• AI KATANA
  • Posts
  • AI app Character.ai is catching up to ChatGPT in the US

AI app Character.ai is catching up to ChatGPT in the US

Also: How Artificial Intelligence Gave a Paralyzed Woman Her Voice Back

Author

AI KATANA
September 12, 2023

Welcome!

In the AI landscape, Character.ai, an app for creating AI personas, is nearing ChatGPT's mobile user base in the US with 4.2 million monthly users against ChatGPT's 6 million. Meanwhile, breakthrough technology from UC San Francisco and UC Berkeley has given a paralyzed woman the ability to communicate through a digital avatar, directly translating brain signals. Additionally, Meta's AI model, Llama 2, is touted as a frontrunner by Benchmark's Bill Gurley, but OpenAI faces copyright infringement lawsuits from writers. On the tools front, NVIDIA launches TensorRT-LLM to enhance Large Language Model performance, and in venture capital news, Pahdo Labs raises $15M for an AI-driven, anime-inspired game platform.

Sliced:

  • 👩🏻 AI app Character.ai is catching up to ChatGPT in the US

  • 💜 How Artificial Intelligence Gave a Paralyzed Woman Her Voice Back

  • 📈 Benchmark’s Bill Gurley: Meta AI model Llama2 has 'most momentum’

  • ✍️ More writers sue OpenAI for copyright infringement over AI training

Character.ai, an app allowing users to craft their own AI personas, is rapidly growing in popularity, nearing ChatGPT's mobile app user base in the U.S. Market data shows Character.ai's mobile apps have reached 4.2 million monthly users in the U.S., in contrast to ChatGPT's 6 million. Since its May 2023 debut, Character.ai has significantly grown from its initial 1.7 million installs in the first week, indicating successful user retention. However, globally, ChatGPT remains dominant, especially on web platforms. Interestingly, Character.ai is more popular among younger demographics, with 60% of its users aged 18-24. Despite a decline in visits in previous months, ChatGPT's web traffic is rebounding. Funded with a massive $150 million in Series A, Character.ai, founded by AI veterans from Google, is poised for continued growth. Whether it's a lasting trend or a fleeting fad remains uncertain.

After suffering a brainstem stroke at 30, Ann was left paralyzed and unable to speak for 18 years. Recently, researchers at UC San Francisco and UC Berkeley developed a cutting-edge brain-computer interface that allows Ann to communicate more fluidly through a digital avatar resembling a human. This technology synthesizes speech and facial expressions from brain signals, and for Ann, it can decode these signals into text at a rate of nearly 80 words per minute. Notably, this is the first instance of synthesizing speech or facial expressions directly from brain signals. The team also used a recording of Ann's voice pre-injury to give the synthesized speech a personalized touch. The ultimate aim of this research is to design a wireless version of this interface, thereby enhancing the independence and social interactions of patients like Ann.

Bill Gurley from Benchmark noted that founders regard Meta's new AI model, Llama 2, as the frontrunner among large AI models. Gurley expressed astonishment at incumbent companies' attempts to deter governments from supporting open-source products like Llama 2. Launched in July with Microsoft, Llama 2 is available for free for commercial use and research. Meta stated that their aim with Llama 2 isn't monetization and it can be accessed on Microsoft’s Azure. While some tests showed Llama 2 surpassing other AI models, it didn't outperform GPT-3 in generating code. Interestingly, Microsoft, despite collaborating on Llama 2, has invested heavily in OpenAI which introduced GPT-4 in March. Google also released its AI model, PaLM 2, in May.

A group of U.S. authors, including Pulitzer Prize-winning Michael Chabon, are suing OpenAI, alleging the company used their writing without permission to train its AI-powered chatbot, ChatGPT. This is the third such lawsuit against OpenAI, and other tech giants like Microsoft and Meta Platforms have faced similar copyright infringement claims. OpenAI, along with other companies, contends that using copyrighted content for AI training falls under "fair use." Despite these controversies, ChatGPT gained significant popularity, registering 100 million active users in January. The authors claim their works, representative of high-quality writing, were essential for training and have sought damages and a halt to OpenAI's alleged unauthorized practices.

🛠️ AI tools updates

NVIDIA has introduced TensorRT-LLM, a tool designed to enhance the performance of Large Language Models (LLMs) inference on NVIDIA GPUs. This innovation, borne out of collaborations with several leading tech firms, is integrated into open-source software optimized for Ampere, Lovelace, and Hopper GPUs. The software includes a deep learning compiler and other key features that enhance the execution of LLMs. With a user-friendly Python API, TensorRT-LLM offers developers flexibility and ease in optimizing and deploying new LLM architectures, without necessitating deep knowledge of C++ or CUDA. MosaicML and Databricks, among others, testify to its efficiency. The software provides significant boosts in performance and energy efficiency, evidenced by benchmarks on NVIDIA Hopper architecture. As the AI ecosystem continues to innovate, TensorRT-LLM facilitates easy adaptation, supporting efficient LLM inference at scale without the need for cumbersome manual adjustments by developers.

💵 Venture Capital updates

Pahdo Labs has secured $15 million in funding, led by Andreessen Horowitz, to develop an anime-inspired role-playing game and user-generated content (UGC) platform. The startup, based in New York with a diverse international team, has codenamed its project "Halcyon Zero". The game will allow players to design their own anime worlds using AI tools and procedural generation. The platform, built on the Godot Engine, will launch on PC with plans for cross-play on other platforms. Pahdo Labs, which began in 2021 and has previously raised $2.5 million, is actively engaging with fans on TikTok, highlighting their development journey and aiming to inspire budding game developers. The founder, Daniel Zou, cites his own experiences with online RPGs and the growth of anime in the West as key inspirations for the project.

🫡 Meme of the day

⭐️ Generative AI image of the day