How to talk about "AI" without adding to the anthropomorphization

The op-ed outlines a structured approach to replace anthropomorphizing language related to AI with more accurate descriptions. This includes categorizing common phrases and suggesting alternatives that focus on the functionality of systems rather than attributing human-like qualities.

One major change is the recommendation to avoid using terms like 'artificial intelligence' in favor of more descriptive phrases such as 'probabilistic automation.' This shift requires a conscious effort to adapt language habits, which may initially feel clunky but can lead to clearer communication about the technology's capabilities.

The authors emphasize placing agency with users rather than the systems themselves, which can help clarify roles and responsibilities in interactions with technology. For example, instead of saying 'the model shows bias,' the recommended phrase is 'the model reflects bias,' which accurately indicates that the responsibility lies with the data and the users.

While adopting these new language practices may require effort, the long-term benefits include improved understanding and discussions about AI technologies. This is particularly relevant for engineers who build or operate these systems, as precise language can lead to better documentation and training materials.

Overall, the op-ed advocates for a cultural shift in how we communicate about AI, aiming to foster a more accurate and responsible dialogue around the technology. Engineers and developers are encouraged to practice these new conventions to enhance clarity and reduce misconceptions in the field.