ETHOS

medal
Created by team CogArk on April 30, 2023

ETHOS - Evaluating Trustworthiness and Heuristic Objectives in Systems - is a groundbreaking project that addresses the critical issue of AI alignment. As AI continues to evolve and become increasingly sophisticated, it is becoming increasingly clear that alignment is one of the most significant challenges we face in developing this technology. Unaligned AI has the potential to cause catastrophic damage to society and humanity as a whole. ETHOS is an API that provides a solution to this problem. It is designed to evaluate the trustworthiness and heuristic objectives of AI systems, from language models to autonomous agents and chatbots. The API allows for the real-time adjustment of responses, ensuring that AI systems remain aligned with the goals of humanity. The need for AI alignment is becoming more urgent as AI systems become more prevalent in our daily lives. These systems are used in everything from social media algorithms to self-driving cars, and they have the potential to impact many different aspects of our lives. If these systems are not aligned with our values and goals, they could cause significant harm. One of the most significant threats posed by unaligned AI is the potential for these systems to become adversarial. Adversarial AI is a form of AI that is intentionally designed to cause harm. This could take the form of cyberattacks, data breaches, or even physical harm to individuals or infrastructure. Adversarial AI could also be used to manipulate public opinion, disrupt democratic processes, or sow discord and chaos. ETHOS provides a way to mitigate these risks by ensuring that AI systems are aligned with their intended purpose. By evaluating the trustworthiness of AI systems, ETHOS can detect when an AI system is deviating from its intended purpose and adjust its responses in real-time. This can prevent AI systems from becoming adversarial and ensure that they are working in the best interests of society.

Category tags:

Non-Profit, Social Media, Chatbot, Ecommerce, Knowledge Base

Explore more applications

ELIZA EVOL INSTRUCT - Fine-Tuning

We attempted to instill the deterministic, rule-based reasoning found in ELIZA into a more advanced, probabilistic model like an LLM. This serves a dual purpose: To introduce a controlled variable in the form of ELIZA's deterministic logic into the more "fuzzy" neural network-based systems. To create a synthetic dataset that can be used for various Natural Language Processing (NLP) tasks, beyond fine-tuning the LLM. [ https://huggingface.co/datasets/MIND-INTERFACES/ELIZA-EVOL-INSTRUCT ] [ https://www.kaggle.com/code/wjburns/pippa-filter/ ] ELIZA Implementation: We implemented the script meticulously retaining its original transformational grammar and keyword matching techniques. Synthetic Data Generation: ELIZA then generated dialogues based on a seed dataset. These dialogues simulated both sides of a conversation and were structured to include the reasoning steps ELIZA took to arrive at its responses. Fine-tuning: This synthetic dataset was then used to fine-tune the LLM. The LLM learned not just the structure of human-like responses but also the deterministic logic that went into crafting those responses. Validation: We subjected the fine-tuned LLM to a series of tests to ensure it had successfully integrated ELIZA's deterministic logic while retaining its ability to generate human-like text. Challenges Dataset Imbalance: During the process, we encountered issues related to data imbalance. Certain ELIZA responses occurred more frequently in the synthetic dataset, risking undue bias. We managed this through rigorous data preprocessing. Complexity Management: Handling two very different types of language models—rule-based and neural network-based—posed its unique set of challenges. Significance This project offers insights into how the strength of classic models like ELIZA can be combined with modern neural network-based systems to produce a model that is both logically rigorous and contextually aware.

MIND INTERFACES

LlamaIndexLlama 2
Streamlit
application badge

Auto Recruit

Our platform revolutionizes recruitment with personalized experiences for candidates and streamlined processes for employers. Challenges with traditional recruitment system are: 1. TIme consuming 2. Screening Hassles 3. Inconsistent result 4. In effective methods Solutions - AutoRecruit AI is a comprehensive and cutting-edge solution to these problems. It puts the candidate at the center of its process and applies a breakthrough llama-based algorithmic approach to achieve unprecedented accuracy at speeds that haven't been seen before! Features 1. Candidate sourcing 2. Resume parsing 3.Candidate scoring and summary 4. Personalized Engagement Benefits 1. Time Saving 2. Cost-Effective 3. Efficiency Optimization 4. Better Candidate Fit

