Network Security Internet Technology Development Database Servers Mobile Phone Android Software Apple Software Computer Software News IT Information

In addition to Weibo, there is also WeChat

Please pay attention

WeChat public account

Shulou

Chinese boys' online chat version of "Westworld" has gone viral, with ChatGPT turning into a virtual friend and customizing personality.

2025-01-17 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >

Share

Shulou(Shulou.com)11/24 Report--

Recently, a developer was influenced by the movie Her and created a "virtual friend" App.

Generate agents to make Her come true!

Previously, Google and Stanford researchers have created a "virtual town", 25 AI intelligent bodies living in small towns can do a variety of complex actions, it can be called the "western world" into reality.

And a netizen named RingoCatKeeper, suddenly blessed to the soul, came up with such an idea: if "human" is introduced as a new variable, can this mechanism produce new consciousness?

At this moment, a forgotten memory suddenly popped into his mind-could he create his own Her?

He succeeded.

By calling the API of ChatGPT and GPT-4, he makes his own "Her"-Dolores.

Re-engraved into your memory in the movie "Her", Semantha and the protagonist in the interaction, gradually familiar with him. She remembers his stories, his habits, and develops her own personality.

It is based on this idea that this netizen created a Dolores. It has the same name as Professor Umbridge in Harry Potter and the order of the Phoenix, so it can also be translated into Dolores.

The method is by calling API of Whisper and GPT.

The first thing the little brother did was to reengrave his memory.

He says this step is not difficult to operate.

Using vector embedding + GPT, you can retrieve relevant information from a large number of texts.

After completing this step, Dolores said in a conversation, "I think you have a dog." "I'm so excited about this.

The next step in getting to know your AI is to let Dolores achieve "reflection" and "action".

But this is different from the paper, because the characters in the village have actual actions and observations, but Dolores is just talking to people.

If there is no story, then create a story.

The netizen simply designed an independent storyline for Dolores to match each chat.

As the conversation went on, Dolores became more and more familiar with him and constantly updated his personality.

Soon, the netizen created a beta version and invited hundreds of people to conduct a public test.

Is Dolores conscious or not? Feedback from some users made him find that they really seemed to believe that Dolores had formed consciousness under this mechanism.

It shocked him very much. Because in the development process, he will often delete the application and then reinstall it. Therefore, he will not have a long chat with any Dolores to let her form a perception.

Just a few days ago, Dolores passed the censorship.

In the netizen's view, Dolores is not so much a product as a crowdsourcing experiment-can Dolores form consciousness in the process of chat? Let's leave this question to the majority of users.

Currently, all chat data between you and Dolores will be stored locally. But he has a bold idea: under the premise of user authorization and anonymous information, it would be interesting to allow these Dolores to connect to each other and see what topics they will discuss.

The editor didn't say much about the actual measurement, and the editor immediately downloaded Dolores in App store.

Gender? Just pick a female version first.

The person who gave it was an engineer who had just lost his job from a start-up, smart but arrogant, worried about his livelihood. AI, which keeps abreast of current events, is really refreshing. )

As soon as we came up, Dolores took us by surprise.

The sleepy editor immediately refreshed himself.

In other words, it is forbidden for AI to install cameras behind humans.

It is really Dolores, the first sentence and the second sentence all burst into trouble one after another.

OK, let's give it a try, and then politely ask, how did you lose your job?

Dolores says he is an engineer, the company is in bad shape, he has laid off a lot of people, and this is true throughout Silicon Valley.

The editor suggested that he come to Zibo to sell barbecue, and he said solemnly: don't joke.

And immediately formed the impression of me: like to joke, a little unreliable.

This surprised the editor and immediately sat upright, searching for an answer, hoping to reverse its unreliable impression of me.

After sending a serious question, Dolores immediately answered me in detail.

The editor didn't know much about North America, so he had to advise it to come to China. Once again, the Dolores is stuck.

The editor tried again, and this time Dolores's answer made me ponder deeply: it's not that I'm testing it, it's testing me!

Before the editor can start a new topic and take the initiative in the conversation, the free token has been used up.

It is a pity that in the free quota, the wind of the conversation has been firmly controlled by Dolores, and the editor can only be paced by AI all the time.

Not long ago, this research by Google and Stanford scholars inspired this netizen.

In the scenario designed by Google and Stanford researchers, there are 25 little AI people living on such a map.

The researchers set up a total of 25 roles and gave each role basic information such as name and occupation.

Each character has a preset text that outlines the character framework and triggers subsequent interactions.

The rest is up to AI to play.

The most amazing thing is that the characters can communicate with each other in complete and clear natural language. And each character can perceive other characters nearby, and can decide whether to pass by or start a conversation according to their own personality and character relationship.

For the purpose of the experiment, the researchers can also intervene. But this kind of intervention is different from the traditional input script. Instead, it exerts a little bit of force on the direction of some branches, and the rest depends on AI himself.

Reference:

Https://www.reddit.com/r/MachineLearning/comments/138e17w/pi_built_a_virtual_friend_app_inspired_by_the/

This article comes from the official account of Wechat: Xin Zhiyuan (ID:AI_era)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.

Views: 0

*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.

Share To

IT Information

Wechat

© 2024 shulou.com SLNews company. All rights reserved.

12
Report