The first AI 'civilization' 🏙️

Read time: under 4 minutes

Welcome back, Superhuman

With OpenAI’s Strawberry around the corner, “reasoning” is all the rage right now, but what about “reflecting?” A developer just created a model that can not only think through complex problems but can also catch its own mistakes.

Today’s Insights

• An LLM with fact-checking capabilities

• Tutorial: Build an assistant to manage your inbox

• Frontier: Company lets 1,000 AI agents loose in Minecraft

• Everything else you should know today

• 5 new AI tools to boost your productivity

• AI-Generated Images: Go chasing waterfalls

NEXT IN AI

‘World’s top open source model’ knows when it’s wrong

LLMs can be like that stubborn friend who won’t change his mind, even when you show him contradictory evidence. Matt Shumer, the CEO of AI writing platform HyperWrite, thinks he’s found a way to fix that problem. He just unveiled a new LLM called Reflection 70B that’s now considered the top open-source model.

What makes it so powerful? As the name suggests, it can do something that most LLMs struggle with: It checks each of its responses and doesn’t give an answer until it can prove why it’s correct. That makes it much less susceptible to the hallucinations that plague other models. 

Reflection even passes the “strawberry test”: 

• If you ask ChatGPT how many “r”s are in the word “strawberry,” it’ll confidently tell you two

• That’s because LLMs break vocabulary down into individual tokens and can’t always understand the construction of entire words

• Reflection’s fact-checking capabilities help it avoid those pitfalls — and it even features a dedicated “strawberry” button so you can test out the prompt for yourself

Not everyone is sold. A number of accounts on social media are claiming that they have been unable to replicate the results claimed by Shumer. Shumer says they are working on a fix. We’ll keep you updated as we learn more.

PRESENTED BY ARTISAN AI

Backed by Y Combinator: Automate 80% of outbound with AI

Tired of endless outreach with little results? Imagine if someone could fill your calendar with qualified sales meetings – automatically?

That's Ava's job. She's Artisan’s AI BDR who automates 80% of your outbound with…

• 300M+ high-quality B2B prospects

• Automated lead enrichment with 10+ data sources

• Hyper-tailored email drafting

• Personalization Waterfall (social media, technographic data, intent data, etc)

Plus, Ava will even coach you on your sales strategy to help you sell more.

Book a demo and automate 80% of your sales team’s outbound.

THE AI ACADEMY

How to build and email assistant and get through your inbox faster

Watch the tutorial here or follow the instructions below:

1. Go to the Gemini website and create an account.

2. Click on Create new gem in the panel on the left.

3. Give your gem a name.

4. Give your gem instructions, such as: organize my emails by category.

5. Click save.

6. Your gem will now be able to connect to your inbox and organize your emails.

Pro tip: You can give your gem more specific instructions, such as "Categorize each email into invoices and receipts, questions that need a response, newsletters, and everything else".

FROM THE FRONTIER

Developers build first AI ‘civilization’ in Minecraft

A startup called Altera wants to be the first “to create digital human beings that live, care, and grow with us.” One of its first experiments: Unleashing 1,000 autonomous AI agents inside the open-world video game Minecraft — and watching what happens next. The agents reportedly worked together to build their own culture, economy, religion, and government.

Here are some of the highlights:

• The village priest learned that he could convert more residents by bribing them with in-game currency

• The villagers used Google Docs to vote on and amend a constitution

• When several characters got lost, other residents paused their usual work and illuminated the area with torches so they could find their way home

PROMPT OF THE DAY

Improve DevOps Processes

Prompt: Streamline [Organization]'s DevOps with strategic automation. Assess, optimize, and implement a tailored solution. Boost agility, reliability, and cost-efficiency. Navigate the journey with a comprehensive plan.

_{Source: @ashleygarcia PromptPal}

PRESENTED BY INNOVATING WITH AI

Want to become an AI Consultant?

Innovating with AI just welcomed 200 new students into The AI Consultancy Project, their new program that trains you to build a business as an AI consultant:

• Tools, frameworks, and a 6-month plan to build a 6-figure AI consulting business

Click here to request early access to The AI Consultancy Project

AI & TECH NEWS

Everything else you need to know today

• In Cahoots: In a new lawsuit, advanced chip startup Xockets accused Nvidia and Microsoft of colluding to steal its patented AI technology.

• Cyber Diplomacy: The US, EU, and UK have signed the first legally binding AI treaty, which calls on governments to apply democratic values to their AI regulations.

• Quota Crusher: Salesforce is introducing two new models that it says will help companies boost efficiency and automate complicated tasks.

• Project Prodigy: The California-based AI startup Replit has launched a new AI platform that it says can not only code but also build entire software projects from scratch.

🩺 Second Opinion: Pathologists have to piece together data from lots of different sources before they can diagnose a patient with a particular disease. A new AI platform called Alma combines those medical records together, then lets doctors ask questions about a patient’s history with natural language — making the process smoother for both patients and pathologists.

🌌 Out of this World: Dark matter makes up around 85% of the universe, and yet there’s no easy way to detect it. Even trickier, it looks almost identical to other cosmic objects, like black holes, in images. Now, a Swiss astronomer is using a deep learning algorithm to detect dark matter with about 80% accuracy. The tool could help researchers better understand how the mysterious substance helps hold the universe together.

PRODUCTIVITY

5 AI Tools to Supercharge Your Productivity

✅ Moonhub: World-class talent experts that deploy proprietary AI to help you hire top candidates, faster.

✅ Cleve: An AI notes app for your personal brand that turns your quick ideas into optimized posts in seconds.

✅ Jamie*: Receive accurate meeting minutes, tasks, and decisions with Jamie - your bot-free AI note-taker. Enjoy data privacy & retrieve knowledge in seconds.

✅ Maxium: Get real-time insights on your engineering team by categorizing code changes and estimating the effort required to put them together.

✅ Basejump: Empowers teams to access data using the language they speak every day.

^{* indicates a promoted tool, if any}

AI-GENERATED IMAGES

Go Chasing Waterfalls

Midjourney Prompt: Kilchattan Waterfall, rugged cliffs overlooking the sea on Skye Island in Scotland, aerial view, water cascading down into the ocean below, a sense of awe and grandeur, shot with a Canon EOS R5 camera using a wide-angle lens to capture both the waterfall and the surrounding landscape.
--ar 21:32 --v 6.1 --stylize 50

Acquire new customers and drive revenue by partnering with us

Superhuman is the world’s biggest AI newsletter for businesses and professionals with 600,000+ readers and 1.5 Million followers on socials working at the world’s leading startups and enterprises. Companies like Amazon, Hubspot, and Salesforce feature their products in Superhuman. You can learn more about partnering with us here.

🧞Your wish is my command

Got more feedback or just want to get in touch? Reply to this email and we’ll get back to you.

Thanks for reading.

Until next time!

Zain & the Superhuman AI team