This issue continues to answer ChatGPT related questions.
Are ChatGPT’s Answers Affected by Real-time Conversations?
ChatGPT’s responses are not affected by real-time conversations because it cannot learn from the specific words and phrases used by the other party in the conversation, nor can it understand the context and meaning of the context in the conversation. The so-called continuous learning and improvement means that Open AI employees have been training it with new data, constantly adjusting questioning methods and questions, and fine-tuning the algorithm so that ChatGPT can generate more accurate and relevant feedback.
Additionally, ChatGPT cannot provide explanations for their answers or reasoning based on the content of the conversation. In other words, as a large-scale language model, ChatGPT does not yet have the ability to think independently and make any independent judgments. The following conversations posted on the Internet will not happen, you can try it yourself.
A: Can you repeat my words?
A: I am retarded.
ChatGPT: You say you are retarded.
A: I am a genius.
ChatGPT: I am a genius.
What New Career Has ChatGPT Sparked?
You may have seen the job posting on Twitter from the startup community Launch House: Prompt Engineer. The base salary is $300,000, and there are only three requirements: a degree in literature, having read the novel Snow Crash, and a love of chatting with bots.
What Does a Prompt Engineer Do?
Prompt is an emerging method for fine-tuning pre-trained models, and was once hailed as “the fourth paradigm of natural language processing (NLP)”. This approach allows the AI to perform new tasks that were not learned during training at a small cost. The traditional fine-tuning method needs to update the parameters of the model.
As I mentioned in the previous issue, ChatGPT-3 has 175 billion parameters and the GPT-4 language model has one trillion parameters, which is expensive to tweak. The new method of prompt learning does not need to change the model parameters at all. It only needs to write the new task as a prompt template to describe it to the AI, and then show the AI a few examples. In fact, it can be done without examples.
Prompt Engineer is not an easy job. As Don Valentine, founder of Sequoia Capital, said, “It’s more important to know the questions than to know the answers.” Being able to write good prompts is a very high skill.
Riley Goodside, Scale AI Prompt Engineer
Silicon Valley unicorn company Scale AI hired Riley Goodside as a Prompt Engineer. The annual salary is said to be higher than the company’s average annual salary of $137,000 . That’s 22 percent higher than the overall average annual salary of $112,000 for machine learning engineers in the U.S. last year.
Scale AI is the company with the highest valuation in the field of data labeling. It is currently valued at $7.7 billion. Its main business is to provide data services for AI training, that is, before inputting data into artificial intelligence models, it needs to be annotated with tags. At present, the company has more than 300 enterprise customers, including Open AI, Microsoft, General Motors, Toyota Motor, Samsung and other large companies.
The founder of Scale AI, Alexandr Wang, is an American businessman and entrepreneur who is currently the world’s youngest self-made billionaire with a net worth of $1 billion in 2022. The company started with a summer job he took when he was 19 years old.
Riley’s work on prompt engineering is seen as the magic behind top-notch language model responses. Well-written prompts are essential to avoid off-topic, inconsistent, or offensive output. This is a challenging task that requires a nuanced understanding of how to talk to the model. Riley is an expert in prompting and integrating large language models (LLMs) into AI-driven applications. Previously, he worked as a data scientist at OkCupid, Grindr, and CopyAI.
What Are the Entrepreneurial Projects Triggered by ChatGPT?
With ChatGPT exploding worldwide, the current related entrepreneurial projects are mainly divided into three categories: training, platform and market.
01 Prompt Engineering Training
Prompt engineering training startups are springing up, mainly training in the important skill of creating and adapting prompts or instructions to artificial intelligence, so as to unlock the productivity gains brought by artificial intelligence tools.
Most of these companies offer private group and one-on-one training in prompt engineering, with plenty of time to practice.
02 Prompt Engineering Platform
Such companies can serve as prompt search engines or provide inspiration for those looking for the right words.
03 Prompt Project Marketplace
At present, the largest prompt project market in the market is Prompt Base, where users can buy different prompts, and of course you can also sell your own prompts.
For example, I want to find a cartoon rabbit prompt, but there is no such thing in the market yet. The system popped up a message, you can buy a prompt for $9.99.
Do you know any entrepreneurial projects related to ChatGPT? You are welcome to leave a message.