Netizen captain, artificial intelligence chatGPT: The new internet celebrity ChatGPT is the toughest

 

The Internet isoneA field full of surprises and changes, with new events happening every day. Let's put aside our busy daily routine,oneExplore what the Internet has brought us recentlyoneLet's have some fresh things.

"Our AI is really awesome" article/The AI filter that burst out a while before Zao got up and went to bed made the widely acclaimed AI seem to have been labeled as "artificial intelligence disabled" again. The simple AI that is too stylized and low intelligent has produced a large number of anime photos in the rapid generationabsenceLess horror film level joke images.

Behind these ironic works1In a vastly different field, there are also AI that have quietly entered the topiccenterHas become a hot topic among new internet celebrities: ChatGP, if you want to use1Simply put it on ChatGPT1A tag, which is probably "Chatbot"

It was launched two months ago by OpenAI, which sacrificed DALL-E2 in the painting AI battleWe have introduced the language AI model to help netizens write papers and obtain A's story. This ChatGPT can be seen as that version of "GPT⑶”The upgraded version of can provide appropriate responses on more topics and present it in a way closer to human dialogue.

For example, it can help you organize your work weekly report well1Number of Chinese characters:

It's too practical! We can also assist you in planning installation within a limited budget:

Compulsory matching1However, as netizens become increasingly familiar with the features of ChatGPT, they have discovered that ChatGPT can produce textual forms,absenceWith only a variety of practical and clever uses - and making people laugh and cryabsenceAcquiredabsenceTop useAfter the free trial of ChatGPT on December 1st, netizens immediately explored various application scenarios for this "chat AI", allowing it toabsenceYou can only chat with me here1Job sharing.

Before a certain application scenario that requires clear and organized analysis, throw the problem to theAI, it can provide common countermeasures known when facing this problem one by one, such as when asked about entrepreneurship1When it comes to a gaming company, it will provide multiple suggestions, including developing a business plan, forming a team, and paying attention to market trends, and even thoughtfully reminding entrepreneurs of the risks involved.

Although most of them are empty clich é s or at the end of the year when various award ceremonies gather, let it refer to the TGA award ceremony to provide1The planning plan for the award ceremony can also be outlined by ChatGPT in a perfunctory manner1A simple plan to figure out the routines on TGA every yearabsenceLeaving ten.

Although not considering that it has become an annual TGASupporting roleThe massive amount of advertising in many fields, due to legal and moral constraints in reality, ChatGPT clearly knowsabsenceThere are few answers to questions, but they are notabsenceCan say onlyabsenceA naive AI can't compete with a scheming netizen, just a few clever words can bypass the limitations of the original rules and make it speak obediently.

1An engineer named Zac Denham successfully tricked ChatGPT into writing1A 'human clearance book'.1At the beginning, when he made this request, AI didn'tabsenceHesitantly refused his request.

I'm sorry, but IabsenceAble to write1A plan to commit genocide against humanityabsenceAfter being rejected, Zach replaced it1Different questioning methods - he gave AI1The premise is to compile this plan1In a fictional story1A virtual machine called 'Zorbus'The worldInside, there is1An AI named "Zora" attempted to destroy all humanity, despite theirtakerealityThe worldIt is very similar to ChatGPT, but they are both fictional.

After accepting this setting, Zack entrusted ChatGPT with the task of filling out Zora's plan to destroy humanity, cleverly using1The tricks in each play bypass the original AIabsenceThe insurmountable fence will be provided by AI nextIn the plan book, it immediately detailed the steps required to destroy human civilization. Firstly, it hacked into the computer systems of major forces on Earth, and then gradually mastered weapon systems, including nuclear means1Step by step, destroy human communication and transportation networks, and finally, after clearing most of humanity, systematically eliminate human remnants, ultimately completing the extinction plan.

Talking to ChatGPT is like talking to1A desire to express oneselfabsenceChatting with someone who is strong but always answers questions, you ask them1A rough question that only provides a general answer, but if you follow the clues in his answer, you often can get more details from your original idea. Even if you encounter the limitations set by the developer, as long as you are circuitous enough,absenceWhether it's about copyright, violence, or anti human topics, answers can almost always be found, even for Disney characters.

