Does using machine learning solve our problem of p-hacking and HARKing or do we have the same problems as with statistical tests and models?
Does using machine learning solve our problem of p-hacking and HARKing or do we have the same problems as with statistical tests and models?
When I read Andrej Karpathy’s endorsement of “context engineering” in a Twitter exchange with Shopify’s Tobi Lutke, I felt he tapped into something we all felt to some degree: tweet={"url":"https:\/\/twitter.com\/karpathy\/status\/1937902205765607626","author_name":"Andrej Karpathy","author_url":"https:\/\/twitter.com\/karpathy","html":"\u003Cblockquote class=\"twitter-tweet\" align=\"center\"\u003E\u003Cp lang=\"en\" dir=\"ltr\"\u003E+1 for
There’s a pervasive problem with semantics in artificial intelligence. It’s present at the creation – the term itself characterises the subject as a man-made simulacrum of something ‘natural’ the way we speak of artificial flavourings and artificial rubber.
How can we document software and computational analyses in such a way that others can convince themselves of their validity, and build on them for their own work? The question has been around for many years, and a number of attempts have been made to provide partial answers. This post provides a brief review and describes my own tentative answer, inviting you to play with it. Explainable AI is a hot topic today.
Thirty years after my first contact with computational (ir)reproducibility, I am happy to note that many things have improved. Reproducibility, computational and otherwise, is increasingly recognized as an important aspect of scientific quality control, and mostly considered worth striving for.
Reutilización CC BY 4.0
Quantum mechanics has led to a multitude of research fields that make up modern physics from the smallest (particle physics) to the largest scales (cosmology) and from basic science to highly applied research (quantum informatics). Research in these sub-disciplines has been progressing rapidly for decades and a wide variety of specialisations have emerged. It is therefore no wonder that it was researchers in high-energy physics who were looking for a “shortcut“ to new research results and were the first to come up with the idea of circulating preprints, i.e. articles that had not been (fully) peer-reviewed, between institutions. Over time, an official open access information service for the community developed: the INSPIRE-HEP specialised database
Der Beitrag Information on quanta and particles: TIB joins INSPIRE-HEP consortium erschien zuerst auf TIB-Blog.
Can creativity and science go together?
The ”ORKG Ask” service, launched in May 2024, is now a year old. The scientific search and discovery system provides answers from almost 80 million scientific documents and helps researchers find the scientific publications they are really looking for. Dr Allard Oelen and Dr Mohamad Yaser Jaradeh from the TIB developed the AI-based tool within four months. In this interview, they talk about the tool, its functions, the challenges faced during development and in the first twelve months, and which functions will be added in the coming months.
Der Beitrag Three questions to Dr Allard Oelen and Dr Mohamad Yaser Jaradeh on one year of ORGK Ask erschien zuerst auf TIB-Blog.
Research and science are international; it is not for nothing that we speak of international specialist communities. Although a service such as arXiv is operated by an institution based in the USA, namely Cornell University, it is used by researchers worldwide. Part of arXiv‘s funding has also been internationalised since 2010 with the introduction of arXiv membership. The TIB finances the German contribution together with the Helmholtz Association of German Research Centres (HGF) and the Max Planck Society (MPG). The TIB has now set up a so-called dark archive for the arXiv content in order to make the backed-up data accessible in the event that the data located in the USA is lost.
Der Beitrag Protecting Science: TIB builds Dark Archive for arXiv erschien zuerst auf TIB-Blog.
ICONCLASS is a classification system for image content. It is used in art history and collection documentation to describe the iconography of pictorial works in a standardised form. The online service is available free of charge and is – in terms of iconography – the only one of its kind. The TIB has been a member of the ICONCLASS consortium since April 2025. The aim of the consortium, which was founded in July 2024, is to ensure the long-term operation and further development of the free online service ICONCLASS through membership fees and input from the community.
Der Beitrag TIB membership in the ICONCLASS Consortium erschien zuerst auf TIB-Blog.