Using A.I. to generate a podcast discussing research papers and iNaturalist projects

Across the board, Scientists can be some of the worst communicators. It usually starts off great, but falls apart when we try to distill the whole idea down to a few take-home points that will stick with the audience while keeping it concise and brief. I'm nearing 20 years of observing how most Scientists try this and get hung up on details. There's only one seminar from 2008 that's ever really struck that "WOW" chord with me where I can honestly say I still remember the Professor's name, the topic, and the metaphors he used to describe his results wrapped around a 'story-telling' framework.

Which brings me to A.I. Specifically Google's Gemini A.I.. A colleague played a prank on a friend and then shared with me how he did it. A light-bulb went off in my head - a practical science communication tool. One function of the tool is that it can produce a completely A.I. generated podcast in the style of two people discussing the take-home points of the science publications/references that is 'fed' into it.

Have a listen at these completely A.I. generated podcasts based on some of my research publications:

Invasive zebra & quagga mussel risk in Canada

• The podcast - https://shorturl.at/AyoZE
• The original papers: [1],[2]

Deep-sea coral and sponge conservation & seamounts in Pacific Canada

• The podcast - https://shorturl.at/V0oz3
• The original paper: [3]

I also tried providing the A.I. with URL links to my Marine Biodiversity of Canada projects [4],[5],[6],[7]

• The resulting podcast - https://shorturl.at/N0O4j

Google's A.I. is doing a better job as a science hype man than any scientist I've ever met. I overlap with enough non-science circles that I will make sure to re-state the obvious; the podcast was not real, the podcast hosts were not real people, A.I. is a computer algorithm that 'read' the documents and then created the podcasts with simulated people talking in laymen terms. The only real aspects were my published papers I used to 'train' the A.I.

There are some quirks and bugs with the A.I. speech algorithm but the results are scary realistic. The bigger picture issue is the net-negative narrative associated with A.I. in the public consciousness. Perhaps the the ticket to combating the negative narrative associated with A.I. is to only feed peer-reviewed evidence into the machine and given, I was the original source, have the original authors confirm if the podcast is accurate. I can confirm the science in my A.I. podcasts was 95% accurate with only some things about sponges/corals being slightly off.

References and Sources:

[1] National Aquatic Invasive Species (AIS) Risk Assessment for Zebra Mussel (Dreissena polymorpha) and Quagga Mussel (Dreissena rostriformis bugensis), April 2022
[2] High-resolution freshwater dissolved calcium and pH data layers for Canada and the United States
[3] Modelling the environmental niche space and distributions of cold-water corals and sponges in the Canadian northeast Pacific Ocean
[4] Marine Biodiversity of Canada project
[5] Marine Biodiversity of Pacific Canada project
[6] Marine Biodiversity of Atantic Canada project
[7] Marine Biodiversity of Arctic Canada project

由使用者 jackson_chu 在 2024年09月26日 13:42 所貼文