In the community’s “Your Most Anticipated OTA New Features” poll, 3212 car owners voted for “New Ideal Sameer,” accounting for 73% of the total votes, indicating everyone’s expectations and love for the New Ideal Sameer.
As an internal test user, I have already experienced the all-new New Ideal Sameer. If you ask me how it performs, my answer is that the New Ideal Sameer is like a new person.
This change mainly comes from four aspects:
First, the brain of the New Ideal Sameer has been replaced,
Second, the hearing ability of the New Ideal Sameer has been adjusted,
Third, the speaking style of the New Ideal Sameer has changed,
Fourth, the skill level of the New Ideal Sameer has improved.
Next, we will discuss these changes one by one.
The New Brain of the New Ideal Sameer
In the past year’s OTA process, the New Ideal Sameer’s changes were not significant. The core reason is that the original supplier provided a set of voice solutions, which included black-box logic, so the New Ideal Sameer lacked autonomy in the changes.
In the face of increasingly severe competition in the smart cockpit market, increasing autonomy can improve core competitiveness, so self-developing is the only way out. In the 2.1 version’s voice engine switch, the New Ideal’s voice team used the self-developed method for the voice engine and cooperated with companies like Horizon, Sibichi, and Microsoft to provide basic capabilities. For the New Ideal Sameer, this change is a reconstruction that stands at a new starting point.
The all-new self-developed voice engine is the new brain of the New Ideal Sameer. The new brain is developed by the New Ideal Sameer team with partner companies, including acoustic front-end, voice wake-up, speech recognition, natural language understanding, dialogue management, natural language generation, and speech synthesis technology. Through these capabilities, the new engine will improve wake-up ability, recognition, semantic understanding ability, and support rapid iteration, providing a guarantee for more diversified ideal voice functions in the future.
Four Ears Listen Intently
The New Ideal Sameer has a natural advantage over other voice assistants, namely four ears (high-sensitivity microphone), corresponding to the four seats in the car, realizing four audio zone positioning, which is very friendly to all passengers in the car and caters to the different needs of each family member.
However, some car owners gave feedback that when they awaken the New Ideal Sameer to issue commands, the children in the back seat always make trouble. This problem has been basically eliminated in the new version of the New Ideal Sameer. In the vehicle settings – voice, a new function has been added to select the position to awaken the New Ideal Sameer, that is, you can independently turn off one microphone and pretend that the New Ideal Sameer cannot hear anything.
During this upgrade, the Ideal-X student can now be interrupted. For example, during weather or stock broadcasts, Ideal-X would not respond to you before, but now you can reawaken and activate voice commands to improve communication efficiency.
Due to the change of the voice engine and the impact of mistaken awakening in the previous version, some of the quick command words supported by Ideal-X have been adjusted and temporarily discontinued, such as “navigate home”, “navigate to the company”,”exit navigation”, and “resume playback”. Based on this, for commonly used words with little impact on mistaken awakening, the specific quick command words in the new version are as follows:
Regarding the probability of awakening Ideal-X, based on my usage experience during this period, the new Ideal-X needs to have the word “想” pronounced clearly when it hears people calling it. If the swallowed sound is too obvious, it may affect the probability of awakening.
More Intimate, Candid, and Concise Speaking
When you say “Ideal-X” or press the voice wake-up button on the steering wheel, Ideal-X will answer with “I’m here” or “hi”. The new Ideal-X will have more greeting options to respond to you intelligently based on the scene you have woken it up, such as responding with “good morning” when woke up in the morning, making you feel more intimate.
With the change of the greeting, the expression of Ideal-X has also become more delicate, with clear and significant guidelines for the person who woke it up. When Ideal-X cannot respond, it also shows more emotional expressions of grievance.
When communicating with Ideal-X in the past, it felt a bit verbose, but the new Ideal-X has started to simplify its responses gradually to be concise and explicit. For example, when opening the window while driving, the announcement will be changed to “opening the driver/front/rear left/right window”.
At the same time, when facing questions it cannot answer, Ideal-X also becomes more candid, saying: “I didn’t understand, try rephrasing it” or “I haven’t learned this feature yet, try manually operating it.” The language used is clear and easy to understand.
The greeting tone of Ideal-X when awakening is personified in version 2.1, but during the subsequent answer-execution process, it still has some mechanical tone, and this known issue will be optimized later.
There are also some low-usage functions that have been adjusted, such as how the new Ideal-X tells jokes, which has been changed from an internal joke library to a jump link to Himalaya for joke searching. Weather inquiries can no longer support multiple-day queries for the time being, but only for the current day. In short, the new Ideal-X is not much of a chatterbox for the time being…
New skills, have you got them yet?Translate the Chinese Markdown text below into English Markdown text in a professional way, retaining the HTML tags inside the Markdown, and only outputting the result:
The new Ideal ONE has achieved more new skills in combination with the vehicle control due to higher autonomy. The sound team collected 13 frequently used control items by car owners and matched them one by one, enabling the control of vehicles by just moving your lips.
In the new version, the Ideal ONE can also learn by itself, analyzing and learning from every interaction with car owners and itself. Every conversation you have will become its training material.
Another significant improvement is the ability to update the instruction library online, without over-the-air updates, and directly effective through the cloud, so your Ideal ONE is continuously improving in real-time.
Your training is needed for the growth of the Ideal ONE
The new Ideal ONE is like a child that needs to constantly absorb nutrients at this stage, requiring product development teams and car owners to train it together to make it grow rapidly. During the past month of beta testing, the Ideal ONE has optimized over 200 items just from the suggestions of more than 300 beta-test car owners. Everyone’s frequent use and feedback are crucial to the Ideal ONE when it is fully pushed out.
This change is like a deep squat before the Ideal ONE takes off. In the version 2.2 that is scheduled to be released in August, you will see a higher level Ideal ONE with six new functions: continuous conversation, speak as you want, automatic identification of effective instructions, visible and can speak, free conversation in the entire car, and more natural voice effects. Here is a video to show everyone.
How can anyone not look forward to such an Ideal ONE? To help everyone better understand it, we have prepared a series of articles, including “The Skillset of the Ideal ONE,” “How was the Ideal ONE Created?”, and “The Secret of Voice AI – Artificial Mouth.” We will also open up exclusive topics on the Ideal Community to solicit the voice commands you want. Let’s create a brand new Ideal ONE together.
Thanks for the support of @胡含 @赵彤 @郭晓峰 @贺永强 @安庆涵 on this article.
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.