**SEO Title:** Pocket FM to Launch Advanced Language Model by Year-End
**Meta Description:** Pocket FM aims to develop a large language model by year-end, enhancing storytelling through AI technology and strategic investments.
**URL Slug:** pocket-fm-language-model-launch
**Headline:** Pocket FM Set to Unveil Advanced Language Model by Year-End
Pocket FM, an audio series platform founded in 2018 by Rohan Nayak, Prateek Dixit, and Nishanth KS, is gearing up to launch a large language model (LLM) by the end of this year. The startup, which has already labeled and categorized its proprietary datasets, is currently testing an early version of its model.
Prateek Dixit, co-founder and chief technology officer of Pocket FM, shared insights on the ongoing development, stating, “We’re currently testing a very raw model, which will take some time. We’re also working on acquiring graphic processing units.” The company is in the process of implementing reinforcement learning techniques for the LLM, with expectations for the model to be ready within five to six months.
To support this initiative, Pocket FM plans to purchase between 30 and 50 Nvidia A100 or H100 GPUs, which range in price from $8,000 to $25,000 each. Despite the significant investment, Dixit views this as a strategic move rather than merely a cost decision, emphasizing its importance for scaling storytelling capabilities with AI.
In addition to hardware upgrades, Pocket FM is enhancing its infrastructure, hiring skilled AI engineers, and increasing its research and development (R&D) budget. Currently, the company allocates 8-13% of its revenue, approximately $26 million, to R&D, with 40% of that dedicated to AI projects. This investment is expected to rise by 1% to 2%.
The startup intends to build its LLM on an open-source foundation model, such as Meta’s Llama 3, specifically tailored for storytelling. While Pocket FM has been utilizing open-source models fine-tuned for its genre-specific needs, the company has reached a point where content quality has plateaued. “With our own model training, we can achieve a significant leap in quality,” Dixit noted.
Popular genres on the platform include drama, fantasy, and thrillers, with writers producing thousands of stories in these categories. The LLM will be utilized for various applications, including story creation, comic development, character arcs, and even AI-generated videos. “The idea is to take these foundation models and fine-tune them for different writing styles, which also applies to comics,” Dixit explained.
Earlier this year, Pocket FM launched Pocket Toons, a webcomic platform powered by an AI studio named Blaze, which produces comics “20x faster at one-third the cost” by automating processes like background rendering, scene composition, and coloring while maintaining artistic creativity. The company has also been leveraging natural language processing (NLP) to enhance its content’s SEO performance.
In conclusion, Pocket FM’s ambitious plans to develop a large language model signify a strategic investment in AI technology, aiming to elevate storytelling quality and expand its content offerings in the competitive audio series market.
**FAQ Section:**
**Q: What is Pocket FM’s plan for its large language model?**
A: Pocket FM plans to launch its large language model by the end of the year, focusing on enhancing storytelling capabilities and content quality through AI technology.