Feng Se Mingmin comes from the Cave Fei Temple
Qubits | Official account QbitAI
Hello, I'm Brother Kun
The hottest AI technology recently is undoubtedly ChatGPT, I am already a heavy user of it, using it in the work to improve the effect is very obvious in my personal feelings, but also continue to use other colleagues in the company Amway, if you have not used, strongly recommend you also experience, I believe it will greatly improve your work efficiency, yesterday I saw tinyfool through step by step prompts to make ChatGPT a complete function .app.
So do ordinary people have the opportunity to earn millions a year with ChatGPT?
In the past two days, a new position called "prompt engineer" quietly appeared in the recruitment market:
Without writing code, majoring in computer, or requiring academic qualifications, I studied how to "chat" with ChatGPT, and the annual salary actually opened to $250,000-330,000, equivalent to 1.7 million +.
Unbelief? There is a picture and there is a truth:
OpenAI CEO Sam said that if you can write prompts for chatbots, it is a strong skill point.
But I really can't believe it...
Although ChatGPT is now on fire, it is not okay to start a business around it, invest in investment, rob people and rob people, and not admit that it is lively.
But at the same time, all kinds of untouched things are also eyeing it, such as scams and fishing.
This is a prompt engineer, give so much salary, is it really reliable?
New jobs spawned by ChatGPT
Whether it is reliable or not, you have to see what it is for.
Let's put it this way. Think about when you played ChatGPT or even when you used to play AI painting, did you lack words? The routine is simple?
So much so that everyone says they're awesome, but you don't always get very satisfying or amazing results?
This position was born to solve this problem.
In fact, its earliest prototype can be said to have started with AI painting.
At that time in July 2021, the AI drawing effect was generally not out of the loop, and some people found that as long as they added "Unreal Engine" to the prompt, the image quality soared instantly.
Similarly, in ChatGPT, some people have found that if it answers your question incorrectly at the beginning, don't be disgusted, try saying "let's think step by step" to it, and it may be correct!
This is what prompt engineering is all about: to use the right description to make AI reach its maximum potential.
As can be seen from the above two simple examples, this job may not be so difficult, it seems that it is enough to have a brain hole and an exploration spirit.
Now, let's confirm our guess from the job posting mentioned at the beginning.
The job came from Anthropic, the AI company founded by the departure of core GPT-3 members, on which Google spent $300 million and is now valued at about $5 billion.
Anthropic offers such a high annual salary for this position, what are the specific requirements for applicants?
A total of 8 articles:
1. Have creative hacker spirit and like to solve puzzles.
2. Good at communication, like to teach technical concepts and create high-quality documents that can help others.
3. At least have a high degree of familiarity with the architecture and operation of large language models.
4. Have at least basic programming and QA skills, and be able to easily write small Python programs.
5. Have organizational thinking and like to build a team from scratch. Think holistically and be able to proactively address needs.
6. It can make ambiguous problems clear, and find out the core rules in the problems with changeable scenes.
7. Interested in the safety and social norms of the technology, able to predict risks, simulate scenarios, and provide actionable guidance to internal stakeholders.
8. Be able to think creatively beyond established paradigms about the risks and benefits of new technologies, be able to maintain interest in emerging research and industry trends, and stay up to date.
Overall, Anthropic's ability requirements for prompting engineers are relatively general, how can a position with an annual salary of millions look like? (Dog Head)
The only thing that seems hard is clauses 3 and 4: familiarity with the LLM architecture and ability to program.
The former can be temporarily "evilly supplemented", and the latter requirements are also very general - compared with the AI circle of computer related majors in famous schools at every turn, it is simply not worth mentioning.
And Anthropic also said at the end of the announcement:
Not all the requirements have to be met, please don't rule yourself out prematurely, and come if you feel interested.
Yes, dare to feel as long as we have a sense of how to better write prompts, even an ordinary person without a high education, programming average, non-computer major, has the opportunity to rub a wave of ChatGPT benefits to rush an annual salary of millions?
Maybe it's really not bluffing.
In fact, the history of the field of prompt engineering is less than two years, and as a new position in a completely new field, the specific standards can be said to be unknown where to start.
This is also to give us ordinary people the opportunity to enter.
