RIYADH: The importance of transparency and responsibility when using artificial intelligence came under scrutiny at the Global AI Summit.
Ali Hosseini, chief technology officer for PWC Middle East, told Arab News the consultancy company had created “Responsible AI” — an approach to managing risks associated with AI-based solutions.
The initiative gave customers a clear picture of how the company used their data, he said.
“We take this to customers and we actually share the experience in terms of how we’re using it internally. So there are a number of areas in terms of general education for the employees and (in) what kind of cases they can use AI and in what kind of context they can depend on the output,” he explained.
Hosseini said companies thinking of implementing AI must ensure “the empowerment of the employees, self-responsibility, and AI use.”
“We give (employees) the right tools coming from the right kind of credible sources to use on the day-to-day automation of tasks or augmenting their knowledge,” he told Arab News, concluding the interview with a key takeaway.
“(There is) a level of self-responsibility that people need to basically take an education in order to use AI … We advise the organization to use (AI) as giving you a basically augmented decision, but not the full decision … The decision-making point is always the human, not AI.”
Ali Hosseini, chief technology officer for PWC Middle East
In a panel discussion at the summit, Priya Nagpurkar, Vice President of the Hybrid Cloud and AI Platform at IBM Research, said AI was created to “enhance and support human capacity, intelligence and expertise and not to replace it, and do so in a very transparent, explainable and responsible way.”
IBM has created watsonx.governance, an AI and data platform which monitors, directs and manages organizations’ AI activities.
Similar to a “nutrition label,” IBM creates factsheets which document AI model metadata across its lifecycle.
“These factsheets are a way of extracting the key facts that went into the data curation in that part of the lifecycle,” she explained. “(A) concrete example is, let’s say you are building an AI application to look at loan applications. The type of facts you want to know about the model you want to use are if there was bias in the data that went to training that model, was it evaluated? And was there a range of variation?”
The GAIN Summit, organized by the Saudi Data and AI Authority, takes place from Sept. 10-12 at the King Abdulaziz International Conference Center, under the patronage of Crown Prince Mohammed bin Salman.