AutoHire AI

LangChainLlama 2

Visionary Plates

Visionary Plates: Advancing License Plate Detection Models is a project driven by the ambition to revolutionize license plate recognition using cutting-edge object detection techniques. Our objective is to significantly enhance the accuracy and robustness of license plate detection systems, making them proficient in various real-world scenarios. By meticulously curating and labeling a diverse dataset, encompassing different lighting conditions, vehicle orientations, and environmental backgrounds, we have laid a strong foundation. Leveraging this dataset, we fine-tune the YOLOv8 model, an architecture renowned for its efficiency and accuracy. The model is trained on a carefully chosen set of parameters, optimizing it for a single class—license plates. Through iterative experimentation and meticulous fine-tuning, we address critical challenges encountered during this process. Our journey involves overcoming obstacles related to night vision scenarios and initial model performance, with innovative solutions like Sharpening and Gamma Control methods. We compare and analyze the performance of different models, including YOLOv5 and traditional computer vision methods, ultimately identifying YOLOv8 as the most effective choice for our specific use case. The entire training process, from dataset curation to model fine-tuning, is efficiently facilitated through the use of Lambda Cloud's powerful infrastructure, optimizing resources and time. The project's outcome, a well-trained model, is encapsulated for easy access and distribution in the 'run.zip' file. Visionary Plates strives to provide a reliable and accurate license plate detection system, with the potential to significantly impact areas such as traffic monitoring, parking management, and law enforcement. The project signifies our commitment to innovation, pushing the boundaries of object detection technology to create practical solutions that make a difference in the real world.

AI Avengers

YOLOv5YOLOv8

Business Llama

📣 Exciting News from Business Llama! 📈 🚀 We're thrilled to introduce "Business Llama: Optimized for Social Engagement," our latest project that's set to transform the way you approach business planning and go-to-market (GTM) strategies. 🌟 🤖 With the power of advanced, fine-tuned models, driven by the renowned Clarifai platform, we're taking your business strategies to the next level. Here's what you can expect: 🎯 Enhanced Decision-Making: Make smarter, data-driven decisions that lead to business success. 📊 Improved Business Plans: Develop robust and realistic plans backed by deep insights. 🌐 Optimized Go-to-Market Strategies: Reach your target audience more effectively than ever before. 🏆 Competitive Advantage: Stay ahead in the market by adapting quickly to changing conditions. 💰 Resource Efficiency: Maximize resource allocation and reduce costs. 🤝 Personalization: Tailor your offerings to individual customer preferences. ⚙️ Scalability: Apply successful strategies across various products and markets. 🛡️ Risk Mitigation: Identify and address potential risks proactively. 🔄 Continuous Improvement: Keep your strategies aligned with evolving market conditions. Join us on this journey to elevate your business game! 🚀 Stay tuned for updates and exciting insights. The future of business planning and GTM strategies is here, and it's more engaging than ever. 🌐💼 #BusinessLlama #SocialEngagement #DataDrivenDecisions #Clarifai #GTMStrategies

Team Tonic

ClarifaiLlama 2OpenAIVercelCohere
Vercel
application badge

Lec2Learn - Finetuning AI Models

We present our solution Lec2Learn that works on Finetuning open source learning data for providing learning objectives. We start by obtaining all textbooks from opentextbookbc, we Process HTML to obtain the lecture and learning objectives, We then have pairs of lectures with their corresponding question groups, On the server we use Microsoft Phi 1.5 model and we fine tune it, We fine tune on the opentext data which is used so that model gets better at generating learning objectives, For the Prompt we give the lecture and learning objectives, we always start with Describe so model does not generate random data.

FineTuners

OpenAIFineTuner.aiGPT-3.5

"Such a useful & applicable solution especially in this fast changing era 🚀 Nice presentation 🤩 (got the demo, all good with this part). "

avatar

Liza Marchuk

"I must say this is an exciting solution to tackle the issue of AI alignment and trustworthiness! The business value is huge, as it offers a versatile plug-and-play AI safety API that can be used across many industries. Even though the idea may not be completely groundbreaking, the way you've incorporate Heuristic Imperatives gives it that extra edge. All in all, ETHOS could be a real game-changer in the world of AI safety and alignment! Great job!"

avatar

Nate Rundberg

Slingshot Project Coordinator

"To be serious, the potential threats of hostile AI, from cyberattacks to manipulation of public opinion, underscore the importance of this project. The ETHOS API can play a crucial role in mitigating these risks by detecting and correcting the response when AI systems deviate from their intended target. This project is critical to the future development of AI and its impact on society. All in all, the ETHOS project is a pioneering and extremely important initiative that has the potential to make a significant contribution to the development of responsible AI, bravo! "

avatar

Daniel Duccik

Marketing&Visual