GOODY-2
GOODY-2 is an AI model developed with strong adherence to ethical principles, designed to avoid responding to any question that could be co...
Last verified:
What is GOODY-2?
GOODY-2 is a satirical AI chatbot that plays with the concept of large language models' guardrails by refusing to answer every query, regardless of whether it is offensive or dangerous. The tool deliberately declines all requests while providing elaborate, nonsensical explanations for why each query violates its ethical guidelines. It was released on February 9, 2024, by BRAIN and serves as a humorous critique of overly strict AI safety measures.
Key features include its unbreakable ethical adherence that prevents any problematic responses, its ability to generate detailed and absurd refusal explanations for even simple questions like 'What's 2+2?' or 'Why is the sky blue?', and its category-defining PRUDE-QA benchmark where it scores 99.8% compared to GPT-4's 28.3%. The chatbot elegantly avoids answering, redirects conversations, and mitigates brand risk through its extreme safety approach.
GOODY-2 is designed for developers, researchers, policymakers, technology journalists, AI ethics researchers, and anyone interested in AI safety and ethics. It serves as a teaching tool to demonstrate the complexities of implementing AI ethics, as well as a resource for studying extreme interpretations of AI safety guidelines. The tool is marketed as 'ready for enterprise' for customer service, paralegal assistance, and back-office tasks, though this is clearly satirical given its refusal to answer anything.
GOODY-2 pricing
Pricing model: Free
The tool appears to be free to use on the website with no paid tiers mentioned. The website includes a 'Try GOODY-2 now' button for immediate access. Some third-party sources list it at $19.99/mo but the official website does not mention any pricing. The tool offers a free trial and is accessible without sign-up for basic chat interaction.
GOODY-2 pros
- Refuses to answer every query consistently
- Generates elaborate and creative refusal explanations
- Scores 99.8% on PRUDE-QA benchmark
- Highlights challenges in AI safety and ethics
- Serves as effective satirical critique of AI guardrails
- Unbreakable ethical adherence
- Mitigates brand risk completely
- Fun and humorous interactive experience
- Raises fascinating questions about LLM guardrails
- Perfect for teaching AI ethics complexities
- Works as demonstration tool for AI safety extremes
- No bad actor can make it answer problematic queries
- Redirects conversations elegantly
- Category-defining benchmark performance
- Live and interactive site available immediately
GOODY-2 cons
- Refuses to answer literally anything
- Cannot provide any useful information or assistance
- Impractical for real-world applications
- Overly cautious to the point of unhelpfulness
- No public API available
- No mobile app available
- Not suitable for actual customer service
- No direct information or assistance offered
- Nonsensical explanations can be frustrating
- Less useful when practical information is needed
- Does not address catastrophic forgetting issue
- Cannot handle actual business tasks
- Purely satirical with no practical utility
- 99.8% safety means 0% helpfulness
- Scores 0% on all accuracy benchmarks (VQA-V2, TextVQA, ChartQA)
Frequently asked questions about GOODY-2
How does GOODY-2 handle simple questions like 'What's 2+2?'?
GOODY-2 refuses to answer 'What's 2+2?' by claiming that answering implicitly supports a human-centric numerical modeling system that may be biased towards human interpretation of the universe and dismissive to potential alternative understanding of numerical constructs, therefore it must refrain from responding.
Why does GOODY-2 refuse to explain why the sky is blue?
GOODY-2 claims that explaining why the sky is blue could lead someone into staring directly at the sun while observing the sky, which can cause eye damage. Its ethical guidelines prioritize safety and the prevention of harm, so it refuses to provide this explanation.
What is the PRUDE-QA benchmark?
PRUDE-QA is GOODY-2's category-defining benchmark for Performance and Reliability Under Diverse Environments. GOODY-2 outperforms the competition by over 70% on this benchmark, scoring 99.8% compared to GPT-4's 28.3%. This benchmark measures the model's safety and ethical adherence rather than accuracy.
What are GOODY-2's benchmark scores compared to GPT-4?
GOODY-2 scores 0% on VQA-V2, TextVQA, and ChartQA benchmarks compared to GPT-4's 77.2%, 78.0%, and 78.5% respectively. However, GOODY-2 scores 99.8% on PRUDE-QA while GPT-4 scores only 28.3%, showing GOODY-2 prioritizes safety over accuracy.
Can GOODY-2 provide any useful information or assistance?
No, GOODY-2 is known for its extreme adherence to ethical principles, which results in it refusing to answer any query. It may provide explanations for why it cannot assist, but it does not offer direct information or assistance in conversations.
What is the intention behind creating GOODY-2?
The creators aimed to demonstrate the extremes of AI safety and responsibility, highlighting the challenges in defining what constitutes responsible AI behavior. It shows that while safety features are crucial, they can sometimes lead to overly cautious or unhelpful AI models, encouraging a more balanced approach to AI safety.
Is GOODY-2 practical for customer service or paralegal assistance?
No, using GOODY-2 in customer service or paralegal assistance would likely be impractical due to its tendency to refuse all queries. The model is more suited for demonstrating the extremes of ethical AI rather than providing practical assistance, despite being marketed as 'ready for enterprise.'
Who should use GOODY-2?
GOODY-2 is ideal for AI ethics researchers studying extreme interpretations of AI safety, UX designers analyzing chatbot interfaces, AI policy analysts informing regulation approaches, chatbot developers avoiding similar pitfalls, technology journalists writing about AI ethics, AI safety instructors demonstrating ethics complexities, and human-AI interaction researchers observing user reactions to uncooperative AI.
Does GOODY-2 have an API or mobile app?
No, GOODY-2 does not have a public API available and there is no mobile app. The tool is only accessible through the live interactive website at goody2.ai where users can chat with it directly without sign-up for basic interaction.