Lore Issue #25: Autonomous GPT-4 + Real-time NeRFs

Good morning.

On the surface, this week's news was slow compared to the flood of the last few weeks. However, some of the most exciting experiments to improve GPT-4 have happened and haven't gained enough attention yet.

Lucky for us, we had the opportunity to chat with Yohei Nakajima, the brilliant mind behind babyAGI, delving into the process of its creation. To our surprise, he's a Venture Capitalist that can't code...

Let's get started.

In today's issue:

• INSIGHTS: Autonomous GPT-4
• COOLEST THING: Luma Unreal Engine alpha
• FEATURED TOOL: Gamma
• INDUSTRY UPDATES: Midjourney censorship, Italy bans ChatGPT, and LangChain raised $10M seed round led by Benchmark
• FURTHER STUDY: Mike Solana says we shouldn't bomb data centers

This week several experiments in Autonomous GPT-4 popped up that caught our attention, which we discussed in a tweet thread today.

These projects, most notably babyagi and Auto-GPT, help you to create "AI founders" capable of collaborating with other GPT-4s, managing tasks, and having memory. You can set them on a mission, and they'll set out to try to make it happen.

We had a chance to talk with Yohei Nakajima, the creator of babyAGI, via voice chat to learn more about his project. He has an incredible story.

Here's a short summary of what we learned:

• Yohei's journey in creating babyAGI was rapid, going from idea -> code -> paper -> tweet thread, all within three days. Remarkably, it was all created by AI without Yohei knowing how to code!
• Yohei used GPT-4 to discuss the idea, then fed it relevant documentation, such as how to use LangChain and Pinecone, and GPT-4 created the code for him. He then put the code into a Replit. There were a few bugs, but GPT-4 helped him debug it.
• Yohei is considering using babyAGI as a VC associate for his fund. He believes it's already capable of performing very basic tasks on its own, potentially even automating entire functions of a small business.

It's possible that these experiments could lead to a significant expansion of what is possible with GPT-4. Andrej Karpathy, former head of AI at Tesla, stated that "AutoGPTs" might be the next frontier in prompt engineering.

This has significant implications for Google and others in tech hoping to catch up. OpenAI now has a very active developer community that is starting to create a flywheel of daily improvements to what you can do with LLMs.

If you combine that with the fact that OpenAI is already working on GPT-5 + rolling out plugins, it's hard to see how Google catches up.

It's possible that in the next 5-10 years, Google Search will become irrelevant. And OpenAI becomes the leading technology company in the world.

So, what does this mean for your business? Here are our insights.

Insight #1: Embrace Autonomous GPT-4

The emergence of Autonomous GPT-4 experiments like babyagi and Auto-GPT presents an opportunity to possibly automate parts of your business. In the future, it might even lead to entirely automated businesses. You should understand this as much as possible and be prepared.

Test these technologies out to see what's possible and better understand the limitations. By embracing them, you can identify new avenues for innovation and productivity enhancement in your organization.

Insight #2: Plan For An OpenAI Future

If our prediction holds true and OpenAI emerges as the winner of the AI race, it's crucial to consider what this means for your business.

Are you well-positioned for a future where SEO might matter less, and the effectiveness of your integration with OpenAI matters more? Reflect on your current strategies and explore ways to adapt and thrive in an environment where OpenAI is the leader in tech.

Insight #3: Rethink Creative Processes with AI

Yohei went from idea to popular project in 3 days. And, the AI created everything from code to the paper to diagrams, and even the tweet thread. What does this mean for open-source projects and how we should all be thinking about creating?

We're personally thinking through ideas that we have that we've assumed would be hard to build, but perhaps could be created + promoted in a matter of days using Yohei's approach.

What's Next?

Over the next few weeks, we plan to experiment heavily with babyagi and Auto-GPT to grow our business. We'll be sharing what works and doesn't along the way with all of you.

