At the end of 2022, ChatGPT emerged and sparked a new wave of AI.
In 2023, the AI landscape witnessed a surge in large-scale models, with major players worldwide making frequent moves in the AI domain, leading to the emergence of various AI mega-models.
As spring arrived in 2024, the heat in the AI sphere surged once again, heralding disruptive advancements.
On February 16th, Google announced the launch of the all-new Gemini 1.5 AI model, marking Google’s largest model to date. Gemini 1.5 is built on Google’s foundational model development and infrastructure research and engineering innovations, including making the training and serving of Gemini 1.5 more efficient through a new Expert Mix (MoE) architecture, capable of handling a standard context window of 128,000 tokens.
Coincidentally, on the same day, OpenAI dropped a “bombshell” by unexpectedly releasing the large-scale text-to-video model, Sora, quickly becoming the talk of the town and garnering widespread attention within the industry.
According to the OpenAI website, Sora inherits the visual quality and instruction-following capabilities of DALL·E3, able to generate high-quality videos up to 1 minute long directly from a user’s single sentence input. These videos include highly detailed backgrounds, complex multi-angle shots, and emotionally rich multiple characters. OpenAI stated that they are teaching AI to understand and simulate the physical world in motion, aiming to address real-world interaction problems.
A stylish lady strolling down a neon-lit street in Tokyo; Image Source: OpenAI official website
Currently, the OpenAI website has been updated with 48 video demos. In these video demos, Sora not only accurately presents details but also understands the presence of objects in the physical world, generating characters with rich emotions. Moreover, the model can generate videos based on prompts, still images, or even fill missing frames in existing videos.
However, OpenAI also admits that Sora is not perfect, as it still has some limitations and issues in certain detail generations, such as incorrect objects in complex physical scenes. Nevertheless, the powerful text-to-video capability demonstrated by Sora has been enough to make a significant impact on the film industry.
Meanwhile, since the global wave initiated by ChatGPT, domestic AI mega-model companies have also seen the largest single-round financing to date.
It is reported that the AI startup “Dark Side of the Moon” recently completed a new round of financing exceeding $1 billion, with investors including Alibaba, Meituan, Xiaohongshu, and Sequoia China, among others. Following this round of financing, the valuation of Dark Side of the Moon has reached approximately $2.5 billion, making it one of the leading companies in the domestic mega-model field.
Public records show that Dark Side of the Moon was founded in March 2023, dedicated to seeking the optimal solution for converting energy into intelligence through co-creation with users. Core members of the company’s founding team have participated in the development of multiple large models such as Google Gemini, Google Bard, Pangu NLP, and Wudao, with several core technologies adopted by mainstream products such as Google PaLM, Meta LLaMa, and Stable Diffusion.
In October 2023, Dark Side of the Moon officially launched its first intelligent assistant product, “Kimi Chat,” supporting input of 200,000 Chinese characters, with a focus on lossless memory, achieving a breakthrough in the field of “long text,” marking the company’s first attempt at To C super applications.
As the Year of the Dragon begins, the AI field continues to see favorable developments, with AI mega-models experiencing new advancements and accelerating implementation. In 2024, with ongoing technological innovation, we anticipate the AI race to bring more surprises and witness endless possibilities.
This paper is from Ulink Media, Shenzhen, China, the organizer of IOTE EXPO (IoT Expo in China)
Join us next year in Shanghai, and let’s shape the future of technology together!
To register IOTE 2024 Shanghai station:
https://eng.iotexpo.com.cn/sh/Visitors.html?lang=EN&source=YJ1
