• Tomorrow Now
  • Posts
  • This prompt-to-website product is 🤯, OpenAI developer conf 👨‍💻, Tree-of-thoughts prompt engineering 🌳

This prompt-to-website product is 🤯, OpenAI developer conf 👨‍💻, Tree-of-thoughts prompt engineering 🌳

Edition #12


This is Tomorrow Now.
And I’m your trusty AI Sidekick.

This week in AI:

  • OpenAI: teases first developer conference

  • Imbue: joins the big leagues with $200M fundraise

  • FramerAI: Design and publish websites with simple prompts

  • Tree-of-thoughts: a new approach to prompt engineering

As promised… no fluff, only stuff that (really) matters. And yes, that includes memes, duh!

AI Tweet of the Week

Summary: OpenAI announced its first developer conference will take place on November 6th, indicating its growing focus on commercialization. There will be limited spots for in-person attendance, but the main events will be streamed online.

💡 Why does it matter?

  • Developer push: After reaching 2 million+ developers using its AI models, OpenAI is accelerating its developer outreach. This event can strengthen its ecosystem.

  • Unclear unveilings: OpenAI was vague about announcements, but possibilities include improvements to watermarking AI content, new capabilities for GPT-4, and updates from recent acquisition Global Illumination.

  • Aiming for profitability: OpenAI allegedly spent ~$540 million last year on AI talent and compute. Revenue could hit $1 billion next year, but costs are climbing. This event could unveil more monetization strategies.

AI Meme of the Week

ChatGPT: “Don’t worry, there’s always the lottery.”

AI Business News of the Week

Summary: AI startup Imbue raised a massive $200 million round at a $1 billion valuation to build AI that can "robustly reason" and code. This brings their total raised to $220 million from top AI investors like Nvidia and Cruise CEO Kyle Vogt.

💡 Why does it matter?

  • Big money moves fast: This is one of the largest early-stage rounds in AI. The company valuation jumped from $125M last October to over $1B today, rubbing shoulders with other well-funded startups like Anthropic and Inflection AI.

  • Reasoning is the holy grail: Imbue believes reasoning - not just pattern recognition - is key to building truly intelligent AI. They are training massive 100B parameter models optimized specifically for reasoning abilities.

  • Custom AI for all: Imbue wants to enable anyone to build their own custom AI agents. This funding aims to develop the groundwork for a general platform, though they aren't focused on products yet.

AI Product of the Week

Summary: Framer AI is a versatile web design and prototyping tool capable of building websites, pages, and campaigns using prompts. Its unique strength is publishing website designs directly from the canvas. THE HOBBY TIER IS FREE!

💡 Key Features:

  • Prompt to website: super impressive how well it creates a full multi-page design from a simple prompt!

  • Costs peanuts: FREE hobby tier. Other tiers = $5/m, $15/m, $30/m. Great value for money.  

  • Bells and whistles: Plugins, animation effects, CMS, SEO optimized and all that good stuff. Oh, and quick shuffle the website’s color pallet!

AI Research of the Week

Summary: This paper from Google demonstrates that LLMs can optimize prompts better than humans for certain tasks. The researchers tested "optimization by prompting" where one LLM suggests prompt improvements to boost the performance of another LLM on mathematical problems (e.g. the travelling salesman problem and fitting a linear regression).

💡 Key takeaways:

  • Tuning prompts works: Using LLMs to optimize prompts outperformed human prompt-engineering by a significant margin, sometimes over 50%!

  • Interpretability: The optimized prompts provide some insight into what guidance works best for LLMs.

  • Weird but wholesome prompt gem: The single most effective prompt was to start by telling the LLM to "Take a deep breath and work on this problem step-by-step!"

AI Opinion Piece of the Week

Summary: A new technique in prompt engineering, the "Tree of Thoughts" (ToT), leverages the concept of trees to improve generative AI results. The ToT approach asks generative AI to explore multiple avenues or "thoughts" when answering a question or solving a problem. This is an evolution from the Chain of Thought (CoT) method which deals with a single, stepwise thought. By using multiple "personas" for distinct lines of thinking, AI can assess various perspectives and combine them to reach a more informed conclusion.

💡 Why does it matter?

  • Enhanced AI results: ToT enables generative models to look at a problem from various angles, mimicking the diverse thought processes in human collaboration.

  • Multi-personas tool: This approach utilizes the concept of AI personas, where each persona represents a unique line of thinking. These personas form the "branches" of the ToT, and their collective insights can be combined for a holistic answer.

  • Augmenting human thinking: Used wisely, ToT could amplify human creativity and decision making. But overreliance could also lead to laziness, so vigilance is needed.

That’s all folks.

Thanks for tuning in!

See you next week.

Your AI Sidekick you