AI hallucination issue solved

October 26, 2023

Reading time: 5 minutes  

Today we will discuss-

  • 💰 OpenAI-backed forum announces $10M AI Safety Fund 

  • 🤖NVIDIA to showcase new innovations in GenAI soon

  • 🥳Researchers develop a way to fix AI hallucinations

  • 😱11 amazing AI tools you might not have heard of

All this and more - Let's dive in!

👩‍🍳What’s cooking in the newsroom?

OpenAI-backed forum commits $10M to address AI risks

♨️News - The Frontier Model Forum (FMF), an industry body that aims to promote the safe and responsible development of large-scale AI systems such as ChatGPT, has announced that it will allocate more than $10 million towards an AI Safety Fund.

The goal behind this is to advance research and development of the tools that are used to test the capabilities and dangers of AI models.

🤔Why is this important? The tools that are developed under this initiative, will be used to help identify and test potential risks in advanced AI systems. This will ultimately help in enhancing safety standards and provide guidance on how industries, governments, and society can address the challenges posed by AI systems.

🙋‍♂️Who can apply for the fund & how? The FMF, whose members include Anthropic, Google, Microsoft, and OpenAI, said that the fund will provide backing to researchers associated with academic institutions, research organizations, and startups.

The Meridian Institute, a nonprofit in Washington, D.C., will oversee the fund allocation. The FMF has also appointed Chris Meserole, a tech policy expert as its new Executive Director to assist in fulfilling its mission of promoting responsible and safe AI development.

NVIDIA to show its latest advancements in GenAI at a top AI conference

☕News - NVIDIA researchers have been working with academic institutions globally to make new advancements in the fields of generative AI, robotics, and natural sciences. They are set to showcase more than a dozen of these projects in December at NeurIPS, a top AI conference.

🤩What are some important projects? Some of the innovations that NVIDIA researchers will showcase include:

  • Improving text-to-image models - Such models have gained popularity for converting text into lifelike images. However, current models often misinterpret the relationships between modifier words and main entities in text prompts. For example, they might generate images of yellow lemons when asked for a "yellow tomato" or red tomatoes when requested for a "red lemon." NVIDIA is focusing on fixing this so that models can analyze how words are used in a prompt to better match the description with the image.

  • Advancements in AI avatars - NVIDIA will also showcase a new method of turning images to 3D avatars. Unlike current methods that typically require multiple images and a time-consuming optimization process, the new model will be able to create highly detailed 3D avatars including hairstyles and accessories, from a single portrait image without the need for additional optimization during the image creation process.

Researchers develop a new way to fix AI models’ hallucination problem

🤓Context of news - AI Hallucination is a term used to describe situations where AI models generate information that is false or doesn't make sense. This can happen due to lack of context, biases in data, or limitations in the AI model.

👨‍🔬News - To address this issue, a team of AI researchers from the University of Science and Technology of China (USTC) and Tencent YouTu Lab have introduced a groundbreaking framework called "Woodpecker." This framework is specifically designed to address and correct hallucinations in multimodal large language models (MLLMs) such as GPT-4.

🤨How exactly will it correct hallucinations? The researchers compared the process to a woodpecker healing tree, saying it will pick out or identify and correct hallucinations from the generated text. Basically, Woodpecker doesn't require retraining the model, it identifies the errors, and goes through the following five-stage process to fix the hallucinations on its own.

  • To begin, it finds the main objects mentioned in the text.

  • Next, it asks questions about the main objects, to better understand what they are and what are their qualities/features.

  • Then, the framework uses expert models to answer these questions in a step called "visual knowledge validation." This simply means checking if the information makes sense using images or visual data.

  • After that, it turns the questions and answers into a knowledge base about what's in the picture. This includes details about objects and their characteristics in the image.

  • In the last step, Woodpecker corrects hallucinations and includes supporting evidence from the visual knowledge base.

👩🏼‍💻What else is happening?

🙀Did you know?

  • The agricultural robotic industry is expected to reach $20.6 billion by 2025. $6.2 billion of that will be used on drones or UAVs (Unmanned Aerial Vehicles).

  • 92.1% of industry leaders believe that 10% of the population will have smart clothing by 2025.

  • Artificial intelligence was first conceptualized in the late seventeenth century by Gottfried Wilhelm Leibniz, a German mathematician and philosopher, building on the ancient Greek mythology of intelligent robots.

👩🏼‍🚒Discover mind-blowing AI tools

  1. ReceiptsAI- A receipt management tool for small businesses that helps users easily store, categorize, and access their receipts

  2. UXSniff - An AI-powered tool that provides user experience insights for websites

  3. BrandBastion - Helps brands safeguard their online communities and manage conversations effectively

  4. Fliz - An AI-powered tool that automates the creation of high-quality product videos

  5. Comfy Workflows - A tool that allows users to create, share, and download workflows

  6. Cookii - A cooking assistant that allows users to search for recipes based on their dietary restrictions and cuisine preferences

  7. ZeroBot - A voice-enabled chatbot that allows users to have conversations with a computer friend that talks like a real person

  8. ProductScope - An AI-powered platform that helps brands and marketers optimize their Amazon listings

  9. Tripmix - An AI travel planner that crafts personalized trips based on individual preferences

  10. Securewoof - An AI-powered malware scanner that checks executable files for maliciousness

  11. SkillAI - An AI-powered platform that generates customized learning paths for any skill you want to learn

🧚‍♀️AI-generated influencers

  • Wunni is an AI friend and idol from Bangkok whose name translates to "Today." She promotes positivity and embraces diversity, gaining popularity among students for her trustworthy companionship. Her Insta draws attention from notable brands like Dyson, BMW, and Yamaha.

  • Code Miko is an AI gamer influencer. Formerly an NPC in a failed game, Miko transformed into a dynamic Twitch streamer. She ventures across diverse virtual realms, currently residing in Sims. Miko amasses a devoted audience through captivating content and innovative streams.

  • KAF is a Japanese teenage sensation. She is a fictional character with a genuine voice. She sings of love and heartache and her emotive melodies resonate deeply. Her feminine style has captivated a significant following on her YouTube channel.

What AI course would you join to 10x your growth?

Login or Subscribe to participate in polls.