How values influence decisions in science
By Vanessa Seifert 29 April 2021
Empirical evidence is not always sufficient to determine the models we use
Science involves making choices. Which hypothesis should be put to the test? Which model should be used to describe a system; and which approximations and assumptions should be enforced? Different factors figure in this decision-making process. Do we want the model to produce numerically accurate results that closely agree with our experiments, or to make new predictions? Should the use of approximations result in simple and understandable representations, or should they be justified by our background theories? Such considerations reflect the expectations scientists have for their results. Looking closely at these expectations reveals that scientists are guided by how they value and perceive the different functions of science. Read more here.
Chemical Exposure in the Human Body
Although the population is exposed to chemicals, there are no measurements that permit the situation in Switzerland to be evaluated.
There is no doubt that people are exposed to a cocktail of chemicals on a daily basis. In Switzerland, the extent of this exposure and its impact on our health is poorly quantified or completely unknown. The reason for this is simple: no figures are available. This makes it impossible to monitor the way exposure is developing and to evaluate the efficacy of measures that are adopted.
The Chemical Products Division is working on a national biomonitoring project with the aim of collecting samples from a representative section of the population (in terms of age, language, etc.). Read more here.
Why manufacturing Covid vaccines at scale is hard
By Anthony King - 23 March 2021
The first Covid-19 vaccine candidate went into the arms of volunteers in Seattle in March 2020. It was an mRNA vaccine from Moderna. The mRNA candidate from BioNTech and Pfizer followed in April. By December 2020, these two had become the first vaccines to be approved by the US Food and Drug Administration (FDA). Hot on their heels are rivals based on adenovirus vectors from AstraZeneca and Johnson & Johnson, as well as Sputnik V from Russia.
Early successes in developing vaccines by upstarts like Moderna and BioNTech papered over the struggles of vaccine heavyweights like Merck, GSK and Sanofi. But those companies that have surmounted the challenges of development now face the next phase: manufacturing doses on an enormous scale. And as Matt Hancock, the UK’s health secretary, told the House of Commons on 18 March ‘the process of manufacturing vaccines is complicated, and subject to unpredictability’. Read more here
Coronavirus: Next phase of reopening on 19 April
Bern, 14.04.2021 - The Federal Council is continuing its strategy of taking cautious, gradual steps towards reopening. At its meeting on 14 April, it decided on further reopening measures. From Monday, 19 April, it will again be possible to hold events with audiences and spectators subject to restrictions, for example at sports stadiums, cinemas, theatres and concert venues. Indoor sports and cultural activities will also be allowed, as well as certain matches and competitions. Restaurants will be able to reopen outside seating areas. Although the situation remains precarious, the Federal Council deems the risk associated with these reopening steps to be acceptable. Read more here
DENSE REPRESENTATION LEARNING
Embed your items, not just words!
Following our previous posts on recent progress in Natural Language Processing, we discuss a follow-up idea: can we extend the concept of word embeddings to any collection of items, possibly unordered? More precisely, can we learn representations from item sets, such as the product baskets in online retail or music playlists on streaming platforms? As we will see, the answer is yes, representation learning can also be applied on such datasets. Read more here.