On May 6 this year, iFLYTEK released the iFLYTEK Spark cognitive model.
In the context of ChatGPT's hot out of the circle and the global wave of technology set off by the AI cognitive big model, the Xunfei Spark cognitive big model showed a good product completion and interactive experience "at the beginning of its debut", and the IT home also did the experience test for the first time, which felt quite amazing.
In the more than a month since the launch of iFLYTEK Spark, it has continued to evolve. For example, on June 9, the upgraded version of iFLYTEK Spark V1.5 was officially released, ushering in a new upgrade in many aspects such as open knowledge Q&A and multi-round dialogue.
At the same time, iFLYTEK also launched the Spark App, Spark Assistant Center, Spark Language App, etc.
Today, IT House may wish to do an experience test for the upgraded version of iFLYTEK Spark Cognitive Model V1.5 to see if the fully evolved iFLYTEK Spark is really better to use in terms of experience.
1. Experience the basic functions of iFLYTEK Spark App
With the upgrade of the iFLYTEK Spark Cognitive Large Model V1.5, the iFLYTEK Spark App has also been launched simultaneously, so this experience is mainly carried out on the Android app.
It is worth mentioning that iFLYTEK Spark is currently the only large model that supports full coverage of five terminals (Android, IOS, mini programs, PC, H5), supports "cross-device history synchronization", and users can freely switch between such mainstream system platforms.
Taking the Android version used by Xiaobian as an example, the relevant privacy policy will pop up when you open it for the first time, and then enter the login and registration page. If the machine is not logged in or registered, you need to complete the relevant login and registration steps.
If you have logged in, the login status will be recorded, and you will be directly entered into the main Chat interface after the page is launched.
The most important thing in the main interface is the dialogue interface, the chat box provides some dialogue suggestions, the input box is divided into text input and voice input, and the "+" button in the upper right corner of the interface can be switched between the dialogue interface and the call assistant interface after clicking.
Click the setting button in the upper left corner to enter the setting interface, the first layer is the function settings, where you can create a new conversation window, switch the active speaker, and view the historical conversation and assistant list. Click Settings at the top right to enter a more detailed app function setting interface.
On the whole, the interface design of iFLYTEK Spark App is relatively simple and easy to use.
IT House tried to ask a few questions casually, and the big model of Spark can quickly answer, and the experience is almost the same as before on the web page, very smooth.
In the iFLYTEK Spark App, there are buttons for voice reading, regeneration, copying, sharing, liking and stepping under each answer.
Voice reading is a feature of iFLYTEK Spark App, click the voice playback button, you can hear the AI anchor's reading voice is very natural, the tone is very close to the real person, and there is background music by default.
In addition, you can also switch the reading host in the previously mentioned settings interface to find your favorite voice.
Second, the AI assistant function experience
iFLYTEK Spark Cognitive Large Model V1.5, Xiaobian feels that one of the most worthy functions to share with you is the introduction of AI assistant. When the version is upgraded, there are more than 200 song AI assistants online Xinghuo App, they can not only quickly complete the writing of wedding speeches, but also write copywriting in the circle of friends, do resume polishing, become a weekly assistant, and even incarnate as a love god assist...
These AI assistants cover a variety of scenarios such as "workplace, life, travel, writing, fun, and emotion", and can help users apply AI models to solve problems anytime, anywhere.
So how exactly to use it?
First of all, you need to log in to the home page of iFLYTEK SparkDesk on the PC web side, in the upper left corner, click to enter the "Assistant Center".
On the Spark Assistant Center page, you can see the different categories of AI assistants that have been on the shelves, find the assistant you want to use, click "Add", and it will appear in the assistant list under your own account.
For example, Xiaobian added two assistants, "Writing Assistant" and "Little Story", and then on the iFLYTEK Spark App, you can see them appear in my assistant list.
When talking to the Spark model, you can click the "+" button in the upper right corner to invoke the assistant you added, or you can directly enter "@" in the dialog box, and the interface for calling the assistant can also pop up.
For example, the editor wants to call the "writing assistant" during the dialogue, let it help me write an 800-word essay with the title of "reject mediocrity", after entering "@" in the input box, a few simple steps to get a complete 800-word essay "Reject mediocrity", the whole process is very natural and smooth.
In addition, Xiaobian also added the "Math Problem Solving Assistant", this assistant has the function of automatically identifying the questions on the picture, you can take questions or upload photos of the questions, and then manually select specific questions, you can identify and answer. During the test, IT House uploaded a math problem, and the AI assistant was able to identify the question very accurately and quickly gave the correct answer.