GPT-4 Automation Resources

• babyagi by Yohei Nakajima - View GitHub
• Auto-GPT: An Autonomous GPT-4 Experiment - View GitHub
• JARVIS, a system to connect LLMs with ML community - View GitHub

‌Luma continues to be one of our favorite companies in AI.

This week, they unveiled their latest innovation: an Unreal Engine plugin capable of real-time, fully volumetric NeRFs (Neural Radiance Fields) generation. This groundbreaking development is poised to revolutionize the way we create and interact with 3D environments.

This innovation paves the way for entirely new types of video games and entertainment experiences.

People are already making awesome demos with it.

Other things we found

Midjourney V5 Niji looks incredible for Anime. Just add --niji 5 to your prompt.

Matt Wolfe was able to make a video game demo with zero coding knowledge.

The "first conversational holographic ai" by Looking Glass.

We've tried a few AI presentation tools before, like Beautiful.ai. But unfortunately, we didn't find them to be that useful.

This week we finally found one we like, Gamma.

Gamma offers an innovative way to create engaging, visually appealing content without the need for design or formatting work. Start writing and let Gamma's AI generator turn your notes into polished presentations.

There is room for improvement, but it's a great start. Try it out.

Other Tools We Found

• EnhanceAI - Add AI autocomplete to our website in 2 minutes
• Glasp - A social web highlighter with AI features

Midjourney Censoring For China

A story about Midjourney censorship came out this week that caused a bit of backlash in the AI community when it was noticed that Midjourney allows you to create images of Joe Biden and Donald Trump – but you can't say Xi Jinping. Or China President. Or even Xi.

Midjourney CEO Says ‘Political Satire In China Is Pretty Not Okay,’ But Apparently Silencing Satire About Xi Jinping Is Pretty Okay

As a rule, it’s a good idea to be particularly suspicious of defenses of censorship that — coincidentally — materially benefit the people espousing them. In this case, the argument in favor of cens…

TechdirtSarah McLaughlin

Several notable people in the AI community, including hardmaru from Stable Diffusion, responded by tweeting AI-generated photos of President Xi.

Italy Banned ChatGPT

In an unexpected move this week, Italy banned ChatGPT, the popular AI tool created by OpenAI.

ChatGPT banned in Italy over privacy concerns

The country’s data-protection regulator has serious privacy concerns over the technology.

BBC NewsShiona McCallumTechnology reporter

‌An Italian minister has responded by calling the ban excessive.

We're also hearing rumors that Germany might move to do the same. It feels like things will keep getting more intense with regulation and geopolitics as the AI era begins.  

Funding Roundup

Announcing our $10M seed round led by Benchmark

It was only six months ago that we released the first version of LangChain, but it seems like several years. When we launched, generative AI was starting to go mainstream: stable diffusion had just been released and was captivating people’s imagination and fueling an explosion in developer activity,…

LangChainLangChain

• AI-powered search engine Perplexity AI lands $26M, launches iOS app - Read article
• Blockchain startup Fetch.ai grabs $40M to provide monetization and other tooling for AI-generated information - Read article
• AI startup Fourthline locks down $54M to bring better ID checks and compliance tools to the finance sector - Read article
• Covariant’s robotic picking AI nabs another $75M - Read article

In this article, Mike Solana discusses the rise of AI doomsayers—particularly those within the tech industry.

Bomb the Data Centers (Smart People Agree)

tuesday report #10 // on the AI doomer suggestion we start bombing data centers, speculative nuclear apocalypse fan fiction revisited, and all the news you missed this week

Pirate WiresMike Solana

He criticizes Eliezer Yudkowsky's recent TIME magazine piece, which argued that the risk of nuclear war is preferable to the release of OpenAI's latest chatbot.

Other things to study

• JARVIS, a system to connect LLMs with ML community - View GitHub
• ‌The AI Index Report by Stanford - View Report

           ‌

Closing

See you next week!

-Nathan Lands

If you ❤️ Lore:

1. Share this newsletter 🥰
2. Follow us on Twitter @NathanLands & LoreAiNews.
3. Join our Generative AI community on LinkedIn.