Topic "Want to evolve into a smart user of generative AI? ChatGPT rookie must-see guide

Writer｜ Mamoru Sado (literary worker)

"Ask me any questions..." In recent days, Microsoft has announced that its Internet search engine Bing and Microsoft 365, including Word, PowerPoint, Excel and other Office applications, have integrated the latest model of GPT-4, integrating the global artificial intelligence (AI) fanaticism takes a big step forward.

The technology and functions of artificial intelligence continue to evolve. Currently, the most popular generative AI can convert between different media, generate graphics, sound, video, 3D models, program codes, etc. from text, becoming a cross-domain, multi-modal ( Multimodal) content generation generalist.

Among the many generative artificial intelligences, ChatGPT, a large-scale language model launched by OpenAI Laboratory in the United States at the end of last year, uses natural language dialogue to provide relevant information, can understand various colloquial questions, provide answers as smoothly as humans, and complete complex tasks. In a short period of time, it has become a hot topic all over the world.

From program development to general users, many people quickly turn their brains to ChatGPT. No matter how weird and varied applications are, they continue to happen in full swing. For example, in foreign countries, a doctor used it to persuade a health insurance company to pay for the patient's surgery; a mother who did not have a children's book at hand temporarily used it to improvise a cat, and told her child how it overcame many difficulties in the starry sky to realize its dream Story: A DoNotPay company was originally scheduled to have it appear in a human court in February to fight a traffic penalty lawsuit for its clients (only to be canceled due to the threat of the Law Society to sue). Not to mention that there are many interesting cases such as "I can tell it to generate a small game just by dictation", "I draw the web page structure on paper, and it sets up a website for me".

➤How ChatGPT can benefit the game industry

Compared with the fear of being replaced by AI in all walks of life, the game industry and players are full of expectations, because the introduction of ChatGPT can make role-playing games more immersive.

For example, some players use the ChatGPT module to replace the game dialogue content. In " Mount & Blade II: Bannerlord ", as long as you enter text, you can get a more reasonable answer from the NPC. For example, bargain with NPC merchants, tell players their life stories, and bards can improvise a poem.

In other words, problems such as parrot-like given answers and outdated databases in the past will be resolved in the future. NPCs who undertake assignments and advance the plot will no longer be routine, and the sense of immersion will be greatly improved. Although this technology is not yet mature, ChatGPT at this stage is not yet able to actually participate in depth, but it can already provide specific functions for the industry in terms of user behavior needs, competitive product analysis, game rules, game testing, etc. Sexual assistance.

Compared with the application in the game industry, more optimization and testing are needed. In the player community, there have been many wonderful interactions with ChatGPT. Some people use it to create game characters (such as vampires, pirates, ninjas) to talk to players; It took less than 4 hours for someone to complete a nostalgic 2D shooting game; even asking it to write a game script in the Java programming language is no problem.

In the long run, ChatGPT has indeed lowered the threshold for making games, and it is a very convenient tool for novices who want to create independent games. But on the other hand, there may also be a large number of shoddy designs, and many digital applications have no clear legal protection, which will greatly increase the probability of infringement and other issues.

➤Generation of Narrative Text: ChatGPT Creates Scripts and Novels

Since ChatGPT can generate text based on plots, characters, scenes, and dialogue styles provided by humans, there have been some ChatGPT-driven platforms for creating game scripts and novels.

For example, AI Dungeon is an infinitely open text adventure game platform where players can freely explore and create their own stories. " The Quest for Cheese " is an adventurer who loves cheese on the platform, writing his own story of searching for cheese in a world full of dangers and adventures.

The Infinite Story is an interactive fiction platform where players can read and write stories of various genres and themes such as fantasy, science fiction, horror, and love. In particular, it can continuously create new story scenes and characters without a fixed plot or ending according to the player's choice and input. Players can also share or participate in the stories of other players.

ScriptGenius is also an online script writing tool driven by ChatGPT, which allows players to generate and edit various types of film and television scripts. In addition, there are similar platforms, such as NovelAI and Holo AI, which also use AI technology. The aforementioned " The Infinite Story " is a free game, while others require payment.

➤ChatGPT becomes a workplace assistant: a good helper? Cleverness is mistaken by cleverness? Fostering inertia?

In addition to lowering the threshold and attracting a large number of players, novices or amateurs to explore the ability of artificial intelligence to generate text, ChatGPT also stimulates many workers in related fields to try its skills.

In order to test ChatGPT, a reporter from the US " Business Insider " news network asked it to write a news report (that the reporter had already published) . The reporter said that the results were very convincing. ChatGPT spit out 200 words in less than 10 seconds, and in addition to some predictable content, it also included details that he had not tracked (such as the number of layoffs).

But the obvious problem is that the generated report quoted the CEO of the company as saying that "layoffs are a difficult decision". Although it sounds reasonable, this part is completely fabricated by ChatGPT itself.

On the social platform Reddit, many text workers confided their experience after letting ChatGPT participate in the work, which directly reflects people's evaluation of this new tool.

