Highlights💡 

OpenAI’s plans according to Sam Altman [Link]

1. OpenAI is heavily GPU limited at present

2. The near-term roadmap includes faster GPT-4, Longer context window (Up to 1 million tokens are plausible), and A stateful API (a version of the API that remembers the conversation history)

3. Plugins “don’t have PMF” and are probably not coming to the API anytime soon

4. OpenAI will avoid competing with their customers — other than with ChatGPT

5. Regulation is needed but so is open source

6. The scaling laws still hold (making models larger will continue to yield performance)

Did you know you could do this? (Think of using shared chats as templates or memories for new chats)

Learn AI like it’s 2023 [Link]

AI won’t take your job. Someone using AI will. Best time to level up? Yesterday. Second best? Right now.

Luckily, there’s Brilliant — the interactive app that makes it easy to master concepts in math, data, and computer science in just minutes a day.

Here’s how:

• They have thousands of lessons on tons of topics — from AI and neural networks to data science.

• They break down complex concepts into digestible building blocks that stick.

• Their interactive style keeps you engaged, so it’s easy to build a daily habit.

Join over 10 million people around the world and start building skills in minutes a day. You can try everything Brilliant has to offer for free for a full 30 days. Plus, right now you can get an exclusive 20% off an annual premium membership.

Let’s Verify Step by Step - OpenAI's new impressive blog and Paper [Paper], [Blog]

TLDR: We’ve trained a model to achieve a new state-of-the-art in mathematical problem solving by rewarding each correct step of reasoning (“process supervision”) instead of simply rewarding the correct final answer (“outcome supervision”)

• Large language models often produce logical mistakes, called hallucinations, in complex multi-step reasoning.

• The study compares two methods of training reward models to detect hallucinations: outcome supervision and process supervision.

• Process supervision, which provides feedback for each step in a chain of thought, leads to significantly better performance compared to outcome supervision.

• Process supervision has alignment advantages as it rewards aligned thinking and encourages interpretable reasoning.

• Outcome supervision may reward unaligned processes and is harder to scrutinize.

• Process supervision incurs a negative alignment tax in the math domain, meaning it performs better while being more aligned.

• If the results generalize, process supervision could offer a more performant and aligned method compared to outcome supervision.

Midjourney v5.1 Tutorial - How to Build a Prompt - Including Prompts + Beginner Friendly

Tools & Links 🛠️

• Elai - Create AI videos from just text [Link]

• OpenChat - Your own ChatGPT for anything (Open source) [GitHub], [Demo]

• Roll.ai - Create professional, studio-quality remote videos in minutes, with just an iPhone. [Demo], [App Store]

• UC Berkeley AI Hackathon - June 17th - 18th @ UC Berkeley [Apply here]

• Hyper: AI Companions & VTubers [App Store], [Demo]

• DietPrompt - Motivational speeches, only for you. [Link]

• Graphy App AI - Now it's even faster to generate charts and insights [Try here][Demo]

• Recraft AI - Generate vector art, 3D images, and more [Link]

• Bolt AI - Instant access to ChatGPT on any Mac app [Link]

• Rask AI - Localize videos. Fast. Fun. With AI [Link]

• DeepLearning AI launched a new mini-course in collaboration with OpenAI: Building Systems with the ChatGPT API [Link]

• Concept - Apple is expected to unveil the long-rumored Reality Pro at WWDC on Monday. This is what we think it looks like based on patents, reports, and more [Link]

• Japan Goes All In: Copyright Doesn’t Apply To AI Training [Link]

• Meta Quest Gaming Showcase is back for 2023! Tune in on Thursday, June 1, at 10:00 am PT for the exclusive game reveals and more. [Watch here]

Unclassified 🌀 

• Morning Brew - Get smarter in 5 minutes (it's free) [Link]

• Fiverr - Find the right freelance service, right away [Link]

• WFH.Team - Work from anywhere in the world [Link]

• Add your link here ➜

AI Is Not an Arms Race [Link]

Go from 'we' to 'me' in just one click

Hope you enjoyed today's newsletter!

Brought to you by eeeziii

Did you know you can add Superpower Daily to your RSS feed https://rss.beehiiv.com/feeds/GcFiF2T4I5.xml

⚡️ Check out our Superpower ChatGPT extension on Chrome Web Store and Mozilla Add-Ons Page.