【英文】2022年新兴科技趋势报告（161页）

Advanced machine learning is only possible with signifcant  amounts of data—ofen user data that includes personally  identifable information. Both the collection and the use of this  data raise compliance and ethical risks that companies may  need to address.  One of the main risks is the use of data for purposes that  were not anticipated at the time the data was collected. This  may occur when companies purchase or leverage thirdparty datasets to train their algorithms. From a commercial  perspective, the key is to provide appropriate notice, and  secure appropriate consent, when collecting the personal  information—or ensure that a supplier has done the same,  even though the AI supply chain is increasingly complex.As AI technology becomes more deeply embedded in a wide  variety of tools to support decision-making, reports have  emerged alleging that the algorithms, data, and models used  in these systems may demonstrate biases. 

There has been  an increase in lawsuits against companies and organizations  that use “black box” AI tools to provide guidance for decisions  related to areas such as employment, consumer credit, or  criminal justice.Other developments indicate views might be shifing in  those jurisdictions, however. For example, in 2021, the U.K.  Intellectual Property Ofce (UKIPO) released a report on  consultations it had conducted regarding the impact of AI on  intellectual property. Included in the report was the suggestion that UKIPO could eventually seek legislation allowing AI tools  to be recognized as inventors. According to a UKIPO ofcial,  “We recognise that AI systems have an increasing impact on the  innovation process. We want to ensure the intellectual property  systems support and incentivise AI-generated innovation.”