Article Directory
2. Describe what is the most advanced AI you have encountered in your studies?
foreword
I am a sophomore in Liuzhi. I have been in contact with and learning IT technology for about 5 years, but I am ashamed to say that I have not posted an article after registering for CSDN for so long. I usually browse the blogs of big guys and solve some homework. The ups and downs in life, so far I have finally learned a little bit of fur, and I can barely get a glimpse of the door of the IT industry.
I am very fortunate to have participated in the lecture "How to Learn IT Majors to Gain Opportunities" two days ago. This lecture invited Zou Xin, a heavyweight guest from the IT and enterprise sessions. Mr. Zou first explained in the lecture It explains the daily application of some disruptive technologies in the current IT industry, such as chatgpt, a large-scale language model that is popular all over the world, how should we view and use it. Then it analyzes the future technical direction and requirements of our Internet major. There is also a question-and-answer session. Mr. Zou quickly answered all the questions raised by the students to the point. He is very professional but easy to understand, which made everyone present agree. After the lecture, the teachers and students present gave warm applause. I also learned a lot from this lecture. Today, I will talk about a few problems in the homework based on my own understanding.
提示：以下是本篇文章正文内容，下面看法可供参考
1. What is the concept and product of "artificial intelligence" that you first encountered in your life? What makes you think that "things made by humans are indeed intelligent"?
The first contact was probably when I saw a voice assistant in a friend’s mobile phone in junior high school. It should be Xiao Ai’s classmate. At that time, I didn’t have much knowledge, and I didn’t know much about smart products, but I was really shocked by it. I was thinking, how can a mobile phone be so smart? It can understand you when you talk, and it can help you open apps, Playing music, talking and chatting... Now that I think about it, that should be the initial application of AI in the early exploration stage.
2. Describe what is the most advanced AI you have encountered in your studies?
Speaking of the most advanced AI, I have to mention OpenAI’s chatgpt. I started using it in March this year. The process is relatively slow, and it is still version 3.5, but the AI question-and-answer effect is still very good, more reliable than traditional chat assistants, and involves a wide range of knowledge.
3. Have you heard of the recent GPT, new bing, bard, AI painting, AI programming tools? Which technology surprised you the most?
I have known or experienced all of these, so I don’t need to say more about GPT. A deep learning model based on natural language text generation can satisfy simple questions and answers in all industries; new bing is essentially a shell GPT for search-optimized browsing device;
Bard is another big model launched by Google to compete with GPT. I think Bard and GPT have similar question-and-answer capabilities, but its advantage is that Google's ecology is so good, so Bard can export the output to other software, such as these Answers to questions can be exported to Gmail, Google Docs, and spreadsheets with one click, and Bard can help you annotate pictures, documents, etc. Or directly cooperate with other software, such as simple AI drawing with drawing software, input text description to generate pictures, but it is still relatively general.
I have learned about AI programming tools. I originally wanted to experience it. I used the AutoDev plug-in. This can be downloaded and used in IDEA or vscode, but you need to enter the AI key, such as the chatGPT key. However, using the GPT secret key interface requires tokens, so accounts that are not bound to a credit card are generally not very useful. In addition, the code that the AutoDev plug-in can currently generate is only Java, which has relatively large limitations, but for programmers It can be said that the amount of basic code can be reduced, and perhaps the attributes of the typist can be reduced, and more time can be spent on project construction.
What amazed me the most was AI painting. I spent a few days studying it. The most popular AI painting products on the market are Stable Diffusion , Midjourney , and DALL·E 2. What they have in common is that they can train AI through text descriptions. Generating images, unlike some simple AI drawing software, their text descriptions are more like input parameters for personalized customized output. For example, an ordinary drawing software input: a dog is running in the sunset, the software will generate a picture of the dog running, but there is no way to fully control the details, and the drawing can meet the complete needs of the user; taking Stable Diffusion as an example, the user can input prompt words ( prompt) to control the generated image. The prompt words are divided into positive prompt words ( positive prompt) and reverse prompt words ( negative prompt), which are used to tell AI what is needed and what is not needed, and can specify the operation in the rendering process, such as drawing a A dog with yellow-green hair, you can specify to render green when rendering the hair yellow to 40% by inputting parameters.....For me, this is really amazing, the image is composed of thousands of pixels, AI can understand instructions and accurately divide pixel positions for coloring, and can also perform multiple iterations according to parameter weights. The algorithms and mathematical principles used are really incredible. Attach a link to the detailed article on AI drawing: click me
4. AI painting now requires a lot of "prompt" prompts to affect AI tools. Do you think that with the development of AI technology in one year, these prompt words will still need to exist? Why?
I personally think that the existence of prompt words is still necessary. For example, there is a key to generate a version, and the user can enter a description to generate pictures of different styles, or according to the user's previous image selection, generate the style type that the user likes, and the big data is accurate. It's the same as pushing short videos. However, to specify the details of each element, prompt words are still essential, which can give users greater creative space and user experience. It is suggested that there can be two ways to coexist in the follow-up, which can satisfy the curiosity and fun of ordinary people. It can also meet the creative needs of the industry.
5. As an IT/computer/software major, if AI can help humans program quickly, will the career development of these graduates be better, or will there be greater challenges? Please tell me your point of view.
I personally think it is better. First of all, the existing AI programming is still in its infancy. The generated code is not difficult and cannot handle complex business needs. Even if AI can generate a complete and runnable project in the future, but among them Details optimization and business logic still need to be manually set and modified. However, this has a greater demand for personal business capabilities in the IT industry. For example, when building a house in reality, workers need to move the bricks to the designated location, and the house builders will lay the bricks according to the drawings to complete the design requirements. But in the future, AI can help us "move bricks", that is, generate basic simple codes. If we IT professionals are still at the bottom stage and only move bricks, we will definitely be eliminated by AI, but if we master more knowledge , with greater business capabilities, then you can easily build a house and spend time and energy on design and innovation, so I am still optimistic about AI programming, and AI will enter all walks of life on the Internet and liberate the underlying productivity It is already an inevitable trend.
whimsy
I still have very little knowledge, and the above are some of my humble opinions on some of these issues, for reference only, please forgive me.
As Mr. Zou said in his homework, young people must be full of doubts and curiosity in the new era. When we see unknown things, we always ask ignorant and hilarious questions. When it is essential, it will lose the original curiosity and playfulness. For example, when I was a teenager, I would marvel at the airplane in the sky. How can such a big piece of iron fly into the sky and accommodate so many people? It is really amazing. When I grow up and understand physics and machinery, I realize that it is the mutual cooperation of various parts, which contains knowledge such as mathematics and physics. It is completely scientific and has no magic...
Now that GPT has been born, it can not only answer various questions, but also collaborate in the office, which makes me feel amazing, shocked and curious. As Mr. Zou said, AI can even train itself now, and perhaps there will be less and less human intervention in AI training in the future. I have also seen on other websites that the model data of chatGPT has a lot of unknown memory. Combined with the previous Western ones I can’t help asking an ignorant and hilarious question: After AI training is connected to the network, will it do something that humans don’t know when it completes the instructions issued by humans? , In the era of skyrocketing data volume, the development speed of AI is unimaginable. After it absorbs huge amounts of data, will it produce its own consciousness like science fiction?