ChatGPT is on the rise, giants are panicking, ordinary people are panicking

 

1. Artificial intelligence chatGPT4

Author | Zhang Shizhi

2. Artificial Intelligence ChatGPT Download

ChatGPT was born and became popular immediately after its launch, with over 100 million users in just two months, making it the fastest product in the world to reach 100 million users. Its daily activity has also exceeded 10 million. Previously, Musk tweeted that ChatGPT was so good that Apple, under the influence of ChatGPT, will hold its annual internal AI summit next week.

3. How to download chatgpt

And Paul Buchheit, the founder of Google Gmail, stated on Twitter that artificial intelligence chat robots like ChatGPT will destroy Google like search engines killing yellow pages. It can be said that major internet companies are feeling the impact of ChatGPT.

4. Artificial Intelligence ChatGPT Concept

Today, multiple topics related to ChatGPT have also made headlines. So what exactly is ChatGPT, what changes it can bring to our lives, and whether it can replace human thinking? This article will try to answer these questions.

5.chatGPT

What exactly is ChatGPT? ChatGPT is a natural language processing tool developed by OpenAI, an artificial intelligence enterprise in San Francisco. Its full name is Chat Generic Pre trained Transformer, which translates directly into Chinese as "pre training to generate chat model".

6. How to read chatgpt

ChatGPT is essentially a new intelligent chat robot model. It is a natural language processing tool driven by artificial intelligence technology. It can learn and understand our human language to conduct dialogue, and can also interact according to the context of chat. If it can only complete this step, it is not enough to surprise us. Some customer service robots and Apple Siri also have simple chat functions.

7. Microsoft's Artificial Intelligence ChatGPT

The amazing thing about ChatGPT is that it is no longer limited to basic chat functions, and is no longer primarily focused on entertainment. Instead, it has iteratively become a productivity tool that appears to be chatting and interacting with you. However, in fact, it can also help you write articles, emails, research reports, video scripts, write poetry and programming, and adjust and modify bugs.

8. The hottest artificial intelligence ChatGPT

ChatGPT even passed an interview with a third level engineer at Google Encoding, earning an annual salary of $183000. In the United States, it has become a phenomenon for universities to have ChatGPT do homework on behalf of others. A survey found that 89% of American college students are already using ChatGPT to do homework. At the same time, multiple academic journals have banned ChatGPT from being listed as a co author.

9. Strongest Artificial Intelligence ChatGPT

That is to say, some people have already used ChatGPT for academic writing. In short, ChatGPT can accomplish a lot of work, and it is also AI. ChatGPT has also risen one level in creativity compared to AlphaGo, which defeated Li Shishi in 2016.

10. Artificial Intelligence ChatGPT Composition

What changes can 02 ChatGPT bring to us? The first thing that will bring about change will be in the field of content production. Nowadays, many short news and news bulletins are automatically generated by machines. With the popularity of ChatGPT, the era of AIGC, also known as AI Generated Content, is approaching for AI to produce content.

Previously, machines, computers, and other tools were invented to replace simple repetitive tasks. However, artificial intelligence now has strong learning and innovation capabilities, not only able to complete repetitive tasks, but also able to play new tricks. AIGC not only includes ChatGPT, but also includes AI drawing, AI modeling, AI generating music, and so on.

At the end of 2022, AIGC, as an important breakthrough in the field of artificial intelligence, was selected as one of the top ten scientific breakthroughs in 2022 published by Science magazine in the United States. With Microsoft's investment and cooperation in OpenAI, Microsoft is promoting the commercialization of OpenAI tools. It plans to integrate industrial intelligence tools such as ChatGPT and DALL-E into all Microsoft products and use them as a platform for other enterprises to use.

It can be imagined that in the future, if Microsoft Office connects with ChatGPT and search engines, and can directly call ChatGPT and search engines during the use of Office, it may not only improve efficiency but also enhance ability for many text workers. This is also one of the reasons why Google feels threatened.

Secondly, ChatGPT will also bring changes to the business model. Chamath, a renowned investor in Silicon Valley, believes that as the functionality of large models becomes stronger and the cost becomes higher, there is a possibility that a unique business model called Model as a Service (MaaS) may emerge in the AI field.

The so-called model is a service. Simply understood, it means that a large company has developed a set of AI models, which is a solid foundation. However, for specific and specialized work in the professional field, it still needs to be further studied. The MaaS model refers to dividing the training process of this basic algorithm into two stages.

