In a separate post I have explained our experiment with consolidating the news links that we present daily and weekly. Many of you may have seen news of Google Labs AI app called NotebookLM.
Among other facilities, NotebookLM offers a way to dissect long documents. The ability to produce short summaries is nothing new. Where it differs is its ability to produce things like a study guide, a set of key questions, and what it calls 'A briefing document'. You can load multiple documents and ask it to produce a cross source summary including an outline of points of agreement and disagreement.
What is new and fascinating (at least to me) is its ability to produce an audio podcast from any document with two 'real-sounding' podcast hosts (one male, one female). This may be a foretaste of the scary future when human voices become redundant. The quality of the voices is remarkable (and somewhat eerie). Of course, listening to such a podcast should also warn of us of the difficulty in knowing whether the voices and opinions you hear are 'real' or 'not real'.
So here is another approach to presenting the news of the week - our first AI podcast. This podcast covers this week's links relevant to UNICEF, the UN, and children.
Click here for our AI Podcast of News You May Have Missed This Week The dialogue takes 20 minutes, so you may want to speed it up to 1.5 speed (just click the 3 dots in the upper right corner and choose "Change Playback Speed".
Keep in mind that these voices are not real people, just the AI talking to itself about the news articles listed of our compilation of news links.
Thanks for testing this new technology, Tom; god knows we need new ways to process and present the avalanche of shattering and numbing news coming out of Gaza and other killing fields.
ReplyDeleteThis podcast does a good job of pulling together the news and commenting on its devastating impact on children. I wonder, however, when the faux earnest Ai voices, with their quirky "ums" and "wells" punctuating the narrative , will begin to turn off listeners. The tone is kind of cloying after a while. Not unlike certain BBC news readers.
I'm a fan, for now, and really appreciate your search for ways to engage our colleagues and humanitarian actors.
Thanks Robert. I think your use of the word 'cloying' hits the nail on the head. The fact that the hosts have such a breezy vocabulary is one problem. Another is that the two hosts never change. No one would want to listen to the same hosts week after week. Also, of course is a question of duration. You can listen to short pieces on a simple subject, but it is asking too much for listeners to follow a long 20 minutes podcast covering so many issues. Finally, there is the issue of tailoring to a particular audience. This is the next next function which Google has apparently built in. The current level is clearly intended for a very general audience. It would be interesting to see whether we can tailor it to a specific group, such as retired UN staff who are very familiar with the issues. Stay tuned, interesting times ahead as technology continues to evolve.
DeleteDear Tom,
ReplyDeleteFor me your work was very helpful, forcing us to think more both about AI's potential benefits and its terrifying potential for destructive purposes. With regard to your experiment -
Your expanded listing of the news about children's issues is definitely in line with my thinking about the need to increase humanitarian issues in the public agenda through a compilation of news with a humanitarian focus - the realities/facts, "lived experience" of victims, and the efforts to aid them. [If no objection, I will use even this test listing in my class - I know of nothing that equals it.).
It also raises questions. What would be the challenges of adding a link? What is the optimal level of detail? Are statements by notable people or simply the facts most helpful? Can "lived experience" be more effectively communicated (not just that "50 trucks.." but what the families are eating, who is providing food and real time reports of the challenges)? Clearly if listings such as this were expanded further to include the realities of many emergencies, the list would be massive, but then, while it may be beyond what might be put in xUNICEF, isn't this massive overview of humanitarian issues just what is needed?
AI - perhaps AI is a tool that might make this doable (?).
The test of the blog on Pact for Future was interesting, impressive in its summary and presentation, but like others have said, difficult to listen to or consider credible. The introduction of judgements throughout the dialogue, knowing it was machine reporting, was chilling for me (if the machine can declare it "good", why can's the machine declare it "bad? It would be interesting to ask AI to summarize and critique the Pact for the Future or the list of news items you compiled, from a radical right perspective - what would the voices say then?
Everett
Thankyou so much, Tom. I found it a bit frothy for such a ‘heavy but essential conversation’. But i guess that’s caused at least in part by the tone of the inputs, which could be problematic for UNICEF given its restraint in public statements and documents. But, well worth trying out and I will certainly give it a go. Rozanne
ReplyDeleteAn interesting experiment indeed Tom, thank you, including comments above and your response, especially (for me) on the tone of faux concern that begins to grate. Bottom line in any use of AI for reporting is that it be identified as such (as you have done). Other issues are the generic IDs (white Americans judging from accents + pic?) of the AI hosts, the still prevalent tho improving tendency of generative AI to "hallucinate" info, the biases inbred in its algorithms, and the unknown source material it draws from, leaving one unable to assess its reliability (this is a major critique of all AI-generated information, contrasted, for example, with Wikipedia that annotates its info). A principal argument for AI is its capacity to ingest ever greater loads of info, the bane of the internet and info economy. But, given which entities control it, that is yet another quantity over quality pitch for more disposable consumerism. AI undoubtedly has benefits - sorting through reams of data is certainly one of them - but its ancillary outputs desperately need regulatory guardrails that, like the proliferating lies, deepfakes, etc. on social media, are still conspicuously absent.
ReplyDeleteUgh. I agree with Robert that their voices and the presentation style were cloying and with Ellen that it began to grate on me in the 2nd or 3rd minute. I suppose we are stuck with AI, but its consumption of fossil-fuel generated energy -- along with the horrific wars that are being waged across the globe -- spells doom for grandchildren and their children.
ReplyDelete