News and views
Blogs, columns and opinion from our contributors and editors
OpenAI’s text classifier won’t calm fears about AI-written homework
Educators are worried about ChatGPT being using by students for homework assignments, so OpenAI has released a tool to classify whether text is human- or AI-written. But relying on the classifier’s results is ill-advised, as some basic statistics shows.
US legislators get their data science act together
A bill introduced in the US Congress wants to make funds available to develop data science and data literacy education across the United States. We sit down with education and policy experts to discuss the challenges and opportunities ahead.
Using ‘basket complementarity’ to make product recommendations
Purchase suggestions – e.g., “if you are buying that, you might also want this” – are, to a large extent, informed by the concept of complementarity: that certain products are often bought and/or used together. A journal paper by Puka and Jedrusik sheds light on how these product recommendations can be derived, as Moinak Bhaduri explains.
Data science can help close the ‘digital skills’ gap, or so it seems
A ‘digital skills’ gap is harming employer productivity and growth, according to a survey by engineering body IET. But the ‘digital skills’ that are needed sound a lot like data science skills: statistical understanding, data analytics, AI and machine learning.
ChatGPT can hold a conversation, but lacks knowledge representation and original sources for verification
ChatGPT represents a next step in the evolution of large language models, says Detlef Nauck. However, there are still major challenges - and concerns - to overcome.
Pulling patterns out of data with a graph
Large volumes of data are pouring in every day from scientific experiments, so much so that it is now commonplace to perform dimension reduction in order to reduce a large number of measurements to a set of key values that are easier to visualize and interpret. Enter ‘The Sequencer’, a proposed method to find trends within high-dimensional datasets.
Determining the best way to route drivers for ridesharing via reinforcement learning
A/B testing is often used to evaluate the impact of design ‘treatments’ — for example, are people who see advert A more likely to buy something than those who see advert B? Classical methods typically assume that changing one person’s treatment will not affect others, but what if that’s not the case? A paper by Shi et al. aims to address this problem.
A chat with ChatGPT
‘Hello there! I’m a large language model trained by OpenAI, so I don’t have the ability to experience emotions or have a physical presence. I’m here to provide information and answer questions to the best of my ability. Is there something specific you would like to know?’