Gerd Buta on Humanitarian AI Today
Humanitarian AI Today is a podcast series produced by the Humanitarian AI meetup community. The series covers humanitarian applications of artificial intelligence and machine learning from operational, technical and other perspectives.
Humanitarian AI Today guest host Christopher Hoffman from Humanity Link recently spoke with Gerd Buta from the United Nations World Food Program (WFP) about Winter conditions in Ukraine, difficulties involved in delivering aid and about open data sharing and humanitarian applications of technology and artificial intelligence (the episode published on January 20th, 2023).
Gerd heads in-kind food operations in Ukraine and joined the the interview from Kyiv. With conditions across Ukraine so difficult due to the war and logistics challenging, we reached out to Gerd to give our Humanitarian AI meetup group members a view of how bad things are this Winter, how WFP is helping and what it takes to move aid on the ground from a logistical vantage point.
Something that we’re interested in is how challenges faced by humanitarian actors can be expressed digitally for machine applications. Humanitarian organizations use open data sharing frameworks like IATI (an XML standard developed and managed by the International Aid Transparency Initiative) to document aid activities, funding transactions and results. We’re curious how information about supply bottlenecks for example can look reported in XML code and in turn how question answering applications like ChatGPT can then use the information to answer complex queries posed by humanitarian actors.
Although we didn’t get a chance to broach the subject directly, the conversation was informative and we appreciate Gerd taking time to record an interview.
Due to communications connection issues, our audio cut briefly in several locations. As an experiment we cloned Gerd’s voice using Coqui Studio to generate some missing worlds like “open to”, “are” and “fail and scale”. Although the edits could have been improved further through more training and adjustments, for a podcast series on applications of artificial intelligence it was fitting for us to finally experiment ourselves with voice cloning. The edits can be found at: 8:07, 8:37, 10:44, 11:51, 25:12, 27:31, 28:35, 29:29, 30:17 and 30:22.
