Practicality Over Technological Flair.

From January 7 to 10, 2025, the Consumer Electronics Show (CES) took place in Las Vegas, USA, under the theme "Dive into the Future." When discussing the future of technology, AI stands as the most transformative force shaping our lives. This year, Leitech dispatched a reporting team to CES 2025, engaging closely with the latest hardware and technologies and observing numerous emerging industry trends.

(Image source: Leitech on-site footage)

Last year's CES buzz was dominated by a startup's rabbit R1, an AI hardware dubbed an "AI new species" that concealed powerful AI functions beneath its cute rabbit exterior. However, post-launch, user reviews paled in comparison to the exhibition hype. The main issue was that this "new species" not only failed to enhance our lives but also worsened the user experience due to its immature design.

The story of "AI new species" continues, with CES 2025 showcasing numerous new AI hardware products. Unlike rabbit R1, these products are not only more intriguing but also more practical. So, can these "AI new species" redefine our future lives?

CES 2025 AI Hardware: Ditching Versatility for Niche Focus

Over the past year, startups have launched various "AI hardware" products, such as rabbit R1, AI Pin, PinNote, etc., characterized by deep integration of large models and a focus on comprehensive functionalities. However, Leitech's hands-on experience revealed that these products fell short of their advertised claims, often inferior to AI phones in terms of user experience.

Conversely, the new AI hardware at CES 2025 seems to have abandoned last year's comprehensive approach, focusing instead on niche markets, particularly areas less prioritized by average consumers, making their products more grounded.

I. AI Basketball Shooting Training Robot: Solo Practice Made Easy

Davy Robot, a specialist in robotic training, unveiled Datic 1, a basketball training robot, at CES 2025. Its primary function is to assist users in shooting practice.

(Image source: X)

Datic 1 comprises a square machine body and an extendable robotic arm. During shooting practice, users can position it under the basketball hoop, where it extends its arm to contact and automatically assemble the hoop. Once assembled, the arm retrieves the ball and returns it to the user via the machine's ejection device, enabling continuous practice sessions.

Notably, Datic 1 leverages advanced neural networks and visual large model data, using body sensors to determine the shooter's position and distance, accurately returning the ball, reducing the need for users to run back and forth, and catering to specialized training needs.

However, Datic 1 still faces practicality challenges. For instance, it must be paired with an existing basketball hoop. Despite its 3-meter arm extension, convenience remains limited, necessitating a suitable basketball court. Additionally, its retrieval mechanism works only for hoop-worthy shots; missed shots still require manual retrieval.

Nevertheless, as a basic training tool, AI technology lowers the usage threshold. Compared to professional equipment, it is more portable and affordable, making it an excellent choice for entry-level enthusiasts.

II. Petal Natural Camera: Capturing Nature's Wonders with AI

Nature photography is a significant category in traditional photography, often requiring photographers to carry various lenses and traverse landscapes to capture stunning plants, insects, and birds, with educational value. At CES 2025, Wonder launched the Petal bionic camera, capable of real-time capture and live streaming of flowers, birds, insects, etc.

The Petal bionic camera is equipped with a novel natural intelligence model that recognizes and analyzes subject information during shooting, providing real-time analysis during live streams. Users can even converse with this "camera" using natural language. For instance, recording butterfly information triggers an automatic report + album generation, or understanding bee pollination processes garners explanations from Petal.

(Image source: Wonder)

From an innovation standpoint, Petal lags behind most bionic cameras. For instance, it supports only 12 megapixels and 4K video recording, and its camouflage resembles ordinary green plant cameras. However, Petal's innovative natural intelligence model bridges vision, natural language communication, and speech large models, making it an intriguing prospect. Frankly, Petal might serve better as a teaching tool than a nature photography camera.

III. WeWalk Smart Cane 2: GPT as a Guide for the Visually Impaired

Recently, OpenAI introduced a new visual model to the ChatGPT mobile app, enabling users to combine real-world objects captured by phone cameras with AI conversations for answers, suggestions, guidance, and other assistance. Some pondered: Wouldn't it be fantastic if GPT could serve as the eyes of the blind?

The startup WeWalk presented its second-generation Smart Cane at CES 2025, incorporating the ChatGPT voice large model for the first time. This cane aids visually impaired users with navigation and life advice. It also features hardware optimizations like tactile buttons, speakers, microphones, and lights, the latter reducing nighttime traffic risks for users.

(Image source: X)

In essence, WeWalk Smart Cane 2 is a smart cane integrated with ChatGPT. Its only drawback is the absence of GPT's latest visual model, suggesting WeWalk's inability to independently develop one. Consequently, WeWalk Smart Cane 2 comes in subscription and lifetime versions, priced at $850 + $4.99/month and $1150, respectively. Both exceed RMB 6000, with the subscription version requiring a monthly OpenAI fee.

Overall, this year's CES showcased a myriad of innovative AI hardware products across sports, nature, and accessibility. Besides these intriguing AI hardware offerings, fields like home appliances, wearable devices, TVs, and PCs have also deeply integrated AI, with a grander presence than last year's CES.

From Technological Flair to Practicality: Has AI Hardware Matured?

2024 is widely acknowledged as the "AI Year" in the industry, and CES, the most influential consumer electronics exhibition, witnessed the debut of AI hardware in a new avatar. Rabbit R1 and AI Pin, representing the so-called "AI new species," showcased groundbreaking smart experiences, with AI technology as the core driver. However, consumers soon realized that reality fell short of the hype.

(Image source: Rabbit R1)

These hardware products not only had limited functionalities but also heavily relied on the internet, sometimes being less practical than ordinary smartphones. The AI hardware market's initial test didn't yield a satisfactory answer.

Market research data indicates that the global AI hardware market size reached 263.1 billion yuan in 2024 but experienced a significant slowdown in year-on-year growth. Consumers' interest in these "new species" waned, with some products encountering rapid declines post-launch due to a lack of sustained appeal. The root cause lies in manufacturers' overemphasis on technological flair, neglecting AI hardware's real-world applicability.

At this year's CES, manufacturers shifted from blindly pursuing "all-round" AI hardware to focusing on niche markets, deeply integrating AI technology into traditional hardware and exploring solutions tailored to user needs. This change not only responds positively to market feedback but also embodies AI technology's pragmatism.

The Datic 1 basketball training robot, Petal bionic camera, and WeWalk's Smart Cane 2 demonstrated that this year's AI hardware has moved beyond showcasing large AI models, embracing a functionality- and practicality-centric design approach. AI's role has evolved from a dominant force to an enabler, enhancing rather than replacing traditional devices, providing users with more efficient and intelligent solutions.

(Image source: X)

For instance, Samsung unveiled its "Home AI" system, integrating various home devices into a smart network to enhance user experience. Additionally, Nvidia released the new GeForce RTX 50 series GPU, emphasizing AI's application in graphics processing to boost product performance.

As AI technology progresses, more traditional hardware fields will benefit from AI empowerment. AI hardware reliant solely on large models will likely fade from the market. We often assert that good technology enhances quality of life without requiring specific user actions. AI empowering traditional hardware seems to be moving in this direction.

Source: Leitech