⚡️ LLM Transparency Report

It’s Wednesday. Stanford Human-Centered AI (HAI) released its Foundation Model Transparency Index, revealing that even leading AI model developers like OpenAI and Meta fail to provide ample information about their potential societal impact. Time to dive in.

Today’s Highlights:

• Jasper announces AI co-pilot for marketers

• IBM expands AI partnership with AWS

• OpenAI claims 99% accuracy with new Deepfake detector

DEEP DIVE

Stanford’s Foundation Model Transparency Index

Understanding the transparency of AI models developed shouldn't be like cracking a mystery code. Enter the Foundation Model Transparency Index, an effort of Stanford University's Center for Research on Foundation Models (CRFM).

Their report consists of 100 indicators that highlight transparency within three main domains: Upstream, Model, and Downstream.

Here are some key findings:

• There's a severe lack of transparency regarding resources used in model creation. Developers' scores for data, labor, and compute were quite low, at just 20%, 17%, and 17% respectively.

• Developers fare better in user data protection and basic functionality, with scores of 67% and 63% respectively. However, even these top-performing areas are far from perfect.

• Also, the debate of open vs closed AI models was put to the test. Open models (like Meta's Llama 2 and Hugging Face's BLOOMZ) surprisingly scored as well or better than the best-scoring closed model.

As the need for AI and foundation models grows, so does the need for accountability and transparency—this index is an effort in the right direction.

Keep reading here.

PUNCHLINES

Squared up: Square's new AI system can now "cook up" a restaurant menu or website in minutes.

New Recruits: Amazon’s new AI-powered robots, Sequoia, reportedly reduce fulfillment time by 25%.

Future of Work: Amazon and MIT are partnering to study how robots impact jobs and the future of work.

Aye Aye, Captain: US official reveals North Korea is experimenting with AI in cyber-warfare.

Sugar Screening: Type 2 Diabetes detection is now possible by simply analyzing your voice.

TLDR

IBM expands AI partnership with AWS: IBM Consulting will extend its cooperation with Amazon Web Services to enhance its enterprise solutions with generative AI. The upgrade targets contact centers, cloud value chains, and supply chain processes. IBM Consulting also aims to integrate AWS generative AI services into its Cloud Accelerator and train 10,000 consultants in their best practices by 2024.

OpenAI adds web search to ChatGPT, launches DALL-E 3 in beta: OpenAI officially introduces internet-browsing feature to AI chatbot ChatGPT, powered by Bing. The update follows a browsing hiatus and is available for Plus and Enterprise subscribers. Alongside, text-to-image generator DALL-E 3 enters beta, enabling seamless integration with ChatGPT to assist users in obtaining accurate images based on text queries.

Jasper Unveils AI Marketing Copilot: Marketing platform Jasper announces its new "AI copilot" that includes optimized performance analytics, an intelligence hub, and faster campaign tools. Promising personalization and performance optimization, the tool aims to uplift Jasper's marketing strategy execution. The tool will enter beta testing in November, with more capabilities expected in Q1 2024.

OpenAI claims 99% accuracy with new Deepfake detector: OpenAI introduces a deepfake image detection tool with a reported precision of 99%. Unveiled at the WSJ’s Tech Live conference, the tool is an upgrade from a previous model that was shut down due to high error rates. The solution tackles the challenge of distinguishing AI-generated images from real ones.

TRENDING TOOLS

🔀 AI Flow: Create multi-perspective responses using interactive AI networks

🤖 Morph Prover v0 7B: Open-source conversational assistant for Lean users

🧮 Memora: A vector DB enhanced with multistage reranking

📍 Marauder: Track and recall your movement history

🛠️ Softr AI: Forge new business apps effortlessly

That’s all for today—if you have any questions or something interesting to share, please reply to this email. We’d love to hear from you!

P.S. If you want to sign up for the Supercharged newsletter or share it with a friend, you can find us here.