Colorado Media Newsroom
March 2nd, 2023, 12:12 PM
From Radio Insight:
https://i0.wp.com/radioinsight.com/wp-content/images/2023/03/auddia.jpg?resize=200%2C200&ssl=1 Auddia Inc. a developer of proprietary AI platforms that reinvent how consumers engage with audio, announced today that it has filed a broad and foundational patent application aimed at improving the shortcomings being exposed in existing GPTs such as Azure OpenAI?s ChatGPT and Google?s Bard. The patent application introduces two significant improvements. First, the patent covers the development of new machine learning algorithms that use the rich data inherent in chat conversations to learn how to optimize GPT prompts to more efficiently extract the best results from large language model GPTs. Second, the patent covers capturing the AI generated prompt improvements and the enriched outputs of the large language model GPTs to create new GPTs with industry specific domain expertise.
?The best way to think about how our new patent application impacts the space is to think about the current large language model GPTs, such as ChatGPT and Bard, as highly knowledgeable people who know quite a bit about everything but are not really experts in anything,? said Jeff Thramann, Executive Chairman of Auddia. ?When these people need expertise in a specific area, they tend to consult experts. What our IP does is create the experts.?
Although Auddia?s patent pending process is applicable to every field, i.e., healthcare, education, law, etc., the Company is focusing its use of the technology in supporting the Company?s mission of reinventing how consumers engage with audio. This will be executed by introducing ChatMusic into the Company?s audio superapp, faidr, as the Company?s differentiated music player. Currently, faidr leverages the Company?s existing AI to offer a highly differentiated premium, commercial-free AM/FM listening experience in addition to offering exclusive content through faidrRadio. Podcasting will be added to faidr this quarter with differentiation and margin in podcasting being introduced through Vodacast innovation in the back half of 2023. The addition of ChatMusic will continue the Company?s strategy of leading the audio superapp space with differentiation and margin.
?Our interest in ChatGPT started when we asked the platform to build a 5-hour playlist of Elton John songs to see what would happen. Incredibly, it gave us a solid playlist. We then asked it to arrange the songs to avoid jarring transitions, ChatGPT responded but perhaps not optimally. We then probed whether it could mimic the famed Pandora algorithm that creates playlists of similar artists. Within seconds we were looking at that playlist, but it still had limitations,? said Theo Romeo, CMO of Auddia. ?Immediately, the entire team realized that a chat interface to music had the potential to be a game changer for how consumers interact with digital devices to generate playlists, adjust them on the fly, mix them together, create new songs in conjunction with other generative AI?s, gather trivia about artists, etc. As we realized the new user experiences for music were essentially limitless and nothing short of amazing, we also realized these game changing capabilities were directly applicable to all the audio content in the faidr superapp. With the addition of ChatMusic to faidr, we will have the only audio superapp we are aware of that has access to consumer data across AM/FM, exclusive content, podcasting, and a music player. Combining this data with the capabilities inherent in our new patent filings uniquely positions faidr to be the ultimate audio content curation platform for consumers. The vision and opportunity here is incredibly exciting.?
Although the chat experience was somewhat mind blowing to the Auddia team initially, the team quickly realized there is an art to optimizing chat conversations to extract the best results from GPTs. The team also realized that simply using a generalized artificial intelligence platform, such as ChatGPT, is not the right solution to optimize a chat interface for a music player or any other specialized use case. Armed with this knowledge, the Auddia team architected a patent pending platform that trains an AI to optimize chat prompts to achieve optimum GPT outputs, while simultaneously using the chat conversations and GPT outputs as training data to create proprietary GPTs in any field, to include a MusicGPT that has domain specific knowledge in music.
?Although ChatGPT is an incredibly interesting technology, it has limitations and is not optimized to music or any other field. A case in point is that ChatGPT has no access to real time information. The current version has nothing beyond 2021,? said Peter Shoebridge, CTO. ?This is where Auddia?s IP comes into play. Our patents are aimed at optimizing chat conversations and utilizing the inputs and outputs of a GPT in real time to train our own proprietary GPTs with domain expertise in music, podcasting, AM/FM, and all other audio content that is constantly being updated through the information embedded in the timely and relevant user conversations that are always happening.?
The power of the Company?s patent applications is in the network effects inherent in the process. Every user interaction generates a conversation between a GPT and the user that can be captured and used to train and update a proprietary GPT with specific domain expertise.
