How lazy can people be now?Ah, no, how smart is that?(After all, the fundamental difference between humans and animals is whether or not tools can be made and used.) If you encounter anything, you can always ask big or small. You don't even need to enter it manually, because language input already simplifies this step for you.
Writing **, writing speeches, making PPTs, formulating travel strategies, and even brain teasers and emotional problems. It can help you out.
All you need is one AI!Speaking of which, we know that ChatGPT has completely brought the AI of large language models to a big hit, but you don't just know ChatGPT, right?In an outdated phrase: that's really out.
First of all, a brief foreshadowing, what is a large language model?(For a long time, you can skip it if you understand.) ）
Generally speaking, large language models, also known as LLMs, refer to pre-trained language models with a size of more than a certain magnitude, and deep learning models based on massive unlabeled text data. In layman's terms, it is to train through a large number of parameters, and then adjust and feedback what it has learned to you under your instructions. (Input, Learning, Integration, Output).
This article takes a look at how three of the more well-known large language model AI "fight on the field and show their powers". When it comes to large language models, ChatGPT is "bearing the brunt".
Remember, it was a night on November 30, 2022, when purple MSI fell from the sky to reorganize the Internet. In just 5 days, millions of users have registered, and it is well-deserved to become "phenomenal popularity".
You think it's just a flash in the pan?no,no,no!It's definitely a seeded player, it's just a crazy turn and a cool cow is fried, isn't it?!(HHH is a bit exaggerated).
1.Extremely strong language processing skillsWhen you use it or learn about it, you really have to sigh :!So awesome?It can achieve near-human language comprehension and generative ability, is largely barrier-free, and the dialogue feedback is of high quality. The semantics, grammar, and pragmatics have reached a high level, the dialogue logic is clear, the context is coordinated, and the ability to complete multiple rounds of dialogue under the continuous optimization of instruction output, and the memory is very strong.
2.Convenient way to interactAptly named:"chat"。It's fun to have a conversation with the user (e.g., sometimes you feel like you're playing a monkey). Anytime, anywhere online access, all questions are answered, messages are replied in seconds, the attitude is sincere and cordial, and most of the solutions provided are reliable and effective - it is difficult to find them with lanterns!
3.It covers a wide range of fields
ChatGPT relies on large language models, and the large language model is not only because of the huge training data but also the huge number of parameters (which can be understood as the knowledge learned by the model during the training process). The corpus of its training covers a wide range of materials, from books, news articles, science** to Wikipedia, social**, it dabbles in everything. In other words, it's like an encyclopedia, and it can help you answer everything from programming professional work to emotional entanglements. 4.Highly intelligent
Far ahead in terms of user intent, knowledge links, etc., it can simulate human intelligence, although it is still weak artificial intelligence, but you will not find it too retarded. 5.Painting is powerful
Compared with other AIs, ChatGPT's level in painting can be said to be a dimensionality reduction blow. As long as you output simple statement instructions, you can draw a high-quality picture for you. However, sometimes due to differences in language habits, you can't find a suitable word to be found. For example, dumplings may be understood as buns. 1.The phenomenon of "hallucinations" and reliability issuesWhat is the phenomenon of "hallucinations"?
The AI responds confidently as if it doesn't match its training data, and the common symptom is: serious nonsense. The psychological term is: Fictional disorder. It may fabricate facts and generate false information, and its authenticity and reliability cannot be guaranteed.
2.Logical reasoning is still weakAs the big brother who is in the front, we are still sure of ChatGPT's reasoning ability. However, in the face of complex logical reasoning or fuzzy problems, it can also be deflated.
3.Sensitive, extreme, or even harmful speechJust like the information on the Internet is mixed, minors may receive bad information, and ChatGPT also faces such a problem under unsupervised learning. Sensitive, extreme, and harmful information contained in the pre-trained corpus may be collected and amplified and fed back to the user. It leads to ideological and behavioral extremism and endangers social stability. （emm...Zoom in a bit)
4.Difficult to deployAs a super-large model with hundreds of billions of parameters, ChatGPT is really expensive, with high development and operation costs, and extremely high requirements for computing power for training and deployment. Otherwise, you won't be able to take it
In addition, after all, as a foreign application, domestic use is relatively troublesome, and the universality is not strong. 5.Legal issues –Intellectual Property, Privacy
ChatGPT's answer does not indicate the source of the information, and the data has been processed, which infringes on intellectual property rights, and the current legislative protection of the law for related fields is not perfect enough.
In addition, in order to further optimize ChatGPT, OpenAI may excessively collect users' personal information and data during the interaction process, which may infringe on personal privacy.
6.Socio-ethical issues – replacing sexual panicAt the beginning, ChatGPT was born, and many people were worried about whether they would be replaced by it, from low-level legal professionals to programmers, as if everyone felt the danger. In fact, in general, there is no need to be too anxious at the moment, after all, it is still in the stage of weak artificial intelligence. All that can be done is to collect and integrate data, and it has not yet risen to the stage of creation.
For example, programmers know that the programs written by ChatGPT are full of vulnerabilities and will not rely on it completely, so it can be seen that it can only be used as an auxiliary tool at present. However, if you don't try to use AI at all, you will inevitably lose your competitiveness and risk being replaced. After all, when both people and AI can do it, AI may be more cost-effective Needless to say, who the capital chooses. And, even if AI doesn't replace you, people who use AI will. Everyone is rational and polite. On March 16, 2023, the "Wenxin Yiyan" was officially released, and it will be fully opened to the whole society on August 30. The AI industry, which has the most investment from China's major manufacturers, is still attracting much attention, and at that time, everyone was waiting for this product that benchmarks ChatGPT. Whether we are looking forward to it being ugly or brilliant, we always have the mentality of watching the excitement and not worrying about it. 