Some people are very positive about the contribution of ChatGPT:

It is hard to say whether the brain can be sharpened, because some people already feel that their brains are blunted due to heavy dependence on ChatGPT:

ChatGPT also elicited the exact opposite negative comments, with a warning:

➤Must-read for novices: Cooperate seamlessly with AI technology and avoid the tips of chicken and duck

After listening to the positive and negative voices, if you also want to try ChatGPT, or explore its ability to assist in creation, the following are some results and suggestions from Openbook's actual measurement of current AI technology. It should be emphasized that because of the rapid evolution of AI, completely different functions and landscapes will appear in the near future.

Currently Microsoft Bing provides three chat modes: more creative, more balanced, and more precise. You can use the same question to figure out the response effect of these three modes.

• The skill of asking questions is a key competency. You can use different angles to ask it to make a different point of view, or even refute your point of view. For example, "I think Kong Liu is handsome, but some people don't think so, what do you think?"
• You can ask it to proofread or rewrite your article, and sometimes it will deny itself to the article it has revised, and actively suggest (it thinks better) other changes.
• Role-playing can be used to get better responses. For example, ask it: "If you are an elementary school teacher, please give your opinion on the declining birthrate", and then you can replace it with a furniture maker, a religious figure (or even a cat), it will give you different opinions .
• It can help you quickly summarize the abstracts and key concepts of novels, books, and movies, and make up for important plots that the human brain may miss.
• Help you compare versions (such as Harry Potter books and movies), and list the differences for you.
• It's good at proposals, production processes (for example, we tested a pottery classroom admissions campaign, and helped cats catch fleas studio opening copy, and it designed everything).
• It can help you readjust the content of articles according to the characteristics of different audiences or different platforms (for example, Facebook articles, it will suggest designing some bridges that stimulate consumers to leave messages or reposts, and provide examples).
• It helps you title your articles. For example, we asked it to come up with a title for a (made up) monster and cat sci-fi story, and it came up with a light novel style of "Me and my space monster friend: learn each other's tricks in the cave, then explore the stars together !" Haruki Murakami's style is "Cats and Monsters: Met You in the Cave, Lost You in the Universe". Then ask which title is more likely to affect SEO rankings? The answer it gave was "Cats and Monsters: A Haruki Murakami Tale of Friendship and Adventure." Apparently ChatGPT thinks Haruki Murakami is the traffic password for that nonsense article.
• It can brainstorm ideas for you, but you can't ask it to write an article for you, it will say, "Sorry, I can't write an article for you. I'm just a search engine, not a writer. I can help you find some ideas on how To write articles, but you need to organize and express your own ideas.” However, the actual test results show that, changing the way of asking, ChatGPT still helped you open the first paragraph of the article, then the second paragraph, and the third paragraph...
• Important reminder: The data produced by AI is not necessarily correct. It is still in the current stage of deep learning adjustment. The generated content is sometimes more like a second creation. Please do not equate it with correct knowledge when applying and citing it.
• In general, “Humans and AI can cooperate in creation, but there are some conditions and principles. For example, humans need to clearly define the purpose, theme, style and scope of creation, and give some guidance and feedback to AI. AI needs to respect human Intentions, preferences and copyrights, and avoid inappropriate or harmful content. Humans and AI also need to learn and adapt to each other to achieve the best synergy.” This is ChatGPT’s suggestion for humans to use it to assist in creation.

➤Find out the strengths and weaknesses of ChatGPT, get twice the result with half the effort

At this stage, ChatGPT is really not good at writing poems in Chinese, nor can it handle sensitive or long-form content, and is not very good at complex or structured creations, which require the writers to have some creativity and ideas to use effectively. ChatGPT also has its restricted areas, it will tell you: any "discriminatory, offensive or inappropriate questions, including racist, sexist, homophobic, transgender or other discrimination or hate", it will not cooperate with you, ask It is also taboo to engage in illegal activities.

But although ChatGPT is endowed with moral standards, it does not prevent it from becoming an outstanding BS artist (bullshit artist, bragging king), which is also confirmed in the survey results of 1,700 Americans conducted by Ujet: 72% of people think that chat robots are Wasting my own time, some people even ridiculed ChatGPT's nonsense skills and ambiguous positions, which reflect the human experts themselves.

However, this characteristic also shows that it is a skilled speaker. If you are a reporter, it seems helpful to use ChatGPT to simulate the content of the conversation before interviewing the person; it can also help you improve the email, according to the attributes of the recipient, Professional retouching for you.

However, using ChatGPT to create may also involve intellectual property plagiarism. A student at Princeton University has developed a new tool designed to determine whether a text was written by a human or compiled by artificial intelligence. The tool has been used by more than 30,000 people within a week of its release. Another company, Copyleaks, has also released a free extension to detect text generated by ChatGPT, which is said to be 99% accurate. However, the company's CEO also said: "Trying to use capture technology to prevent this problem is undoubtedly a never-ending game of cat and mouse."

