Post AVunMTcS8a2aVcqVuK by csb@podcastindex.social
(DIR) More posts by csb@podcastindex.social
(DIR) Post #AVueaBydd0VKG9gjMu by dave@podcastindex.social
2023-05-22T12:16:27Z
0 likes, 0 repeats
@csb Ok, I was just asked by a journalist if it’s possible to know how many podcasts there are about AI. Is this something that your new plug-in could answer?
(DIR) Post #AVufRDEvfDc5WXaebw by csb@podcastindex.social
2023-05-22T12:26:02Z
0 likes, 0 repeats
@dave nope, because several podcasts about AI have inventive titles without "artificial intelligence" in title, see https://fnd.io/#/us/search?mediaType=podcasts&term=artificial%20intelligence as example.However to find estimation do SQL query on your DB akin to:SELECT COUNT(*) FROM podcasts WHERE podcast_description LIKE '%artificial intelligence%'
(DIR) Post #AVum2wUB0EpOI761hI by dave@podcastindex.social
2023-05-22T13:40:04Z
0 likes, 0 repeats
@csb But, isn't the LLM supposed to be able to find those inventive titles?
(DIR) Post #AVumICOY8QQJybjtGS by clq@clq.im
2023-05-22T13:42:48Z
0 likes, 0 repeats
@dave @csb Given that LLMs are supposed to come up with stuff independently rather than reverse-engineering things I suppose the way to do this would be to ask AI to "come up with 500K inventive titles for AI podcasts", then do a fuzzy'ish match with actual titles. A lot of potential for false positives though.
(DIR) Post #AVunHSAM73m6YQJBR2 by csb@podcastindex.social
2023-05-22T13:53:54Z
0 likes, 0 repeats
@dave my pluggin delivers only top 10ChatGPT has old data: “ As an AI with a knowledge cutoff in September 2021, I don't have the most up-to-date statistics on the number of podcasts specifically dedicated to AI. As of my last training data, there were dozens of popular podcasts focusing on AI, including "Artificial Intelligence with Lex Fridman," "AI Today," "The AI Alignment Podcast," "The AI Podcast" by NVIDIA, "Eye on AI," "Talking Machines," and "AI in Business," among others.”
(DIR) Post #AVunMTcS8a2aVcqVuK by csb@podcastindex.social
2023-05-22T13:54:49Z
0 likes, 0 repeats
@dave maybe ask Google Bard that fresher LLM with Palm2asking Bing Chat rather useless as it is basically searching web for old top 100
(DIR) Post #AVuvKbpVRHrD1mZWb2 by steven@podcastindex.social
2023-05-22T15:07:08Z
0 likes, 0 repeats
@dave @csb his plug in just queries the search api when a user asks it to to. It isn't being trained on the index data.
(DIR) Post #AVuvKcYAlDaPGJTCDY by csb@podcastindex.social
2023-05-22T15:24:06Z
0 likes, 0 repeats
@stevenyez, Steven, otherwise podcast index website would go down to China town i strive to extend it so that transcript(s) of given episode(s) is/are fetched and then full power of LLM is unleashed: summarizing, splitting into chapters, analyzing ideally podcast index dot org could have it’s own LLM trained internally @dave
(DIR) Post #AVuwZupDCIXSEqIT4a by steven@podcastindex.social
2023-05-22T15:26:16Z
0 likes, 0 repeats
@csb @dave thanks for the clarification. Although I don't use any of these systems, I'm interested to see what else you add to the plug in.
(DIR) Post #AVuwZvbQJ36SeMqyDg by csb@podcastindex.social
2023-05-22T15:38:06Z
0 likes, 0 repeats
@steven we should follow each other
(DIR) Post #AVuwmSJCX1CtYrItBw by csb@podcastindex.social
2023-05-22T15:40:22Z
0 likes, 0 repeats
@steven shureand i have zero technical problems - i am master both in Python and OpenAPIv3’s YAML: i just am not shure what to put in endpoint descriptions so that GPT - the LLM of ChatGPT - would recognize it and associate by calling that new endpoint for transcripts, it’s like poetry writing @dave