Technology Review 6/21: Why we need to make AI fairer and more open


Artificial intelligence is the most powerful technology of our time – and therefore also the most dangerous. Not because at some point it will take over the world and subjugate humanity. What makes it so dangerous today is the fact that it controls more and more of the services and apps that we use every day, training with data that is riddled with prejudice, discrimination and bias. Women receive fewer job advertisements than men, and black people are more likely to be accused of crimes than whites. These systems reproduce and cement social inequalities without our noticing.

And AI is becoming more powerful: Large language models such as GPT-3 generate texts that in part appear as if they were written by human hands. Very soon you could be controlling conversation systems, parts of the Internet search and much more. But malicious actors are also able to run semi-automated disinformation campaigns and thus spread more hatred and agitation with fewer staff and cause more destabilization than ever before.

For the title series of the new edition 6/21 of Technology Review (from Thursday on the well-stocked kiosk or online can be ordered) we spoke to researchers all over the world. You have recognized this “blind” danger and want to integrate more fairness, justice and openness into the AI ​​systems: Artificially generated data, various language models, deep learning algorithms in combination with rule-based systems – these are approaches that give rise to hope The power of artificial intelligence to tame. And they show that we are not simply at the mercy of newer technological developments.

In our second focus we are dedicated to the digital city. What advantages do cities really enjoy today through digitization? We didn’t find the highly networked metropolis in Germany, but we did find some projects that have potential – for example a city app in Karlsruhe that residents could use for all kinds of services, if the legal framework allows it. We will also show how networked sewage systems can help prevent sewage from simply being channeled into nearby rivers when it rains heavily. The intelligent water pipes also result in savings in terms of retention basins, for example.

More from MIT Technology Review

More from MIT Technology Review

More from MIT Technology Review

More from MIT Technology Review

Predicting the aroma of a coffee from green coffee beans that have not yet been roasted with a specially developed sensor and associated app? The Colombian-Israeli start-up Demetria wants to do just that. This means that they are not the only ones who want to “digitize” the popular pick-me-up. However, sensors, data and machine learning should not only ensure better coffee. You could also shake up the production conditions for the popular pick-me-up. But how well does it work to determine and even better control the complex aroma properties of coffee with digital tools?

This article is from issue 5/2021 of the Technology Review. The magazine will be available from July 8th, 2021 in stores and directly in the heise shop. Highlights from the magazine:

Back in 2014, a British research group estimated that 700,000 people around the world die every year because the bacteria they are infected with are resistant to the antibiotics they are treated with. While Covid-19 drew our attention to the threat posed by viruses, microbiologists have long been concerned that we are forgetting about bacteria like the plague pathogen Yersinia pestis. “Antimicrobial resistance may not seem as urgent as a pandemic, but it is just as dangerous,” said Tedros Adhanom Ghebreyesus, Director General of the World Health Organization, in November, calling it “one of the greatest health threats of our time.” What can we as a global community do from a global one Learn about a pandemic in order to get the next epidemic under control at an early stage?