News of the week selected by Impactscool – March 30th
30 March 2020 | Written by La redazione
The human footprint on the environment does not end in the atmosphere, but also leaves a heavy mark on the soil. If we were able to restore and safeguard the soil, we could be able to absorb more than 5 billion tons per year, roughly how much the United States emits at the same time. Research published in Nature Sustainability has revealed that if we could manage the soil properly it could contribute to a quarter of the absorption of CO2 that occurs on dry land. The good news is that to allow this process, the only thing you need to do is leave the ground as it is, in this way about 40% of this capacity would happen naturally. The bad news is that to feed a growing population that wastes a third of the food it produces, more and more land is needed for agriculture (to produce food for humans and farm animals). Will we be able to respond adequately to this challenge?
According to a new study conducted by Stanford Engineering researchers, the technology behind the major automatic speech recognition systems in the US makes twice as many errors in the interpretation of words spoken by African Americans than in the interpretation of the same words pronounced by whites. The system takes advantage of artificial intelligence that is trained through databases, if these contain partial data, therefore, in this case, with a few words spoken by black people, the AI will suffer from these problems. With the ever more widespread diffusion of this technology, used for the management of banking systems, for looking for a job or for the transcription of legal processes, a failed system could lead to serious consequences in the life of people belonging to ethnic or non-ethnic minorities. mother tongue.
Google has announced a new real-time transcription feature for its Translate app for Android phones. The feature will allow users to get instant translations of speeches, lectures or monologues live in one of – for now – eight languages, including English. Until now, Translate only allowed the conversion of relatively short speech fragments. The function is not yet available for Italian but it shouldn’t take long for the implementation of our language.
Drones are increasingly popular and not just for recreational use. For each drone that is used to make videos or stunts, there are as many that are used by law enforcement, firefighters, to carry out monitoring operations and much more. Despite the ever better performances there is one thing that these devices are not able to do at their best, avoid obstacles. A research team from the University of Zurich has developed a camera that can manage images in an intelligent way to understand when an object launched towards a drone (or towards which the drone is going) has a trajectory that would lead to a collision, ordering move to the on-board computer. The on-board camera only records the moving pixels, in this way the data load is lower and the processor is able to detect the threat in just 3.5 milliseconds against the 40 normally needed, sometimes not enough to avoid the impact.