Tinius Digest February

ChatGPT detected if the written essays were generated by itself with an accuracy of over 92 percent. Two other systems for detecting AI texts failed to do so and only had an accuracy of 20 percent.

Plagiarism detection may need to shift its focus from similarity checks to verifying the origin of content. As evidenced by this study, possibly AI tools can offer a more straightforward yet effective solution by predicting if the text is produced by AI or not.

40 out of 50 essays composed by ChatGPT demonstrated a remarkable level of originality, stirring up alarms about the reliability of plagiarism check software used by academic institutions in the face of recent advancements in chatbot technology.

The researchers emphasise that applying large language models in education, such as the OpenAI ChatGPT and Google Bard AI, offers numerous possibilities to improve student’s educational experience and facilitate teachers’ tasks.

Google Search is the go-to news discovery platform for U.S. adults—regardless of age. Almost 50 percent use Google to start researching a significant news event.

In February, 14 percent of Gen Z adults reported using TikTok to start researching major news events, significantly higher than the share of all adults (2%) saying the same. Gen Z also prefers Instagram substantially more than older generations.

To boost their chances of counting Gen Zers as paying subscribers, news publishers must distribute more compelling content that lives outside of their websites, the study concludes.

The industry’s current flux regarding data collection may present an opportune moment for regulators to identify new ideas and alternative ways of doing digital advertising with less tracking.

Much academic research has demonstrated that the digital advertising status quo significantly impacts privacy, democracy, society and the environment. There is limited evidence to suggest that the efficiency and efficacy gains to advertisers and publishers of this system outweigh the societal impact.

Alternative models can rely on less personal data. Contextual advertising and local profiling are models that depend on less collection of personal data. Still, it is unclear whether they are effective compared to the current digital advertising model. This could make it more difficult for advertisers and publishers to justify investing in alternative models.

The Digital Services Act will introduce provisions for individuals to have access to more transparency over ad targeting. However, this is unlikely to address issues related to indicating and applying targeting preferences, especially the challenge of trying to do this across the complex and opaque ecosystem of companies that process individuals’ data for advertising purposes.

After five minutes of training, a classification model reliably identified over 55,000 VR users by hand and head motion alone. The accuracy was 73 percent after ten seconds and 94 percent after 100 seconds.

To preserve privacy in VR, there is a need to develop secure privacy technologies that may be deployed to enable the use of VR without revealing private user information. 

To preserve privacy in VR, there is a need to develop secure privacy technologies that may be deployed to enable the use of VR without revealing private user information. 

81 percent feel that it is unavoidable that they will lose some privacy in the future because of what new technology can do. An equal number of people (81%) believe social media companies have too much power.

60 percent fear technical progress is destroying our lives. This number is considerably lower in Denmark (49%) and Sweden (41%) and more prevalent among younger people. At the same time, connectivity has become an integrated part of people’s lives worldwide. 71 percent can’t imagine life without the internet. 

The level of concern about privacy varies hugely from country to country. The most indifferent inhabitants are primarily located in Asia and Latin America. The most sceptical is situated in Europe and North America—with the noticeable exception of Japan, South Africa and South Korea.

In 2022, illicit addresses associated with illegal activity received $20.6B (SEK 214m/NOK 214m). This amounts to a share of 0.24 percent of total cryptocurrency values. Remember that this only includes identified addresses and doesn’t capture proceeds from non-crypto native crime involving cryptocurrency as a mode of payment (e.g. conventional drug trafficking).

From 2020 to 2022, the sums associated with identified unlawful activities increased by 145 percent. So do illicit transaction activities related to sanctioned entities (e.g. entities based in North Korea and Russia). In 2022 this amounted to 43 percent of all illegal transactions.

The value of ransomware payments fell significantly from 2021 to 2022 as more victims refused to pay. The lack of transparency surrounding ransomware attacks makes it difficult to estimate the total extent—but the share of ransomware payments to identified addresses was down 40 percent from 2021 to 2022.