Here's the English translation:
In simple terms, the media's advice to everyone is "beautiful but useless, don't buy it at all."
The latest news is that the first batch of AI Pin customers are already returning the product on a large scale. ### The well-known tech media The Verge reported that from May to August this year, the number of AI Pin returns exceeded new sales. If the Vision Pro facing a wave of returns is experiencing a steep decline, then AI Pin is facing a "free fall."
Once aimed to kill smartphones, now facing mass returns
Data shows that the total sales of Humane's AI Pin and accessories have just exceeded $9 million (about 64.388 million RMB). The shipment of AI Pin is about 10,000 units, however, this is only 10% of Humane's planned target of 100,000 units by April 2025.
As of June this year, about 8,000 AI Pins had not been returned; as of today, the number of devices still in customers' hands is about 7,000. ### This means that about 1,000 AI Pins were returned in 2 months - a number higher than the new shipments of AI Pin in these two months. At this rate of returns, it might not take a year before AI Pins become hard to find in the market.
It's worth noting that internal sources revealed that once an AI Pin is returned, the company cannot refurbish it, cutting off the path for second-hand sales. A large number of AI Pins will completely become electronic waste.
Fortunately, Humane still has time to make mistakes, as it has raised over $200 million from Silicon Valley executives. Even if not a single AI Pin is sold, it won't affect the company's operations in the short term, but in the long run, a completely failed AI device will definitely affect investors' confidence.
In fact, a simple review reveals that the failure of AI Pin was somewhat inevitable. As a standalone wearable device without a screen, AI Pin supports network connection through eSIM, does not rely on phones or other personal terminals, uses voice as the core interaction method, and also supports laser projection, gestures, and touch (on the body surface) as three interaction methods.
From the "screenless" product design, it's clear that the AI Pin team completely lacks awareness of user needs and human nature: ### Vision is the main way humans obtain information, and electronic products without screens can only be accessories, like AirPods. When smartwatches, smart home controls, smart speakers, and even smart cars are emphasizing the value of "screens," AI Pin goes against the trend, and it's no wonder it fails.
Limited by the immaturity of machine vision technology, the information AI Pin can understand about the real world through its camera is still very limited, and visual interaction is difficult to be fully usable in the short term. The system design that heavily relies on cloud-based large models makes AI Pin face connection and latency issues. Adding the time for model generation, in the demonstration at MWC, AI Pin often needs to wait for several seconds to give a response.
We won't elaborate on the unbearable battery life, cumbersome and inefficient interaction methods, and lack of application ecosystem. In the face of a series of shortcomings, AI Pin's high price of $699 became a magnifying glass for problems, not to mention the additional $24 monthly subscription fee.
Theoretically, everything AI Pin can do can be done through a smartphone, and considering the more powerful computing power, screen, and software ecosystem, smartphones will probably do it better. This is also the challenge faced by all standalone AI hardware. Although AI Pin's design concept of observing the world while hanging on the chest is highly creative, it is still far from replacing smartphones at this stage. This looks more like an engineer's "fantasy" product, with enough avant-garde ideas, but really impractical.
As AI large models accelerate and combine with hardware, AI phones, AI PCs, AI TVs, AI educational hardware, and other AI hardware emerge and gain market recognition. However, many oddly shaped AI hardware also appear as "new species," walking on the same path of pitfalls as AI Pin.
Countless "AI Pins" are on the road to failure
A chief analyst at the market consulting firm Forrest once shared an interesting view: No one wants to show up at a costume party in casual clothes. This is most appropriate to describe the current AI hardware market. Every company is catering to industry trends, showcasing their latest achievements in AI, and many are passively following the trend just to "not seem out of place."
Some traditional products have indeed brought better user experiences after adding AI functions, such as AI phones that have become smarter, AI TVs with improved picture and sound quality, gas stoves that automatically control temperature, and so on. ### But some products are purely forced innovations, giving a feeling that the hype is greater than practicality.
The Canadian startup GluxKind Technologies has showcased an AI stroller called "Ella" that doesn't need to be pushed manually. This stroller can automatically brake, issue alerts, and self-drive when empty. When out and about, Ella also has an automatic rocking function, combined with customizable soothing sounds inside the body, claiming to achieve the effect of lulling babies to sleep. With the addition of new concepts, the price of this AI stroller has directly soared to $3,300.
There are many AI strollers that focus on self-driving, but from the current perspective, expensive AI strollers are still difficult to popularize and cannot break into the mainstream market. Moreover, these companies have overlooked the degree of importance parents place on their babies, or rather "AI". Trusting to leave a baby alone in a stroller that might malfunction and cause accidents at any time, it's likely these parents are not biological.
AI strollers are a typical example of "product development detached from reality." There are many similar cases, and it's not that AI can't be applied to these products, but it needs to find a more suitable entry point, which is what we often call the difference between "having AI" and "good AI." If this problem cannot be solved, it can be foreseen that the AI Pin's failure will occur on more AI hardware.
What does ideal AI hardware look like?
Currently, AI hardware can be roughly divided into two routes: one is "AI + hardware," which is the traditional products with added AI functions mentioned above; the other is "AI native hardware," represented by independent AI terminals like AI Pin and Rabbit R1.
The combination of AI and hardware is a strategy adopted by enterprises to create uniqueness in the competitive and homogeneous consumer electronics market, which is commendable.
The first category, "AI + hardware," currently has better market feedback because the corresponding products are already mature. Using AI to transform products is "icing on the cake," such as AI phones using AI to enhance imaging experience, or AI PCs using AI to improve work efficiency. They don't need to redefine demand scenarios, and the difficulty and risk are much smaller.
The second category, "AI native hardware," adopts a completely new product form, although it has enough freshness and even "gimmicks," it also faces greater challenges. In addition to the old topics such as the education market and defining categories, ### the primary challenge is how to handle the relationship with smartphones, whether to join the smartphone ecosystem or try to replace smartphones.
AI Pin and Rabbit R1 are representatives trying to replace smartphones, but they only extract the AI from smartphones and make it into a device for specific scenarios, highly overlapping with smartphone AI functions. However, in terms of other functions, their performance is usually not as good as smartphones.
Smartphones are still the core of personal computing devices. What AI native hardware may need to do is not to replace smartphones, but to fill the gaps in the ecosystem built around smartphones.
For example, AI friend, which positions itself as an "electronic friend"