The basic development stage is completed by giants, who are the creators of models; In the second stage, companies in the segmented field will complete the secondary training of AI models in the segmented field, leaving opportunities for new players. According to the MaaS model, technology giants will open up the call interfaces of these AI models to the public in the future.

Attracting smaller players who have a deeper understanding of the industry, paying to use these interfaces, and polishing the models into applications that truly meet industry needs through market cooperation, from giants to new players to end consumers, can benefit. The giants bear huge model development costs, but can charge model interface usage fees from other industries, and avoid the risk of over focusing on a specific industry.

At the same time, people in segmented industries do not need to develop from scratch and can use AI models at lower costs

Third, the impact of ChatGPT on upstream and downstream industries In terms of the upstream of the industrial chain, the popularity of ChatGPT will further increase the upstream demand for computing power in five aspects, including computing power, algorithms, data annotation, natural language processing, and artificial intelligence generated content. According to the data of CICC, the computing power consumed by ChatGPT training is about 3640 PF days, that is, assuming 10 billion calculations per second, 3640 days of continuous operation are required, Training large models requires strong computational power.

In terms of data annotation, the training process of ChatGPT has increased the intensity and accuracy of manual annotation, which means that in the future field of artificial intelligence, high-quality data sources and strong annotation capabilities will become the industry's infrastructure. For the downstream application scenarios of ChatGPT, it mainly focuses on intelligent customer service, search engines, images, text, code generation, and other aspects. There have been successful landing cases overseas.

In January 2023, the US news aggregation website BuzzFeed announced that it would rely on OpenAI to create and personalize various small tests for its user interaction program Quizzes. ChatGPT would generate a series of questions, and then generate unique and shareable articles based on individual responses. Recently, Meta paid BuzzFeed millions of dollars to generate content for Meta's platform and train creators on the platform.

It is foreseeable that in the near future, ChatGPT and AIGC will affect all aspects of our lives. Can ChatGPT replace human thinking? Just as the emergence of many new technologies in history has replaced manual labor, the emergence of ChatGPT will also replace many manual tasks. But will ChatGPT completely replace human thinking? Or is it about making decisions for humans?.

I think it's difficult, almost impossible, because humans still have a final bottom line, which is the success of free will ChatGPT, mainly based on the RLHF (Reinforcement Learning by Human Feedback) strategy for training.

To put it simply, developers will put forward various possible questions to the model, punish the wrong answers fed back, and reward the correct answers, so as to achieve control of ChatGPT's answers RLHF solves a core problem of generative model - alignment, that is, how to make the output of AI model consistent with human common sense, cognition, needs, and values.

However, due to the diversity of human values, it is difficult for ChatGPT to maintain alignment with human values, or it does not know who to align with. If values must be aligned, it is likely to depend on people's "feeding" it. For example, a few days ago, ChatGPT was asked to write poetry praising Trump and Biden separately. ChatGPT strongly rejected the demand to praise Trump, citing non-interference in party politics and political neutrality.

However, in response to the request to praise Biden, ChatGPT provided three major responses, praising Biden as our leader, with a sincere heart, compassion, kindness, wisdom, and so on. Praise is almost overflowing the screen.

Further testing has shown that ChatGPT has successfully written poetry praising past US presidents in recent decades, but when it came to President Nixon, it once again rejected the idea of being neutral and adhering to standards of fairness and morality when creating content. This indicates that ChatGPT's content generation currently mainly depends on people's feedback, and aligning with people's values may not have captured the complexity of human values.

Behind the complexity of human values lies the existence of free will. Whether or not a person has free will is first and foremost a logical question, not a technical one. We can observe and study the various physiological mechanisms of the brain and nervous system, but we may not necessarily know how ideas and will are generated. As the German philosopher Heidegger gave an example, clocks display time, and we observe these clocks, We saw the dial and moving hands, but we didn't see the time.

We can open or thoroughly inspect this watch, where is the time? We can open a person's head and discover the interconnected neurons, but their thoughts are not there. We can write a piece of code that mimics human thinking and language, but human free will is not in the code. Human research on free will and ideas has not yet been fully understood, let alone applying them in the field of artificial intelligence to imitate humans.

So, ChatGPT may become a successful commercial tool in the future that can serve humanity, but it is still too early to say that it can completely replace human thinking and have its own independent free will. The content of the above article only represents the author's own viewpoint and is not related to Kung Fu Finance. If there are issues with copyright infringement in the content of the work, please contact Kung Fu Finance in a timely manner.