New AI proposal: INDIA
#EU has the #AIAct (first in the world). #USA has recommended the hiring of a Chief AI Officer for every Federal Agency. But, right now, I love the new India CAS framework proposal for AI. Most importantly it has a section "Ensuring Human Control through 'Manual Overrides' and 'Authorisation Chokepoints". So important as we increasingly delegate decisions to AI without having a moment to reflect with common sense or even think through unintended consequences or tradeoffs (not just risks!!!!). #humanagency
Second, under Distinct Accountability (separate point from Transperancy and Explainability), it talks of liability for AI malfunctions and unintended outcomes! So important - we often talk of profits, commercialisation and reach (assuming it means the same as benefits). I would have loved to see planetary #sustainability highlighted here too. #accountability #liability
Third, it talks of mandated standardised reporting of failures and incidents (as opposed to just #usecases). While the number of AI incidents is increasing the reporting is miniscule. We also haven't a real clue on how many incidents are bug fixes (oversight, unknown, negligence), just malicious ones (since accountability is diffused in this case). #AIincidents
Fourth, it also highlights the need for a specialised, agile, regulatory body! #agilegovernment frameworks are so critical. #Regulations can never keep up with the pace of #technologies.
While it does not mention AI systems, the lens through which AI is viewed is complex adaptive thinking. Well done #India....now to pass the regulation and implement it. Also wishing for some diversity in the team :) #AIS #AIsystems #Systemthinking #complexadaptivesystems
More articles like this at www.melodena.com
