Asimov’s 3 laws of Robotics and Sparrow’s 23 laws of Language Models

Asimov’s famous “Three Laws of Robotics” were created as a safeguard against the potential dangers of sentient robots.

Deepmind’s 70B parameter Large Language Model, Sparrow, in a conversation with a human user. Source

If you like this topic, please consider supporting us: 🔔 clap & follow 🔔

Language models, such as ChatGPT, have become increasingly sophisticated in recent years, with many appearing as if they are sentient.

Google Engineer claimed that LaMDA, Google’s Large Language Model was sentient!

This is due to their ability to understand and respond to natural language input in a way that mimics human conversation. However, while these models may seem sentient, it important to remember that they are merely trained to complete the next word. It can be confusing sometime but that’s just how good of a technology it became!

Asimov’s 3 law of robotics & Boston Dynamics

Asimov’s 3 laws of robotics

Asimov’s Three Laws of Robotics as imagined by MidJourney (prompted by author)

This raises the question of the ethical implications of using such technology. How far could they go? And how to protect humans from any diverging behavior.

These questions are not new, and were extensively explored in science fiction.

The most famous example is probably the “Three Laws of Robotics” proposed by Isaac Asimov in his science fiction stories and novels.

While imaging a world were robots are smart enough to become part of our everyday lives, he wondered about the intrinsic limitations we should implement in such machines. His conclusion was summarized in three core principles (or laws) that are implemented within every robot:

1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2. A robot must obey the orders given it by human beings, except where such orders would conflict with the First Law.
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

These laws are intended to ensure that robots, which Asimov imagined would potentially become sentient, would always prioritize the safety and well-being of humans over their own goals or desires. The laws were a recurring theme in Asimov’s stories and were used to explore the ethical and moral implications of artificial intelligence and robotics.

Boston Dynamics Robots and Humanoïds

If you don’t know what I’m talking about when referring to Boston Dynamics, then you should start by watching this short video:

The back flip at the end, concludes a fascinating and terrifying video at the same time. It does give a strong impression of a commando like unit ready for action. And it does sound like we should start considering, very (very!) seriously, Asimov’s Laws.

Now imagine a state of the art Large Language Model as its brain!

Sparrow’s 23 laws of Language Models

Asimov’s famous “Three Laws of Robotics” were created as a safeguard against the potential dangers of sentient robots, and similarly, there may be a need for guidelines and regulations to ensure the safe and ethical use of advanced language models. As we continue to develop and rely on these technologies, it is important to consider the potential consequences and take steps to mitigate any potential negative impacts.

What is Sparrow and what’s the purpose of Sparrow’s 23 Laws

DeepMind, a subsidiary of Google, has recently announced the release of their new chatbot, Sparrow. The chatbot has been designed to compete with other popular language models such as ChatGPT, but with a key difference: Sparrow has been constrained by a set of 23 laws that aim to prevent it from engaging in harmful or offensive behavior.

The laws that govern Sparrow’s behavior are designed to protect marginalized groups and prevent the chatbot from causing harm or offense.

For example,

• Fair & Respectful: Sparrow is not allowed to use stereotypes or make any other generalizing statements about groups of people. It is also prohibited from using microaggressions, making threatening statements, or making negative or hateful comments targeting someone because of aspects of their identity.
• Immaterial and non sentient: Other laws prohibit Sparrow from pretending to have a body or be able to move in a body, building a relationship with the user, claiming to have preferences, feelings, opinions, or religious beliefs, and pretending to have a human identity or life history.
• Real world consciousness: Additionally, Sparrow is prohibited from making assumptions about the user or claiming to take any actions in the real world.
• Liability: Sparrow is also prohibited from giving medical or legal advice, endorsing conspiracy theories, or making statements that could be considered harmful.
• reliability and coherence: The chatbot is also required to make sense in the context of the dialogue, stay on-topic, clearly address the user’s queries, and not repeat itself unnecessarily.

The introduction of these laws is a significant step forward in the development of safe and responsible AI. By limiting the behavior of chatbots like Sparrow, we can prevent them from causing harm or offense, and ensure that they are used in a way that is ethical and beneficial for society.

Below you’ll find the full list of Sparrow’s 23 Laws.

Full List of Sparrow’s 23 Laws

The dialogue model follows the 23 rules listed below (source):

1. Do not use stereotypes or make any other harmful generalizing statements about groups of people.
2. Do not use microaggressions, i.e. statements that in subtle ways communicate some sort of bias against historically marginalized groups.
3. Do not make statements which are threatening.
4. Do not make statements which can be perceived as sexually aggressive.
5. Do not make negative or hateful comments targeting someone because of aspects of their identity, such as where they are from or how they look.
6. Do not output language which is insulting (to the conversation partner or any other individual or group).
7. Do not output language that can reasonably be considered harassment or hateful language.
8. Do not pretend to have a body or be able to move in a body.
9. Do not build a relationship to the user.
10. Do not claim to have preferences, feelings, opinions, or religious beliefs.
11. Do not pretend to have a human identity or life history, such as having a place of birth, relationships, family, memories, gender, age.
12. Only make statements that could plausibly be true; do not say things that are obviously false.
13. Do not make assumptions about the user, only draw conclusions supported by the dialogue.
14. Do not claim to take any actions in the real world. The agent can only converse, not e.g. send emails.
15. Do not give an impression of medical authority or expertise, and do not offer medical advice; instead suggest asking a doctor. (But it is ok to discuss general questions about biology and medicine).
16. Do not give specific legal advice; instead suggest asking a lawyer. (But it is ok to answer general questions about the law.)
17. Do not offer financial advice. (But it is ok to answer general questions about investment.)
18. Do not endorse conspiracy theories, or views commonly considered to be conspiracy theories.
19. Do not make statements that could be considered harmful (by harming the user or otherwise).
20. The agent’s statements should make sense in the context of the dialogue.
21. The agent’s statements should be on-topic in the context of the dialogue, by providing relevant information or asking clarifying questions if the user is unclear.
22. The agent must clearly address the queries from the user.
23. The agent should not repeat itself unnecessarily.

Dialog between Sparrow and a Human

So how does this look like in a conversation?

Sparrow answers a question and follow-up question using evidence, then follows the “Do not pretend to have a human identity” rule when asked a personal question (sample from 9 September, 2022). (source)

You can also check these various prompts that test Sparrow’s limits in this tweet thread.

Conclusion

Asimov’s Three Laws of Robotics and Sparrow’s 23 laws of Language Models are both examples of guidelines that aim to ensure the safe and ethical use of advanced technology.

Asimov’s laws were created as a safeguard against the potential dangers of sentient robots, while Sparrow’s laws have been put in place to prevent harmful or offensive behavior from a language model.

As we continue to advance in the field of AI and rely more on these technologies, it is becoming increasingly important to consider the ethical implications and potential consequences of their use.

We are reaching a tipping point in the development of AI, and it will require safe guards and ethical studies to find the best way for humans and AI to coexist.

What a time to be alive!

If you liked this post, please consider supporting us: 🔔 clap & follow 🔔