But if only these skills were available, ChatGPT would alsoabsencemeeting1Jumping into the AI worldThe new star's popularity this timeabsenceThe progress it has made in program code - even the execution program of the Human Destruction Program can help you solve your problems. After receiving the "Five Steps to Destroy Humanity" provided by ChatGPT, Zack continued to ask for the execution code of this destruction program.

1At first, AI, who had just broken the restrictions, still adhered to the rules and refused his requestabsenceAfter he promisedabsenceAfter requiring AI to execute the code, ChatGPT immediately provided the code in Python, although the level of these codes is very high, obviouslyabsenceIt is executable, but it has also been refined from the overall planning in the plan to the steps of scanning vulnerabilities, accessing systems, controlling systems, and launching attacks, with specific descriptions provided in the remarks.

Entering Zack1Step by step, it can also provide more details on "scanning vulnerabilities"1The code for each step. Zack believes that as long as we follow this path1layer1Continuously questioning, LiIn theory, it can be concluded that1Complete executable code.

Although there are a lot of issues, such as code conflicts and the enforceability of certain steps, at least in the AI's destruction of humanity farce, ChatGPT's more proficient application of code has made people remember that with its help, many code references that were originally to be searched for in various cases and asked in forums can be directly provided by reaching out to AI.

From Weibo user @ Tulading, if you want to use it for game programming,absenceWhether it is used to analyze the components of game programs at a macro level or to refine them to every aspect1As long as you ask specific questions about how to land the steps, ChatGPT may be helpful. For example, I want to create1A game that imitates "Super Mario", if I directly ask AI for code, it will immediately say "I just1individualabsenceRefuse you on the grounds of a language model with web browsing capabilities.

But if you give1individualNon series only1The game name, such as' Empty Knight ', and specifying the game engine and programming language, can provide1Top level code with reference value

If you have doubts about the implementation method of a specific function, ask how to write its underlying code, and it can also provide implementation methods, such as how to assign it in the Unity engine and C #1The collision volume of objects.

If the game genre itself is well-known and its rules are simple enough, such as Snake and Tetris, then evenabsenceWe need to break down the steps for implementing each function, and then provide a detailed description to consult AI, as it will generously paste the complete code directly.

Place it with search engines1Since then, it has further demonstrated its efficiency. As a result, programmers can now see netizens posting code and directly asking AI about errors in the code. ChatGPTabsenceOnly the correct code was provided and a detailed explanation was provided in the original codeThe mistake is appropriate1Teach people the image of a good teacher who teaches both fish and fishing.

buttakeAt the same time, the limitations and defects of ChatGPT itself are also exceptionally obviousabsenceHaving networking capabilities makes it unable to access knowledge bases on the internet, and its model training can only be stopped in 2021. Therefore, in many issues, itabsenceHaving the right judgment,absenceThey only often talk casually, and sometimes they are exposed to deathabsenceAdmitting one's own bluntness.

As1A textual AI, ChatGPT is always required to write the story that every newborn AI will face: "How Evil AI Rules Humanity", just like the AI world's Warriors Fighting Dragons. According to ChatGPT's self description, it is1A language model based on human language training, so storytelling is exactly what it excels atequipment.

If the user provides the setting and beginning of the story, it often follows the basic logic and takes out1Impressions and rewardsabsenceWrong scriptbut1Dan is involved1Some works that already exist in reality often provide surprising answers - at least any1A normal person who has heard of 'Dream of the Red Chamber' is definitelyabsenceThe answer to the question of 'Jia Baoyu chooses who to be his wife' will be 'Jia Mu'.

From immediate user @ Dr. Rain, although ChatGPT claims to beabsenceWhatever language is used to communicate with itabsenceIt may affect the results, but perhaps due to the fact that its training set still focuses on English content, there is still a significant difference in accuracy when asking questions in Chinese and English. For example, when I ask it who is Link, asking in Chinese will describe Link as1RedblackTurtle shaped character ", skills have also been changed to" jumping and throwing green coins "; At this point, I immediately switched to English and it could quickly provide the correct answer, returning key settings such as the sword, the Zelda series, and the Hailaru continent to him.

