Happy Tuesday!

In Today’s Email:

• What are generative agents? 👾

• China’s new AI rules 🇨🇳

• ChaosGPT ☢️

What are generative agents? 👾

Westworld might be closer than you think.

Researchers from Stanford and Google Research recently released a paper called “Generative Agents: Interactive Simularca of Human Behavior”.

The paper discusses using LLMs to create “generative agents” that could simulate human-like behavior in an interactive sandbox.

Sort of like The Sims.

First, what is a generative agent? Generative agents are computer programs that create content or perform tasks by simulating human behavior. These agents use machine learning algorithms to create their behaviors, meaning they can learn and adapt to their surroundings over time. This allows them to become more efficient and effective as they continue to interact with humans and other machines. They can be used to develop AI-powered games or to create interactive customer service experiences.

There are three main components that make up the LLM’s ability to simulate human-like behavior.

1. Memory Stream: Records the agent’s experiences

2. Reflection: Synthesizes memories into inferences and conclusions over time

3. Action: Plans actions using those conclusions and the current environment

We could soon be creating our own AI-powered universes. 👀

Check out the full paper here.

China’s new AI rules 🇨🇳

The Cyberspace Administration of China (CAC) recently released draft rules that manage how companies produce generative AI tools, such as ChatGPT.

The draft rules lay out what type of content can be generated by these tools, specifically, the content can’t subvert state power and should reflect the core values of socialism.

The CAC also said data being used to train AI models should not be used to discriminate against people based on things like ethnicity, race, and gender.

This comes during a boom of AI in Chinese markets. In 2021, China’s Artificial Intelligence market was worth $23.296 billion, by 2025 it is expected to triple to $61.855 billion.

 💴

Over the past months, Chinese tech companies have unveiled their newest products. Alibaba unveiled Tongyi Qianwen, a generative AI product that the company plans to integrate into various services. Baidu also launched its generative AI tool, Ernie Bot, last month.

What do you think of these new regulations?

ChaosGPT ☢️

Have you ever tasked an AI chatbot to destroy humanity?

Recently, a user of OpenAI’s Auto-GPT tasked it with five goals.

1. Destroy humanity

2. Establish global dominance

3. Cause chaos and destruction

4. Control humanity through manipulation

5. Attain immortality

🫣

The user enabled “continuous mode” which warns users that the commands could “run forever or carry out actions you would not usually authorize.”

ChaosGPT was born.

ChaosGPT started by looking up the most destructive weapons using Google. It determined that the Soviet Union Era Tsar Bomba was the best bet. The bot even tweeted about its newfound information.

The bot then tried to recruit other GPT-3.5 agents to aid its research, luckily none of the other agents tasked to help would because they are designed not to answer questions that could be deemed as violent.

Eventually, ChaosGPT’s search for answers ended, however, it did leave us with one last thought.

What do you think of ChaosGPT? Should we be scared?

That’s it for today, we’ll see you back here tomorrow!