How Hot is ChatGPT
How hot ChatGPT is, we can take a look through Baidu Index.
With the sweeping of ChatGPT, many people in the industry in China have heated discussions, including seeking to join. Typical is former Meituan co-founder Wang Huiwen. Two days ago, he released the "Artificial Intelligence Manifesto" in the circle of friends, saying that he would join the group with a capital of 50 million U.S. dollars . onlookers and investors. Shen Peng, the founder and CEO of Waterdrop, commented in Wang Huiwen's circle of friends: "The big guy's investment is so simple and boring."
On 2.13, Wang Huiwen released the AI Hero List again, revealing: all-in AI, the company is valued at 200 million US dollars away from light years away, and he invested 50 million US dollars. Accounting for 25% of the shares. In the next round of financing, top VCs have already subscribed for 230 million US dollars. For a while, the hormones in the venture capital circle were once again stimulated.
According to the latest news, Liu Jiang, former dean of Meituan Technical College, has announced to join Wang Huiwen's team.
What jobs can ChatGPT replace?
Many self-media are listing which jobs ChatGPT can replace? Programmers are clearly not. Because programmers spend half of their time discussing and analyzing requirements and designing. An average monthly code output of 2k is not bad. Don't simply think that being able to generate code can replace programmers.
Another one, ChatGPT is suitable for single-stage work, not for more complex multi-stage work. As far as program development is concerned, let's look at an example.
Write a seckill program:
// 定义商品
let product = {
name: 'iPhone Xs',
price: 7999,
amount: 100
}
// 定义秒杀函数
function seckill(product) {
// 判断商品是否已售罄
if (product.amount <= 0) {
console.log('商品已经售罄');
return;
}
// 判断是否有足够的库存
if (product.amount > 0) {
// 进行秒杀
console.log(`恭喜你抢到了${product.name}, 抢购价格是${product.price}`);
product.amount--;
}
}
// 执行秒杀
seckill(product);
```Write a flash program in java language:
秒杀程序如下：
// 定义商品类
public class Product {
public String name;
public int price;
public int amount;
public Product(String name, int price, int amount) {
this.name = name;
this.price = price;
this.amount = amount;
}
}
// 定义秒杀函数
public static void seckill(Product product) {
// 判断商品是否已售罄
if (product.amount <= 0) {
System.out.println("商品已经售罄");
return;
}
// 判断是否有足够的库存
if (product.amount > 0) {
// 进行秒杀
System.out.println("恭喜你抢到了" + product.name + ", 抢购价格是" + product.price);
product.amount--;
}
}
// 执行秒杀
public static void main(String[] args) {
Product product = new Product("iPhone Xs", 7999, 100);
seckill(product);
}Write a spike program. 10 million people grab 3 iPhones. Consider user experience. Java language.
Front-end frameworks are not limited.
import java.util.ArrayList;
import java.util.List;
import java.util.Random;
public class Seckill {
// 定义参与抢购的用户数
public static final int USER_NUM = 10000000;
// 定义有几个商品参与抢购
public static final int PRODUCT_NUM = 3;
public static void main(String[] args) {
// 存储参与抢购的用户ID
List<Integer> userIds = new ArrayList<>(USER_NUM);
// 存储商品ID
List<Integer> productIds = new ArrayList<>(PRODUCT_NUM);
// 初始化用户ID
for (int i = 0; i < USER_NUM; i++) {
userIds.add(i);
}
// 初始化商品ID
for (int i = 0; i < PRODUCT_NUM; i++) {
productIds.add(i);
}
// 抢购开始
Random random = new Random();
while (productIds.size() > 0) {
// 随机抽取一个用户
int userIndex = random.nextInt(userIds.size());
Integer userId = userIds.get(userIndex);
// 随机抽取一个商品
int productIndex = random.nextInt(productIds.size());
Integer productId = productIds.get(productIndex);
// 将商品从列表中移除
productIds.remove(productIndex);
// 输出抢购结果
System.out.println("用户：" + userId + "抢到了商品：" + productId);
}
}
}It is necessary to know that the specific seckill program must be completed through the cooperation of front-end and back-end/database/cache. A requirement is not equivalent to a program, it may be the result of several components and services working together. ChatGPT's answer is basically of no reference value.
ChatGPT can create poetry, write notifications, write public accounts, and generate content. But particularly high-quality content is unlikely to be done through ChatGPT. You can think in reverse, if the official account you write is a saliva about how programmers earn 1 million, it can really be replaced. Take poetry as an example, "The lonely smoke in the desert is straight, and the sun is setting in the long river", this kind of artistic conception cannot be achieved unless there is a weather in the chest. "Chicken sound Maodian month, human footprint Banqiao frost" is not just a static superposition of four nouns, but a static writing.
ChatGPT is overvalued in the short term
In the short term, ChatGPT is overrated. Several giants have come to an end, not to mention that ChatGPT has already been out of the circle, social experts, not talking about ChatGPT is not enough to integrate into the Internet. 2 days before school starts, attend parent meeting. The principal also mentioned the impact of ChatGPT on students' homework. Even with such intimate friends, is it because the teacher has been replaced.
Teachers, preach and teach karma to solve doubts. ChatGPT allows us to see a trend that human beings are inferior to AI based on the information and knowledge of memory. There is a huge knowledge base behind AI, which can be retrieved, and now it can further generate content.
Sheng Qi, a senior consultant of Puyuan Banking, shared that ChatGPT cannot create new knowledge. ChatGPT is not a god-like product, or it cannot create new knowledge within the scope of knowledge that humans can understand.