?Similar to how Microsoft is investing $10 billion into OpenAI to use ChatGPT in an attempt to leapfrog Google in search, Auddia plans on leveraging ChatGPT, Google?s Bard, and other GPTs in conjunction with our newly filed IP in an attempt to leapfrog Spotify and other legacy music players,? said Jeff Thramann. ?I see the foundational IP we have filed as an incredibly important catalyst and differentiator for Auddia.?
Michael Lawless, CEO, added, ?It is already well known that ChatGPT, Bard, and Microsoft?s integration of ChatGPT into Bing are resulting in some odd and unproductive conversations. This is to be expected from any interaction between humans and AI. The secret sauce to fix these issues is in training additional machine learning algorithms to optimize chat conversations to extract the best results from the GPT models. This optimization process is a critical part of what our patent applications cover, and we are first applying this process to optimizing conversations related to music, via ChatMusic and MusicGPT, and our plan is to apply our proprietary technology to all audio content across our faidr superapp platform.?
?Data is everything in AI,? added Pablo Calderon, VP of Engineering. ?As users interact with a chat interface, we can capture the incredibly rich data inherent in conversational speech and correlate that to GPT outputs. This loop consisting of conversational inputs, GPT outputs, and conversational feedback is a stream of rich, never-ending data from users that we will harness to train not only a proprietary domain specific MusicGPT, but multiple individual audio content generators specific to each users? preferences across all audio content on the faidr superapp.?
?To get a feel for how interesting the ChatMusic interface can be, go to ChatGPT and ask it some music trivia or to build a music playlist. You can then have a conversation to explore the responses further and take the playlist in any direction you want. The experience is quite intriguing, but it is not yet optimized,? said Theo Romeo.
Auddia will be going live with a website for ChatMusic in March and will announce its launch in a follow up press release.
About Auddia Inc.
Auddia, through its proprietary AI platforms for audio is reinventing how consumers engage with AM/FM radio, podcasts, music, and other audio content. Auddia?s flagship audio superapp, called faidr, brings three industry firsts to the audio-streaming landscape: subscription-based, ad-free listening on any AM/FM radio station; podcasts with interactive digital feeds that support deeper stories and create new revenue streams for podcasters; and a proprietary chat interface for music. faidr also delivers exclusive content and playlists, and showcases exciting new artists, hand-picked by curators and DJs. All differentiated offerings address large and rapidly growing audiences.
more (https://radioinsight.com/headlines/249346/auddia-to-launch-ai-powered-chatmusic-into-radio-app-patent-gpt-enhancements/)
https://i0.wp.com/radioinsight.com/wp-content/images/2023/03/auddia.jpg?resize=200%2C200&ssl=1 Auddia Inc. a developer of proprietary AI platforms that reinvent how consumers engage with audio, announced today that it has filed a broad and foundational patent application aimed at improving the shortcomings being exposed in existing GPTs such as Azure OpenAI?s ChatGPT and Google?s Bard. The patent application introduces two significant improvements. First, the patent covers the development of new machine learning algorithms that use the rich data inherent in chat conversations to learn how to optimize GPT prompts to more efficiently extract the best results from large language model GPTs. Second, the patent covers capturing the AI generated prompt improvements and the enriched outputs of the large language model GPTs to create new GPTs with industry specific domain expertise.
?The best way to think about how our new patent application impacts the space is to think about the current large language model GPTs, such as ChatGPT and Bard, as highly knowledgeable people who know quite a bit about everything but are not really experts in anything,? said Jeff Thramann, Executive Chairman of Auddia. ?When these people need expertise in a specific area, they tend to consult experts. What our IP does is create the experts.?
Although Auddia?s patent pending process is applicable to every field, i.e., healthcare, education, law, etc., the Company is focusing its use of the technology in supporting the Company?s mission of reinventing how consumers engage with audio. This will be executed by introducing ChatMusic into the Company?s audio superapp, faidr, as the Company?s differentiated music player. Currently, faidr leverages the Company?s existing AI to offer a highly differentiated premium, commercial-free AM/FM listening experience in addition to offering exclusive content through faidrRadio. Podcasting will be added to faidr this quarter with differentiation and margin in podcasting being introduced through Vodacast innovation in the back half of 2023. The addition of ChatMusic will continue the Company?s strategy of leading the audio superapp space with differentiation and margin.