To put it simply, it has the most basic qualities of artificial intelligence. It also has some advantages of ChatGPT mentioned earlier, but the level can be said to be dwarfed. It has now been upgraded to 4Version 0, exclusive to distinguished members, enjoy 100 questions and answers every 3 hours. Here are some of its individual benefits. 1.Regular channels, approachable
Wenxin Yiyan is a domestic platform that is more suitable for the physique of Chinese babies. A point of immediate reality is that it has convenient access channels and strong universality. For ordinary people, using ChatGPT may require a wave of operations to find a channel, but Wenxin can easily win it with a word. Just log in to the official page, which is more convenient.
2.Five modules, wide range of practicalityLiterary creation, commercial copywriting, mathematical calculation, Chinese comprehension, and multimodal generation are its five major use scenarios. The needs of daily life and work can be met. In line with the habits of Chinese people, there is no technical obstacle in the use of control.
3.Rooted in China, strong Chinese comprehension skills
One side of the water and soil raises one person, and one AI benefits one person. As a large language model in the local market, it has its own advantages in Chinese language and Chinese culture. There are no language barriers, and there are no cultural differences. For example, if you use Wenxin Yiyan, you don't have to worry about it understanding "dumplings" as "buns". Even obscure Chinese characters are easy to understand. In terms of poetry, it can be said that it is better than foreign AI, and it is still a native to know the aesthetic meaning of the Chinese.
4.Strong real-time performance, keep up with lifeIt can support the implementation of hot search Q&A and has strong timeliness. In addition, you can answer the latest information such as the weather conditions in a certain city and the current stock price situation on the day. This is an advantage that ChatGPT does not have, after all, the Internet environment is different, and it still has an advantage in China.
5.Sensitive word blocking to avoid controversyWenxin Yiyan will automatically block the submitted sensitive words, such as inappropriate characters, words and other content will be restricted.
6.Collaborate in real-time to empower high-performing teamsMulti-person collaboration can be realized in Wenxin Yiyan, which greatly facilitates the communication and modification between team members and makes it possible for the team to use.
1.Poor memoryPoor sentence memorization, sometimes unable to remember the previous content in multiple rounds of questions and answers. As a result, a lot of people prefer to use it as a search engine.
2.Weak drawing abilityThere is still a long way to go side by side with ChatGPT (emm.).Look at the picture and speak, the texture of the painting is human, the abstraction of the characters is high, it may be 3The version of 5 is too low).
In addition, often not getting your points, drawing is not what you want, which is also an important improvement point.
Of course, it also has all the common problems of AI, and the previous summary of ChatGPT is also partially applicable to Wenxin Yiyan. The blue heart V in OriginOS 4 is no longer limited to a small chat box, but integrates AI and systems, which is definitely the most practical in AI applications. Its biggest feature is that it is embedded in the mobile phone system, breaking the permission wall. Sidebar, floating icons, widgets three ways to wake up, on call - your intimate smart assistant!
1.Cleared the system permission wallWith the permission to manage local files, you can control the mobile phone in the window of Xiao V, search for the ** in the album, and even send it to the circle of friends, you just need to tell Xiao V. (If the voice input is done in the cloud, it also needs to be sent with one click) is like a console, similar to a command center.
As mentioned earlier, it breaks down the system permission wall, which means simplifies operations. Then the convenience is greatly improved, even the elderly at home can easily operate. Since then, waiting is no longer a pity, and technology no longer forms a chasm Suitable for all ages!2.One-click service, convenient and fast
One-click P picture, one-click text to send to the circle of friends, one-click to find APP permissions, one-click to find albums**, one-click to summarize document content. Mobile phone use, all one-stop getting. 3.Critical summary, saving time and effort
Mind map, SWOT analysis for web pages, ** and other content, compared to reading slowly, small V can easily help you draw key points. It's really practical to use in an emergency, and of course, it's a good choice to be lazy. 4.The translation conversion function is strong
It can turn out Chinese in real time, break the language barrier, solve the problem of language comprehension, no need for translation, no need for frequent operation, you only need a blue heart V. 1.Limited by mobile phone systemSuccess is also Xiao He, defeat is also Xiao He. The blue heart small V landed in the mobile phone system and has been widely used in practice. However, its audience may be much smaller than the first two, and the entrance is narrower. 2.The parameters are smaller
The main cloud model of Blueheart V, with only 70 billion parameters, is far smaller than the 1,760 billion parameters shown in the public information of ChatGPT4. It's a small comparison, of course, and it's not exactly on the same track.
3.The overall level needs to be improvedAfter all, it is built-in on the mobile phone, and the requirements should not be too high, for the positioning of the mobile phone smart assistant, it is indeed not bad, and it is very cost-effective. However, depending on the identity of the large language model, the blue factory still has a long way to go.
I want to talk about a lot of content, but the space is limited, so I hurriedly put down my pen here.
Which of the many AIs is stronger?There is no basis for empty words, and only practice can bring true knowledge. ChatGPT4 can be said to be invincible all over the world at present, far ahead. It may be difficult for some domestic language models to catch up, but in terms of the overall situation, it is still hoped that domestic AI can develop like mushrooms after a rain and chase after it. Furthermore, I believe that the programmers of the big factories are not vegetarians!Although the domestic AI is still poor, over time, it can also have greater achievements!