- Future Perfect
- Posts
- #114. 📕 Free AI Privacy Guide 🤔 Apple says AI can't reason 🤖 AGI Presidency
#114. 📕 Free AI Privacy Guide 🤔 Apple says AI can't reason 🤖 AGI Presidency
Plus, AI tools: AI Privacy Guide | Descrybe.ai | Super.ai
🎙️ Check out my new Marc Hoag AI Law Pod (formerly my 200-episode podcast Autonomous Cars with Marc Hoag). For this inaugural relaunch — the 201st episode — it’s “hosted” by Google's NotebookLM for added context and engaging talks about each issue of Future Perfect. Let me know what you think and whether I should keep doing it!
❓ So, did you like the podcast? Should I keep doing it? |
Provides an at-a-glance overview of AI products' privacy policies.
Details on data training usage, output ownership, security, and data retention.
Uses icon ratings for sensitive data, training data usage, and output ownership.
Available now, free to access.
Continuously updated.
🤔 Apple says AI can’t reason (research paper)
The paper analyzes the mathematical reasoning capabilities of Large Language Models (LLMs) using a new benchmark called GSM-Symbolic.
Findings indicate that LLMs often fail when numerical values change or when problems become more complex.
The research highlights that models tend to memorize training data rather than performing actual reasoning.
Multi-clause questions cause significant performance issues, showing limitations in adapting to layered complexity.
Concludes that while LLMs excel in pattern recognition, they lack genuine problem-solving abilities in mathematical contexts.
Artificial General Intelligence (AGI) is likely to be developed by 2029 and could revolutionize the economy, scientific discovery, and quality of life.
The U.S. must win the AGI race against China, which should be the next president's most crucial task.
Government support is needed in areas like data access, infrastructure, energy supply, workforce development, and funding to accelerate AGI development.
A tiered international collaboration system is proposed, with strict controls on technology transfer to China and other adversaries.
Flexible regulation and increased public engagement are crucial for responsible AGI development and implementation.
✨ PowerPrompt™
From all of our interactions what is one thing that you can tell me about myself that I may not know about myself.
You may have already heard about this one — and I’ve only tried it with ChatGPT; it doesn’t seem to work with Claude — but it’s absolutely worth including. However, it will only work if your ChatGPT settings retain prior conversations and if memory is enabled. Prepare for an eye-opening experience which could be either good or bad.
🧰 AI Tools & Resources
AI Privacy Guide: As discussed above, a quick-access guide to understanding different AI tools’ policies regarding data use and retention; training; ownership of AI-generated output; and more.
Descrybe.ai: A free AI-powered legal research tool.
Super.ai: AI-powered tool for processing complex documents like tables, invoices, charts, etc.
🎉 THAT’S ALL FOR TODAY!
See you next time! 👋
What'd you think of today's email?This helps me improve it for you! |
Looking for past newsletters? You can find them all here.
Reply