?Our interest in ChatGPT started when we asked the platform to build a 5-hour playlist of Elton John songs to see what would happen. Incredibly, it gave us a solid playlist. We then asked it to arrange the songs to avoid jarring transitions, ChatGPT responded but perhaps not optimally. We then probed whether it could mimic the famed Pandora algorithm that creates playlists of similar artists. Within seconds we were looking at that playlist, but it still had limitations,? said Theo Romeo, CMO of Auddia. ?Immediately, the entire team realized that a chat interface to music had the potential to be a game changer for how consumers interact with digital devices to generate playlists, adjust them on the fly, mix them together, create new songs in conjunction with other generative AI?s, gather trivia about artists, etc. As we realized the new user experiences for music were essentially limitless and nothing short of amazing, we also realized these game changing capabilities were directly applicable to all the audio content in the faidr superapp. With the addition of ChatMusic to faidr, we will have the only audio superapp we are aware of that has access to consumer data across AM/FM, exclusive content, podcasting, and a music player. Combining this data with the capabilities inherent in our new patent filings uniquely positions faidr to be the ultimate audio content curation platform for consumers. The vision and opportunity here is incredibly exciting.?
Although the chat experience was somewhat mind blowing to the Auddia team initially, the team quickly realized there is an art to optimizing chat conversations to extract the best results from GPTs. The team also realized that simply using a generalized artificial intelligence platform, such as ChatGPT, is not the right solution to optimize a chat interface for a music player or any other specialized use case. Armed with this knowledge, the Auddia team architected a patent pending platform that trains an AI to optimize chat prompts to achieve optimum GPT outputs, while simultaneously using the chat conversations and GPT outputs as training data to create proprietary GPTs in any field, to include a MusicGPT that has domain specific knowledge in music.
?Although ChatGPT is an incredibly interesting technology, it has limitations and is not optimized to music or any other field. A case in point is that ChatGPT has no access to real time information. The current version has nothing beyond 2021,? said Peter Shoebridge, CTO. ?This is where Auddia?s IP comes into play. Our patents are aimed at optimizing chat conversations and utilizing the inputs and outputs of a GPT in real time to train our own proprietary GPTs with domain expertise in music, podcasting, AM/FM, and all other audio content that is constantly being updated through the information embedded in the timely and relevant user conversations that are always happening.?
The power of the Company?s patent applications is in the network effects inherent in the process. Every user interaction generates a conversation between a GPT and the user that can be captured and used to train and update a proprietary GPT with specific domain expertise.
?Similar to how Microsoft is investing $10 billion into OpenAI to use ChatGPT in an attempt to leapfrog Google in search, Auddia plans on leveraging ChatGPT, Google?s Bard, and other GPTs in conjunction with our newly filed IP in an attempt to leapfrog Spotify and other legacy music players,? said Jeff Thramann. ?I see the foundational IP we have filed as an incredibly important catalyst and differentiator for Auddia.?
Michael Lawless, CEO, added, ?It is already well known that ChatGPT, Bard, and Microsoft?s integration of ChatGPT into Bing are resulting in some odd and unproductive conversations. This is to be expected from any interaction between humans and AI. The secret sauce to fix these issues is in training additional machine learning algorithms to optimize chat conversations to extract the best results from the GPT models. This optimization process is a critical part of what our patent applications cover, and we are first applying this process to optimizing conversations related to music, via ChatMusic and MusicGPT, and our plan is to apply our proprietary technology to all audio content across our faidr superapp platform.?
?Data is everything in AI,? added Pablo Calderon, VP of Engineering. ?As users interact with a chat interface, we can capture the incredibly rich data inherent in conversational speech and correlate that to GPT outputs. This loop consisting of conversational inputs, GPT outputs, and conversational feedback is a stream of rich, never-ending data from users that we will harness to train not only a proprietary domain specific MusicGPT, but multiple individual audio content generators specific to each users? preferences across all audio content on the faidr superapp.?
?To get a feel for how interesting the ChatMusic interface can be, go to ChatGPT and ask it some music trivia or to build a music playlist. You can then have a conversation to explore the responses further and take the playlist in any direction you want. The experience is quite intriguing, but it is not yet optimized,? said Theo Romeo.
Auddia will be going live with a website for ChatMusic in March and will announce its launch in a follow up press release.
About Auddia Inc.
Auddia, through its proprietary AI platforms for audio is reinventing how consumers engage with AM/FM radio, podcasts, music, and other audio content. Auddia?s flagship audio superapp, called faidr, brings three industry firsts to the audio-streaming landscape: subscription-based, ad-free listening on any AM/FM radio station; podcasts with interactive digital feeds that support deeper stories and create new revenue streams for podcasters; and a proprietary chat interface for music. faidr also delivers exclusive content and playlists, and showcases exciting new artists, hand-picked by curators and DJs. All differentiated offerings address large and rapidly growing audiences.
more (https://radioinsight.com/headlines/249346/auddia-to-launch-ai-powered-chatmusic-into-radio-app-patent-gpt-enhancements/)