The study also highlighted a culture of indifference or hostility towards AI ethics, with product managers viewing such considerations as detrimental to productivity, revenue, or product launch timelines. Ethical issues are often only addressed late in the development process, making it difficult to make necessary adjustments. The report also noted that the prioritization of engagement or performance metrics over ethics often requires irrefutable quantitative evidence, which is challenging to produce given that existing data infrastructures are not designed for such metrics.
Key takeaways:
- Stanford University researchers have found that tech companies are failing to live up to their promises to support the ethical development of AI, with safety often taking a back seat to performance metrics and product launches.
- Despite publishing AI principles and employing teams to work on AI ethics, many companies have not prioritised the adoption of ethical safeguards. Workers involved in promoting AI ethics report a lack of institutional support and often face indifference or hostility.
- Ethical considerations are often only taken into account late in the development process, making it difficult to make adjustments to new apps or software. Frequent reorganisation of teams also disrupts ethical considerations.
- Quantitative metrics of ethics or fairness are hard to define and are not prioritised, as existing data infrastructures within companies are not tailored to such metrics.