In addition to choosing the AI assistant you need in the assistant market of the Spark Assistant Center, you can even create your own assistant.
Click "Assistant Creation Center" at the top right of the Assistant Center page on the web page to proceed with the process of creating an assistant.
According to the prompts on the page, fill in the assistant name, assistant type, assistant answer, function description, assistant instruction and other related content, and give a more accurate description of the functional characteristics of the assistant you want to create.
For example, if you want to create a "history assistant", the assistant type is "comment", the assistant selects "need context understanding" when answering, and the function description is filled in "enter historical events and let me comment".
This descriptor appears in the dialog box when using the assistant, letting everyone know what to enter.
In terms of the assistant's instructions, fill in the "You are a historical expert, please give a comment based on the following historical events {{}}}, requiring clear views, fluent writing, and no more than 500 words."
The assistant instruction is mainly to facilitate the large model to understand the main functions of this assistant, if you do not understand the assistant instruction writing, the creation page also provides a detailed assistant instruction writing guide.
After completing the filling in of this information, you can also debug the assistant function in the "Debugging and Preview" box next to it, which is convenient for timely discovery of problems, such as Xiaobian found that the limit of 500 words or less when the assistant instruction was written during debugging was somewhat inappropriate, and it was deleted.
After debugging, there is no problem, you can click "Create".
If you want to list the AI assistant you created and share it with other users, you can also click "Create and apply for listing". Xiaobian is to apply for the "History Assistant" on the shelf, and after completion, you can see that the assistant just created is under review on the Create Assistant page.
The review time is about less than half a day, not long or short, acceptable. After passing the review, in the assistant list on the Xinghuo App, you can also see the AI assistant you just created and has been put on the shelf, and you can also use it normally.
For example, the editor used the "historical assistant" created by himself to let it comment on the "Battle of Red Cliff", which quickly gave a rich and complete review content, and the views were clear and logical.
In general, calling the function of the assistant and the assistant creation center introduces the concept of "co-creation" for the iFLYTEK Spark model, so that the "good ideas" used by everyone for the large model can be presented through the assistant and shared with other users, and the new Prompt "application development" scenario, through the large model to reduce the threshold of application development, so that creativity can be used immediately, which greatly expands the application scenarios of the iFLYTEK Spark model, so that the functional potential can be fully released.
Third, real-time search and voice input function experience
Another important capability of the Xunfei Xinghuo cognitive large model V1.5 upgrade is real-time search. In the past, many large models could not answer many questions related to recent hot events due to the relatively old database. The real-time search of iFLYTEK Spark model can greatly improve the ability to answer current hot events.
During the test, IT House selected several relatively new questions, such as when "Spider-Man: Across the Universe" will be released, and it quickly gave accurate answers.
For example, "In which cities will Mayday's 2023 concert be held", it can also answer.
Even the college entrance examination that just passed, asked him "what is the essay question of the 2023 national college entrance examination paper A", and asked it to analyze the meaning of this essay question. It also answers accurately and beautifully.
It can be seen that after the upgrade of iFLYTEK Spark Cognitive Large Model V1.5, it does have the ability to keep up with current events, and IT House feels that this greatly enhances the practicality and playability of iFLYTEK Spark Large Model.
In addition, iFLYTEK Spark App also supports voice input and free dialogue. If you don't want to type, just press and hold the voice input button in the input box, and the app will convert the speech to text in real time, let go after the voice is finished, and the text question will be sent.
Based on iFLYTEK's powerful speech recognition and transcription technology, the recognition and transcription of human voices by the large model is very accurate and fast, and its answer is also natural and smooth, and the whole process will have a sense of realism in dialogue with real people.
IV. Conclusion
In general, the Xunfei Xinghuo cognitive large model V1.5 has brought the ability upgrade visible to the naked eye, especially the introduction of AI assistants, as well as real-time search capabilities, making it a great increase in practicality. This allows us to see that AI large model products are not only used to play and taste, but can already be applied in many life scenarios, and they are also very easy to use.
At the same time, iFLYTEK Spark model took the lead in realizing the full coverage of multi-terminal App clients, which is also very friendly for users, and you can use it anytime, anywhere, as if you have a smart assistant with you, which is obviously a big plus.
In short, iFLYTEK Xinghuo Cognitive Big Model V1.5 allows us to see that the large model product does have unlimited potential in terms of application, and we look forward to its continued evolution to bring us more useful and practical functions.