Wir haben die Konturen der nächsten Generation von KI-Hardware erfasst. Es ist eine Gitarre mit der Fähigkeit großer Modelle.

Text | Liu Shiwu Here is the original English text, keep the original

While most AI products are still limited to office scenarios and efficiency tools, we might be able to glimpse the outlines of the next - generation AI hardware at this year's Yunqi Conference – not a smartphone, not glasses, but a guitar.

Since we've been engaged in human - machine interaction and AI applications for a long time, we had the opportunity to test this generative AI guitar. Without any guitar knowledge, we tried its "Light - chasing Solo - Mode" – key - pressing and following the light trail. The whole process required almost no learning time, and one could start right away.

In contrast to all other intelligent musical instruments on the market, all the notes displayed on the screen during the test run come from the self - developed large model "Tianpu Le" integrated in the guitar – that is, we were not just "device operators", but enjoyed "making music with AI" all the time.

From "playing a song" to "generating a musical segment"

On September 24th, Quwan Technology officially introduced the world's first intelligent guitar, TemPolor Melo - D (hereinafter referred to as TemPolor), equipped with AIGC functions at Alibaba's Yunqi Conference.

It is not only a musical instrument but also an independently functioning "music generation terminal". Its appearance might mark the end of purely software - based AI music and initiate the beginning of a new era of "model - defined hardware".

In contrast to most "intelligent musical instruments" on the market that rely on smartphone apps, TemPolor is an independently functioning guitar. It is equipped with a rotatable touchscreen and colorful strings, which reflects the idea in the hardware design that one can have a complete experience even without a smartphone.

Actually, there are not a few "intelligent musical instruments" on the market. Products like intelligent guitars and intelligent drumsticks have appeared more and more frequently in recent years and formed a relatively stable user group. In comparison, the real breakthrough of TemPolor lies in that generative AI has re - defined the way people play and compose music.

Because TemPolor is the first hardware product that has integrated AIGC into a guitar.

Users no longer have to spend hours learning complicated chords and music theory. They can simply upload their favorite songs or even a melody fragment they sang to the Tianpu Le large model. Within a few minutes, the guitar will then generate an arranged guitar fingerstyle melody and automatically adapt it to a visible note display on the guitar's screen – "every melody in the mind can be turned into a playable reality".

Moreover, we also tested its AI lead - sing function on - site, such as vocal accompaniment, section - by - section practice, and even automatic key adjustment. It breaks down the traditionally boring instrument learning into a progressive playing experience of "first play along, then sing along, and finally play alone", which significantly reduces the frustration of beginners.

For users with existing knowledge, TemPolor also offers a free - play mode. It fully simulates the fretboard logic of a real guitar. The self - developed colorful strings correspond to the six strings of a real guitar and support single - note playing without sacrificing professionalism.

This design of "low entry barrier + high development potential" makes one recognize certain similarities with the early product philosophy of the iPhone – it's not about building hardware with a multitude of functions, but creating a terminal that re - defines the experience.

Naturally, the emergence of TemPolor didn't come out of nowhere but is based on the in - depth insight and long - term practice of the team regarding the trend of music popularization.

The team of Jia Shuo, the person in charge of TemPolor, successfully developed the app "Changya" in the era of the mobile Internet, which is based on playing styles like guitar accompaniment while singing and music social media. It was the first to introduce the interaction pattern of "being able to sing and play by oneself without knowledge of musical instruments".

In February 2019, the Changya app was officially released in various mobile app stores and reached over 2 million daily active users. It also received the annual award for the best apps from leading mobile manufacturers such as Huawei, OPPO, VIVO, and Xiaomi. Users have created a total of 130 million guitar accompaniments while singing.

Interestingly, we found that the user interface of this app developed six years ago is very similar to the latest intelligent guitar apps on the market. In a way, their team was already involved in defining the intelligent guitar six years ago.

The success of Changya also showed that there is a great demand among non - professional users for "playing music". However, the smartphone in the era of the mobile Internet still had many limitations for "playing music". When the wave of AIGC came, it was difficult to fully convey the new experiences offered by software technology through the smartphone.

Thus, TemPolor was born.

Practical exploration of "model - defined hardware"

As an AI - equipped intelligent device, TemPolor doesn't stand alone.

Its core capabilities come from the self - developed multimodal music generation large model "Tianpu Le" of Quwan Technology, a model that supports the generation of music from text, images, audio, and even videos.

Based on the same model, the Tianpu Le team launched a dialogue - based AI music agent named Tunee in the internal testing phase in August this year. On September 26th, it will be publicly tested worldwide. He is like an "indefatigable and communicative professional music partner" who understands user needs through dialogue - based interaction and quickly generates commercially usable music and even track files.

If we look at the history of technology, the truly groundbreaking hardware products are not simply machines with better parameters, but they re - define the experience.

TemPolor might also re - define the "musical instrument" – it is no longer just a toy but an expression medium. In an era when AI is conquering everything, most companies still focus on improving efficiency with AI (writing code, creating designs, optimizing processes), but only a few can enhance the expression and emotional capabilities of people with AI.

Jia Shuo shared his thoughts at the Yunqi Conference. Tunee and the intelligent guitar TemPolor are two application directions of AI. One pursues maximum efficiency, the other pursues an optimal experience.

Our investigation has shown that there are very few technology companies in the world that simultaneously own a self - developed large model, AI applications or agents, and intelligent hardware. Whether it's the latest rumors about OpenAI's plan to launch its first AI hardware product next year or the continuous attempts of Google and Apple to combine generative AI with end - devices, it shows that leading technology companies are actively crossing the boundaries of pure software business areas and approaching an overall strategy of "model + hardware + ecosystem".

The self - developed model of the Tianpu Le team and its two application directions exactly match these approaches. The only difference is that Tianpu Le has always focused on the music field.

The new ecosystem behind a guitar

The appearance of the intelligent guitar TemPolor is a representative example of the further development of AI application forms: The technology of large models is gradually extending from the cloud to end - devices, from virtual interaction to physical interaction, and from a tool to a creative partner.

The ultimate goal of technology has always been to serve people better. AI can not only write code and optimize processes but also lower the threshold for creativity and expand the possibilities for self - expression.

That's exactly what Tianpu Le tries to achieve – through the close integration of large models, intelligent agents, and hardware, all music lovers, even without music theory knowledge and poor playing skills, can freely immerse themselves in the world of music.