U.S. Patent 11,935,520 titled 'Identifying Shifts in Audio Content Via Machine Learning' supports the core AM/FM AI model used within faidr
Company advances its patent application for using AI prompts to interface with large language models (LLMs) to a non-provisional application
BOULDER, CO, April 08, 2024 (GLOBE NEWSWIRE) -- via NewMediaWire -- Auddia Inc. (NASDAQ:AUUD) (NASDAQ:AUUDW) ("Auddia" or the "Company"), developer of a proprietary AI platform for audio and innovative technologies for podcasts that is reinventing how consumers engage with audio, today announced the U.S. Patent and Trademark Office (USPTO) awarded Auddia with U.S. Patent 11,935,520 for the core AI technology Auddia uses in its flagship faidr app to deliver ad-free AM/FM radio stations to paid subscribers.
"Although we have believed we were leading the AI space with important and patentable technology when we began the process of training our algorithm to deliver a premium ad-free AM/FM listening experience, there is a big difference between filing a patent application and having the application approved," said Jeff Thramann, Executive Chairman of Auddia. "Issuance of this patent validates our innovation and secures our AI for Audio technology as a proprietary platform we look forward to building upon to drive new innovations in support of additional industry first premium audio listening experiences."
In addition to issuance of Patent 11,935,520, the Company has taken the next step in advancing the provisional patent application it filed last year related to leveraging AI to improve large language model (LLM) prompts and the domain specific knowledge of proprietary GPTs. In February of 2024, the Company converted the previously filed provisional application to a non-provisional application. This conversion initiates the process for the patent office to review the application as the next step in pursuing patent protection.
As the Company highlighted last year when filing the provisional application, the non-provisional patent application introduces two significant improvements to the LLM space. First, the proposed patent claims cover the development of new machine learning algorithms that use the rich data inherent in chat conversations to learn how to optimize prompts to more efficiently extract the best results from LLMs. Second, the patent covers capturing the AI generated prompt improvements and the enriched outputs of the LLMs to create new GPTs with industry specific domain expertise. Auddia intends to use this technology to create focused domain specific language models to introduce new B2B and B2C audio experiences.
"As the technology landscape continues to accelerate, we continue to evaluate new technologies to determine how we can use them to introduce unique capabilities and experiences for creators and consumers within the audio space," said Michael Lawless, Auddia CEO. "We look forward to continued validation from the USPTO as we continue to innovate and invent."
名爲 “通過機器學習識別音頻內容的變化” 的美國專利11,935,520支持faidr中使用的核心AM/FM AI模型
公司將其使用人工智能提示與大型語言模型 (LLM) 交互的專利申請推進爲非臨時申請
科羅拉多州博爾德，2024年4月8日（GLOBE NEWSWIRE）——通過NewMediaWire——Auddia公司（納斯達克股票代碼：AUUD）（“Auddia” 或 “公司”）（“Auddia” 或 “公司”）今天宣佈獲得美國專利商標局（USPTO）授予專有音頻和播客創新技術人工智能平台奧迪亞在其旗艦fair應用程序中使用的核心人工智能技術獲得了11,935,520項美國專利，該技術用於向付費用戶提供無廣告的AM/FM廣播電臺。
Auddia執行董事長傑夫·瑟拉曼說：“儘管當我們開始訓練算法以提供優質的無廣告AM/FM聆聽體驗時，我們相信我們憑藉重要且可獲得專利的技術在人工智能領域處於領先地位，但提交專利申請和批准申請之間有很大的區別。”“這項專利的頒發證實了我們的創新，並確保了我們的AI for Audio技術作爲專有平台的安全，我們期待在此基礎上再接再厲，推動新的創新，爲更多行業首創的優質音頻聆聽體驗提供支持。”
除了頒發專利11,935,520外，該公司還採取了下一步措施，推進了去年提交的臨時專利申請，該申請涉及利用人工智能改善大型語言模型（LLM）提示和專有GPT的領域特定知識。2024年2月，公司將先前提交的臨時申請轉換爲非臨時申請。這種轉換啓動了專利局審查申請的程序，這是尋求專利保護的下一步。
正如該公司去年在提交臨時申請時強調的那樣，非臨時專利申請爲LLM領域帶來了兩項重大改進。首先，擬議的專利要求涵蓋了新的機器學習算法的開發，這些算法使用聊天對話中固有的豐富數據來學習如何優化提示，從而更有效地從LLM中提取最佳結果。其次，該專利涵蓋捕獲人工智能產生的即時改進和LLM的豐富產出，以創建具有行業特定領域專業知識的新GPT。Auddia打算使用這項技術來創建特定領域的特定語言模型，以引入新的B2B和B2C音頻體驗。
Auddia首席執行官邁克爾·勞勒斯表示：“隨着技術格局的持續加速，我們將繼續評估新技術，以確定如何利用它們爲音頻領域的創作者和消費者引入獨特的功能和體驗。”“在我們繼續創新和發明的過程中，我們期待美國專利商標局的持續驗證。”
