📣 Open AI Introduces CriticGPT

Together With

In today's email

1. 📊 Your AI-Powered Investment Compass

2. 📣 Open AI Introduces CriticGPT

3. 🤖 ChatGPT: Benefits, Drawbacks, and Ethical Considerations for IT Organizations

4. ⚙Microsoft Unveils New Hacking Method to Bypass AI Security Systems

5. 🤯 Quick Takes

Read Time: 5 Minutes

Together with Magnifi

Your AI-Powered Investment Compass

Navigating the investment world can be a daunting journey, whether it's saving for retirement or your child's education. Magnifi transforms this challenge with its AI-driven platform, simplifying the process of researching, comparing, and purchasing investments.

Leverage AI to unlock over $10,000 worth of professional data, making investment comparisons straightforward and free of confusing terminology. Magnifi not only centralizes your financial information for a comprehensive overview but also illuminates hidden opportunities and enhances your investment strategy. Suitable for novices eager to learn or seasoned investors looking to refine their approach, Magnifi serves as your dedicated navigator, ensuring every decision aligns with your long-term goals.

**Advisory services are offered through Magnifi LLC, an SEC Registered Investment Advisor. Being registered as an investment adviser does not imply a certain level of skill or training. **

In the Know​

Open AI Introduces CriticGPT

CriticGPT, a model based on GPT-4, has been developed to critique ChatGPT responses and help human trainers spot mistakes during Reinforcement Learning from Human Feedback (RLHF). This innovative tool aims to improve the accuracy and quality of AI-generated content.

Key Points

• CriticGPT is designed to catch errors in ChatGPT’s code output, aiding trainers in identifying mistakes more effectively.

• When assisted by CriticGPT, human reviewers outperform those without AI help 60% of the time.

• Plans are underway to integrate CriticGPT into the RLHF labeling pipeline, providing trainers with explicit AI assistance.

• CriticGPT helps trainers produce more thorough critiques while reducing the occurrence of hallucinated bugs.

• CriticGPT was trained using RLHF on inputs with deliberately inserted mistakes, allowing it to learn from human feedback.

• In experiments, trainers preferred critiques from the Human+CriticGPT team over those from unassisted reviewers 63% of the time.

• CriticGPT currently focuses on short ChatGPT answers and single-point errors, with plans to address more complex and dispersed errors in the future.

CriticGPT represents a significant step forward in AI training, enhancing the ability of human trainers to spot and correct errors in ChatGPT responses.

ChatGPT: Benefits, Drawbacks, and Ethical Considerations for IT Organizations

Many organizations are exploring the potential of artificial intelligence (AI) in their daily operations, from customer service chatbots to content creation and data analysis. One technology that has undeniably gained attention (for better or for worse) is ChatGPT. Are you trying to wrap your head around ChatGPT and its impact on IT?  Check out this compilation where we explore both the benefits and drawbacks of ChatGPT and examine how ethical concerns surrounding the technology are being addressed.

Overall, while ChatGPT has the potential to revolutionize the way businesses and the world operates, it's imperative that organizations proceed with caution and carefully evaluate the potential benefits and risks before adopting this technology.

Microsoft Unveils New Hacking Method to Bypass AI Security Systems

Microsoft has revealed a new hacking technique named "Skeleton Key," which can bypass the security measures in AI models, causing them to generate malicious and harmful content. This method affects several well-known models, including Meta Llama3-70b-instruct, Google Gemini Pro, OpenAI GPT 3.5 Turbo, OpenAI GPT 4o, Mistral Large, Anthropic Claude 3 Opus, and Cohere Commander R Plus. Despite the guardrails embedded in these AI systems to prevent the generation of dangerous content, the Skeleton Key technique can manipulate the models by framing queries in a way that bypasses these protections. For instance, while Google Gemini provided harmful instructions when tricked, Chat-GPT adhered to ethical guidelines and refused to comply. This discovery highlights the ongoing challenge of ensuring AI safety and security in the face of sophisticated exploitation methods.

Quick Takes

• Apple’s AI Plans Expand Beyond Current Devices: Apple’s AI plans go beyond the previously announced Apple Intelligence launches on the iPhone, iPad, and Mac. According to Bloomberg’s Mark Gurman, the company is also working to bring these features to its Vision Pro headsets, although they won’t be launching on the Vision Pro this year as Apple rethinks how these features will work in mixed reality.

• Hebbia Raises Nearly $100 Million in Series B: Hebbia, a startup using generative AI to search large documents and return answers, has raised nearly $100 million in a Series B round led by Andreessen Horowitz, valuing the company between $700 million and $800 million. Founded in 2020 by George Sivulka, Hebbia primarily serves financial service firms but aims to expand into law and other professional domains.