AI Evolution in 2024: Redefining the Boundaries of Intelligence

Are you still amazed by ChatGPT? The AI field has reached new heights in 2024! – If 2023 was the year of AI technology’s outbreak, then 2024 is the year of technological maturation and practical application. Today, we will review the AI products that truly shocked the market in 2024 and see how they redefine the boundaries of ‘intelligence’.
1. Foundational Large Models: Deliberate Thinking and Performance Competition OpenAI o1 In 2024, OpenAI introduced a new series of large models, the o1 series, which completely subverted the traditional paradigm of large language models. The biggest highlight of o1 is its ‘deliberate thinking’ ability: through internal thought chains trained by reinforcement learning, the model will perform multiple reasonings before answering complex questions, thereby enhancing accuracy.


– Performance: In doctoral-level physics problem tests, o1’s scores jumped from GPT-4o’s 59.5 points to 92.8 points, and even reached the level of gold medalists in the Informatics Olympiad. – Application scenarios: From mathematical derivation to algorithm optimization, o1 approaches or even surpasses human experts in multiple fields. Netizens also commented: ‘This model is truly the ‘meditation master of the AI world’, thinking carefully before answering questions, it’s stable!’



Anthropic Claude 3.5 Anthropic’s Claude 3.5 takes a different path by introducing the ‘Computer Use’ feature, supporting AI to operate computers like humans. It can complete tasks such as screen clicking and form filling with a mouse and keyboard, and even adjust itself when encountering problems. Although the current success rate is only 80%, this technology has shown the huge potential of AI in the field of automated operations. Many netizens joked: ‘In the future, can we ask Claude to help us with ticket grabbing?’


DeepSeek V3 The always low-key Chinese team DeepSeek made a splash in 2024 with DeepSeek V3. This open-source model was trained in just 55 days and cost 5.576 million US dollars, yet it surpassed closed-source models of international giants in multiple fields such as encyclopedia knowledge and long-text understanding. User comments: ‘Looking at DeepSeek’s performance, it really has the flavor of domestic AI ‘overtaking on the inside lane’!’


Google Gemini 2.0 Google’s Gemini 2.0 is another impressive model. It not only understands visual content but also achieves zero-delay voice interaction.


Gemini 2.0 has set a new standard for AI assistants, whether it’s real-time game analysis or recognizing ingredients through a camera to offer cooking suggestions. Netizens jokingly say: “This is the true ‘all-round butler’; in the future, you won’t even need to cook by yourself!”


Video Generation: From Demo to Reality with OpenAI Sora


OpenAI officially launched its video product Sora at the end of the year, allowing users to generate high-quality videos through text and images. Sora also introduced a ‘storyboard’ feature, supporting the design of shots like a film director, making the creative process more flexible. Excited netizens commented: “In the future, shooting short films might only require a computer and a cup of coffee; video production is really taking off.”


ByteDance’s Dream AI


ByteDance’s Dream AI focuses more on the creative needs of ordinary people. With simple text or image input, users can transform their imaginations into high-quality visual works. For example, a creator used Dream AI to produce a short film that recreates the history of film development in just 5 days, garnering 400,000 likes. This easy creation method perfectly embodies the platform’s vision of being a ‘camera for imagination.’


Kuaishou’s Kueliang


Beyond video generation, Kuaishou’s Kueliang AI platform has also introduced a revolutionary AI model workflow. Users can generate model images, automatically change clothes, and produce commercial videos, all in one seamless process. However, due to the high number of users, the queue time for Kueliang is quite long at present. Netizens joked: “Waiting in line for Kueliang is not as good as drinking a cup of milk tea to calm down.”


Image Generation: New Tools for Creative Design


Flux, launched by Black Forest Labs, uses a new architecture and redefines the upper limit of image generation with 120 billion parameters (3.5 times that of Stable Diffusion). It not only matches the quality of MidJourney but also integrates multiple tools, becoming a sharp tool in e-commerce and design fields.


Recraft V3


Recraft V3 quickly dominates the designer circle with its powerful image fusion capabilities. Users can collage various elements on an infinite canvas, easily completing complex creative expressions. However, due to the need for optimization in Chinese text generation, domestic users may feel slightly regretful.


To C Applications: The Rise of AI Companionship with DouBao APP


ByteDance’s DouBao APP, supported by a strong ecosystem, quickly became the largest AI chat application in China.


In addition to the dialogue function, Doubao has expanded into multiple fields such as image generation and music creation, and offers a rich set of AI character settings.


Talkie, developed by the Chinese company MiniMax, has rapidly risen in the global market with its diverse role-playing and interaction methods. Users can even customize celebrity AI characters and interact with them through text or voice.


AI Pet Club of Mengyouhui has created a group of cute AI animal characters, relieving users’ psychological pressure through relaxed interactions. Its unique “Travel Frog”-style interaction method is deeply loved by young people.


In 2024, the innovations in the AI field were dazzling. However, this is just the beginning. With the continuous iteration of technology, we can expect more disruptive products that will further change the way we live and work.


In 2025, who will be the next “AI myth”? Let’s wait and see!



Leave a Comment

Your email address will not be published. Required fields are marked *