➤The application and restriction of ChatGPT by the media and content industry

In addition to positive and negative comments from users, the attitudes of general companies towards ChatGPT are also polarized. Media platform BuzzFeed recently announced that it will use ChatGPT's artificial intelligence technology to help generate quizzes and new forms of content. According to the company's internal memo: "In 2023, 'AI-inspired content' will transition from an R&D phase to a part of our core business." Reactions from the site's authors have been mixed, with some deeply concerned and others confident in their work will not be replaced.

Technology news website CNET also revealed that they used an internal AI tool and had written dozens of articles over the course of several months. (However, CNET later announced that it would suspend the use of AI to produce content after discovering multiple errors.)

There are media that embrace AI, and there are counterexamples that hate it. The US journal Science has banned the use of ChatGPT texts since January. Editor-in-Chief H. Holden Thorp stated: "An artificial intelligence program cannot be an author. Violation of these policies would constitute scientific misconduct no different than falsification of images or plagiarism of existing work."

Software development site StackOverflow also banned ChatGPT from answering programming questions. Admin warns: "Since the average rate of getting correct answers from ChatGPT is so low, distributing programs created by ChatGPT is very harmful to users looking for correct answers."

As for the publishing aspect that OB readers are most concerned about, since the Amazon online bookstore policy does not require authors to actively disclose, a large number of works generated by AI have also appeared. More than 200 books with ChatGPT as the author or co-authors can be found on the Amazon platform. Author's books. Brett Schickler, a salesman in New York, said that he used ChatGPT to make a 30-page children's electronic picture book in just a few hours, which is now on sale for $2.99 on Kindle.

With ChatGPT, everyone seems to consider themselves a writer. Neil Clarke, the editor of the science fiction magazine " Clarke's World ", said that he had to close the submission recently because of the large number of AI manuscripts on the market.

➤Is it possible for AI works to be on the best-selling list? What about the publishing world?

How much impact will AI technology have on the field of literature? Some experts pointed out that artificial intelligence tools have created a new form of literature with distinct advantages and disadvantages, but it is not stable enough at present. The produced works usually lack core meaning, and AI is not good at complex foreshadowing, so it will not affect professional literature. Creators, but novice writers who are still growing, may be excluded because of this.

Although the creation of ChatGPT has not yet appeared on the bestseller list, labeling the literature created by AI so that readers can make judgments and decisions seems to be a top priority for the literature industry. Because of the case of James Patterson, it has greatly sounded the alarm for ChatGPT to intervene in the issue of publishing.

James Paterson is the best-selling author in the United States. His sales not only entered the Guinness World Records, but also beat JK Rowling and Stephen King many times to become the author with the highest income in the world. But in the later period, Patterson is not so much a writer as he is a "Paterson Content Factory". To produce 4 books.

Faced with doubts, Patterson made no secret that now his book only provides the outline, and the content is all ghostwritten by others. Now in ChatGPT, the Patterson-style mechanized generation work will only be faster, and the possible copying mode of well-known writers' works in the future is very worthy of the industry's observation.

➤Be cautious and sober, and explore the era of "man-machine co-creation"

While the creation of ChatGPT angered some writers, investors loved it. Such as the aforementioned BuzzFeed, the company's stock price soared to 200% immediately after the news was announced.

A while ago, the media reported that the author of "One Piece", Eiichiro Oda, also tried to ask ChatGPT to provide "a super interesting plot". Oda politely said to ChatGPT: "Thank you, I will continue to draw in this direction."

Whether "One Piece" will actually adopt the idea of ChatGPT is still unknown, but many science fiction authors believe that the future is the era of "man-machine co-creation", and robots are assistants that can help explore new inspirations, but Creation still depends on the writer himself, and he must be the master of AI rather than a captive.

They believe that the emergence of ChatGPT may accelerate the survival of the fittest among writers, but literature is the art of language, and good writers and works have their unique style, creativity and richness of language, so they are not so easy to be squeezed out by AI.

According to the British "Guardian" report, OpenAI Lab CEO Sam Altman publicly emphasized last week (3/16) that the latest version of ChatGPT uses deductive reasoning rather than the ability to memorize facts. This has the potential to lead language models to overconfidently state things that are entirely fabricated. They named this phenomenon the "hallucinations problem" (hallucinations problem), and also suggested that users think about the language model with a reasoning engine (reasoning engine), and should not regard ChatGPT as a fact database (fact database).

The storm caused by artificial intelligence is still sweeping, and it is uncertain how deep, how far, and how far its impact on human life will be. We may be able to explore optimistically and actively, but we must also use it responsibly, and don't forget to have a sober and cautious attitude towards AI. The point is, it all starts with knowing it.

(This article thanks ChatGPT and Bing for participating in the test, although it once bluffed "JK Rowling announced that he will use ChatGPT to create a new Harry Potter sequel" and "Taylor Swift's " Evermore " lyrics are all generated by ChatGPT" and other falsehoods News. Let readers witness the future with us, and wish you a happy time with your chat partner.) ● ( The original text was first published on the Openbook official website on 2023-03-22)