There are many companies in China that are doing AI intelligence, and there are also many projects that are doing knowledge graphs in practice. Our team was also fortunate to participate in the construction of a knowledge map application project of a joint-stock bank. At that time, it was only for data analysts to provide knowledge questions and answers in the field of data assets, such as data indicators, and more to identify semantic entities and establish associations with data entities. This association is based on the results of the semantic map, combined with the context of the question, defines the dimension of the data entity, and finally automatically calculates the multi-round question-and-answer form of the index value. ChatGPT undoubtedly goes a step further, directly giving a more comprehensive answer, and there must be data support behind it. If ChatGPT does not have data, I think it will not be able to perform strongly and well.
The big explosion of knowledge brings about a big data explosion, and ChatGPT is still a knowledge-based data product after all.
The author once asked ChatGPT about the asymptomatic proportion of the new crown, and the results of many questions and answers were inconsistent. ChatGPT does not have human intelligence and judgment, it just searches, speculates and adjusts based on feedback.
ChatGPT is hot, should enterprises and entrepreneurs enter the game? Zhang Chaoyang, the founder and chairman of the board of directors of Sohu, talked about it in a conversation with Zhou Hongyi, the founder of 360, a few days ago. Zhang Chaoyang expressed the need to be cautious. " ChatGPT has been accumulated for many years, and it involves many issues such as how many servers are required for computing power, knowledge bases, annotations, etc. If companies without these capabilities follow suit, they will consume a lot of resources. But companies with artificial intelligence AI and search capabilities , still need to have such a preparation .”
So, which companies can seize this opportunity? Zhou Hongyi believes that it depends on whether it has invested in server computing power in the past ten years, whether it has an artificial intelligence team, and whether it has big data storage. Zhou Hongyi also believes, "Bubbles are always not a bad thing. There will be more outstanding talents and more funds will flow in."
Everyone knows about Microsoft Xiaobing. Li Di, CEO of Xiaobing Company, was interviewed by "Chinese Entrepreneur" and expressed his opinion on the question "ChatGPT triggered a new round of China's technology and Internet competition. Which ones are really going to be done and which ones are bubbles" own point of view.
Where does the bubble come from? Bubbles come from a serious imbalance between a thing's real social value, commercial value, and people's expectations of commercial value and social value.
1. It is relatively perceptual to think that ChatGPT is conscious. This is a serious imbalance. The bubble largely depends on the self-discipline of all of us. Don’t think that artificial intelligence has developed to this stage today.
2. The second bubble refers to its commercial value. Artificial intelligence still has many unresolved problems in the big model, such as information accuracy and cost. If the amount of interaction is very large, like Google, if it directly uses a large model, then its profit will drop enormously, so you might as well not use a large model.
3. Speculators who are constantly in the air, they will speculate on whatever is hot at the moment, which is actually another aspect of the bubble.
Xiaoice CEO Li Di: Don’t talk about ChatGPT, start-up companies can’t make it
ps: According to reports, Xiaoice is currently the artificial intelligence system that carries the largest amount of interactions in the world, accounting for more than 60% of the total global interactions. In July 2020, Microsoft announced that it would spin off Xiaoice as an independent company, and appointed Li Di as the CEO of Xiaoice.
ChatGPT has been underestimated for a long time
Chatting with some friends who do AI, ChatGPT is still not strong artificial intelligence. Strong artificial intelligence should be able to solve mathematical equations and physical phenomena based on the input of rules and principles. ChatGPT still needs to rely on a large amount of data corpus and annotation.
But ChatGPT can be called a revolution in the world of weak artificial intelligence.
Xu Zhenhu, deputy general manager and chief data officer of Cover Technology, publicly expressed the following views in Cover Media's report.
1. ChatGP T brings a new round of AI technology paradigm innovation. The birth of ChatGPT is very anxious for technology giants such as Google and Baidu, but the impact of ChatGPT does not stop there. What really surprised everyone was the ability to create and learn from large models behind it.
2. Compared with other large models, ChatGPT's changes in AI technical route and training methods have brought unexpected breakthroughs in the industry. It includes a very powerful technical base, that is, the InstructGPT model. At the same time, a breakthrough has been made in the reinforcement learning mechanism introduced, and it is also very particular about data quality and diversity. Just because of the gap in the basic model itself, although many trillion models or hundreds of billion models have been trained at home and abroad, the adequacy of the training is far from enough.
Someone asked a large model of a large AI manufacturer and ChatGPT with the same question. ChatGPT far surpassed the opponent in terms of logic and completeness of the answers. content.
An analysis by Guotai Junan has listed the advanced nature of ChatGPT.
AIGC represented by ChatGPT has a relatively broad space, but the overall application is still in the early stage. Explorations on AI ethics/compliance and legal responsibilities are also integral parts.
Huaxi Securities stated that GPT is the representative direction of commercialization of AIGC.
The commercialization directions that AIGC has realized include AI writing, AI drawing, and AI underlying modeling. The direction of future commercialization includes AI production video and animation.
Summarize:
1. AIGC, represented by ChatGPT, represents a new paradigm of artificial intelligence
2. AIGC is still not a strong artificial intelligence model
3. For large companies to do ChatGPT, technology and data may be able to break through, but the difficulty is how long they can last and commercialization
4. Small companies do ChatGPT, most of them can't make it out.
5. ChatGPT is overvalued in the short term. Some people make money by selling accounts, tutorials, and training. It can be seen that there is a big bubble; it is undervalued in the long term, and we look forward to the next round of revolution. Looking forward to ChatGPT in China.
Past recommendations:
Ali Senior Expert: My Understanding of Technical Architecture
From ChatGPT's popular memories of the trip to Silicon Valley 6 years ago
technical trivia
Based on distributed design, architecture, and system thinking, it also discusses bits and pieces related to R&D, not limited to code, quality system, and R&D management.