematically downgraded in public examinations and decisions to rescind food welfare riddled with errors. Spreading some less common knowledge in the latest in our AI series, Olivia Varley-Winter describes some approaches for handling AI ethics and how everyone from data scientists to government policy makers has a role to play.

Data, data, data is often the mantra for robust AI model development, but data quality is often more important than data quantity and can throw up some interesting challenges As discussed in the latest in our series on AI ⬇

In my NHS role I've been thinking about long term health forecasting, working on our Dynamic Population Model. We've written up where we are at in a Royal Statistical Society blog, thanks to many people but especially Brian Tarran on his editorial help even after officially finishing his RSS role to ensure it's over the line and published. https://lnkd.in/ea3dYEs9

Today’s state of the art large language models can be pretty impressive so how do they compare with human level intelligence? In the second instalment in our series on AI, Diego Miranda-Saavedra weighs up what they can and can’t do and how we might position them with respect to our own capabilities