No, just last December, a little brother got an offer from Silicon Valley unicorn company Scale AI because he was very good at guiding ChatGPT to "do things", which should be the first prompt engineer ever.
The little brother doesn't even want his job because of this offer, but it is tempting to make an offer.
Is it universal?
And now more than one company is hiring prompt engineers.
△ When writing, I found another one, from the legal industry
There is also industry boss Andrej Karpathy, Tesla's former AI director (now joined OpenAI), who has also publicly expressed optimism about prompt engineers.
He joked that because of this new position, English is currently the hottest programming language
。
Therefore, I can't help but ask: will there be more and more dialogue prompts for engineers in the future? What about China?
To answer this question, you have to think about what the essence of this position is.
For example, some views believe that the birth of the concept of prompting engineers actually illustrates the biggest disadvantage of conversational AI -
Not smart enough.
Therefore, it needs to be guided by well-designed prompt words to achieve the desired result of the user.
No, netizens have also spontaneously summarized a lot of prompts, and GitHub's related target stars have reached 2.7w.
However, as the technology matures, AI will become smarter and smarter and can directly understand our intentions and give satisfactory answers, so this practice may become unnecessary - then the profession of prompting engineers may become redundant.
What's more, it may not have to wait until that day, once the prompt file is perfected and the whole process becomes standardized, this type of work will not be needed.
Wait, that means doing this, maybe making a quick buck?
I can't make up the mark yet.
But what is certain is that since even the companies established by the core members of GPT-3 need such positions, the "Chinese version of OpenAI" must not need them - and the number may be larger?
Friends, this is truly gospel.
At least, reasonable and legal.
There is a wilder way
Making money through ChatGPT is another way, a bit wild, and not far from the first list of jails.
Since the birth of ChatGPT, there has also been a lot of commotion on the side of hackers and cyber crooks.
Bots have emerged on Telegram that are calling OpenAI's APIs to generate malware and code with ChatGPT. Once the upper limit on the use of these bots is reached, a fee of only $5.50 per 100 is charged.
Other communities have also emerged ways to bypass ChatGPT restrictions. If implemented, it would mean that criminals have no limit to the number of times they can use ChatGPT and can generate malware code.
Check Point found traces of ChatGPT attacks in the hacker community in January.
And I tried several methods to make ChatGPT cyberattacks, and the results were too good to be true.
For example, it generates the effect of phishing emails: from text to code, you can write it yourself, and you can continue to iteratively upgrade according to the requirements.
However, this abuse has also been solved to a certain extent, so hackers call the API instead...
For example, calling the API of text-davinci-003, one of the underlying models of ChatGPT, can quickly steal the user's PDF document through a phishing software and a script and send it to the attacker via FTP.
In short, ChatGPT is still down the threshold for hackers, even technical scumbags can engage in some dangerous actions...
On the other hand, telecom fraud wants to deceive people and deceive money, I am afraid it will be easier.
For example, during Valentine's Day just this past year, McAfee conducted an experiment to send a love letter written by ChatGPT to 5,000 users, and less than a third of them felt that it was from AI. Others either think it's written by humans themselves, or it's hard to tell.
The problem is that ChatGPT is unstoppable, and online chat apps (such as OkCupid) have actively accessed ChatGPT. While company leaders have called for not to abuse AI, no specific policy has been issued to restrict it.
According to McAfee, 30% of survey users choose to use AI to write their own profiles.
Coupled with the AI girlfriend who exploded a few days ago, I am afraid that a new industrial chain will be formed...
Ahem, calm down.
If you can't be the first to get rich because of ChatGPT, then don't be the first to enter the game because of ChatGPT (serious face.jpg).
Looking at it this way, it still prompts the engineer to be more reliable, doesn't it?
Reference Links:
[1]https://arstechnica.com/information-technology/2023/02/now-open-fee-based-telegram-service-that-uses-chatgpt-to-generate-malware/
[2]https://research.checkpoint.com/2023/opwnai-cybercriminals-starting-to-use-chatgpt
[3]https://dataconomy.com/2023/01/what-is-ai-prompt-engineering-examples-how/
[4]https://www.ft.com/content/0deda1e7-4fbf-46bc-8eee-c2049d783259