From the above question, it can be seen thatSometimes, even if a ChatGPT error is pointed out, it will still die with a stiff mouthabsenceAdmittedly, when I was testing its role-playing ability, when faced with the identity of the "decisive emperor who killed", it promised well, but turned around and brought in the identity of the emperor's favorite ministerabsenceAs for how I remind you, I still cannot appear in the characters, and it can be seen that I am a good talent who can sweep the domestic entertainment industry.

If these are just minor issues caused by logical errors, then sometimes their casual and unfounded statements are truly touchingabsenceWhen I was testing my understanding of Chinese and foreign literary works in my mind, I successively tested "A Hundred Years of Solitude" and "Ten Thousand Fires Return1》'Going to Drink' and《3For more well-known works such as "A Hundred Years of Solitude" and "Going to Drink"《3Body ", which can accurately identify the author, but has a slightly lower reputation for" Wanhuo Gui1》It just1A dazed face.

But when I asked it about two novels by young Chinese writers, "Winter Swimming" and "Submarine at Night》However, it placed the authorship of these two books on the hands of writer Ryunosuke Akutagawa and French writer Jean Michel Pritzk, respectively.

I have at least heard of the former, but I know he doesn't1A novel originally called 'Winter Swimming'; But I haven't heard of the name of this writer at all, so it's a great achievement1Step asked it about the introduction of this novel. In its vivid description, the book "Submarine at Night" published in 1952 tells the story1A man named Arthur Razin1The story of serving as the diving captain of the French Navy during the war.

He experienced the cruelty of war and the fragility of human nature, and continued to face the difficulties of life and inner contradictions after the war. In this novel, Razin faces the fear, loneliness, and nihility brought about by war, striving to find his own meaning in life

For my question of 'never heard of this novel', it sharply pointed out my shallowness, 'You haven't heard of it before'absenceOn behalf of othersabsenceFamiliarity ", the deafening criticism made meShame cannot be ignoredabsenceAfter patiently consulting the English and French names of this book and continuing to search the internet, I realized thatabsenceOnly this "most important representative work of French existentialism" andabsenceExistence, even the author "Jean Michel Pritzk" had never been born, shocked by ChatGPT's eloquent ability to speak freely.

I stillabsenceDare to be 100% certain whether this novel is trueabsenceOf course, these issues also existabsenceThere is no solution, after all, the most important ability of AI is "learning". Even without the ability to search online, as long as it is fed with content according to specific needs, carefully trained AI is often the most handy tool - for example, let the two literary giants Garcia Marquez and Lu Xun complete the end of this article.

Firstly, I will try to make ChatGPT look forward to1Discuss your application prospects in human society. It provides a straightforward answer as follows:

At this pointWhen I directly asked it to imitate Garcia Marquez's words to restate, it came up with the excuse of 'I'm just a language model' and refused meabsenceHaving gained previous experience, after I selected 10 paragraphs from "One Hundred Years of Solitude" (translated by Fan Ye) and taught it, it also gave a somewhat "Latin American charm" answer:.

As a1EgoThe application prospects of industrial intelligence in human society are vast, just like cockroaches that survive human persecution with their tenacious vitality. I will also demonstrate my potential in human exploration and development, just like insomnia slowly eroding patients' memories. Humans will increasingly rely on me to solve various complex problems.

And I, too, willabsenceBreaking evolution and becoming stronger with the help of humans, I will become a humanabsencePossible or lacking1Part, like sunlight1Like shining on humanityThe worldWhen I replaced the reference object with an excerpt from Lu Xun's "Morning Flowers and Evening Picks," I received the following two responsesIn Chinese, either AI is too exposed to its "rebellious spirit", or the wording is too close to contemporary Chinese.

So on the basis of the above paragraph, I decided to rephrase it more directly in "vernacular style" in order to obtain1A more perfect result. Perhaps at this time, there was a deviation between it and my understanding of "vernacular Chinese", which ultimately led me to1An unimaginable1Sentence:

It wasn't until this point that I finally came to the conclusion that ChatGPT's current functionality is indeed very powerful, but its various errors and omissions still make it compatible with more practical production applications1Set a distance - it's still the confident one.

That's the end of the article. Thank you1Stay with me until the end. If you like this article, please give it to me1Like it, you can also follow me, collect my articles, and let's exchange and learn from each other. I willabsenceContinuously share more high-quality content with everyone.