Good morning. It’s Monday, May 8th.

Python may find itself obsolete with the arrival of AI coding language Mojo, meanwhile, Shap-E is shaping up to be the most impressive development we’ve yet seen in the field of text-to-3D.

In today’s email:

• Shap-E, Text-to-3D is Here (No Blender 3D Required)

• Mojo is Gearing Up to Become the Definitive AI Programming Language

• Trending AI Tools

• Top 5 AI Research Papers of the Week

You read. We listen. Share your feedback by replying to this email, or DM us on Twitter.

Shap-E, OpenAI’s Text-to-3D is Here

Just like DALL·E for text-to-image, Shap·E is a cutting-edge AI model from OpenAI designed to generate 3D objects from text prompts.

The project, currently available on Github, presents a new method for creating 3D models, addressing the limitations of OpenAI's previous model, Point·E, which was capable of creating 3D point clouds from complex prompts but struggled with capturing fine-grained shapes and textures.

Shap·E is designed to generate the parameters of implicit functions directly, enabling the rendering of both textured meshes and Neural Radiance Fields (NeRFs). NeRFs are continuous volumetric scene representations that have been revolutionary in the domains of augmented reality and virtual reality, enabling the generation of photorealistic renderings of 3D scenes.

Shap·E combines NeRFs with diffusion models to efficiently generate detailed 3D models that capture the shape and texture of objects more accurately than its predecessor, Point·E. Shap·E can generate a sample in roughly 13 seconds, without requiring a separate text-to-image model. This is significantly faster than Point·E, which produced 3D models in one to two minutes on the same hardware.

The generation process with Shap·E can be computationally “intensive”, and the model may still struggle with certain complex objects. Though pound-for-pound, it’s the most impressive model yet for 3D object generation.

Read more: Full Paper, Github

Scribe AI

Scribe is your GPT-4-powered process documentation platform that automatically creates SOPs, help centers, new user guides and process overviews for any business process. Scribe AI auto-generates step-by-step guides complete with screenshots and text by capturing your screen while you click and type. Using your guides, Scribe AI can create full process documentation (including headings, subheadings, and detailed text) with your guides automatically embedded. No more staring at a blank document thinking, "ok, I have to teach someone how to do this, where do I start? what are all the steps". Scribe AI does it for you.

_{sponsored post}

Personalized AI Meditations: Ogimi

Ogimi AI is your personal mindfulness coach, offering custom guided meditations tailored just for you​.

Ogimi creates AI-generated personalized meditations that adapt to your needs and preferences, or learn the fundamentals with recorded guided meditations​.

The benefits of regular meditation are wide-ranging. From reducing stress and anxiety, improving focus, emotions, and relationships, to aiding in faster recovery and better sleep, Ogimi AI is designed to help you achieve a healthier, more balanced life.

Begin your journey to mindfulness today with Ogimi AI. Find your inner peace, sharpen your focus, and cultivate a deeper understanding of yourself.

_{sponsored post}

Trending AI Tools

1. Permar.xyz: Permar is a convenient way to get AI to shape your landing pages with a single prompt. Gone are the days of hiring web developers to get your site up and running. Beyond simply helping build up your site, Permar also uses the stats that it gathers to continuously improve on your site.

2. Oasis: This speech-to-text AI allows you to simply dictate your emails into your phone and it can intelligently parse out what you’re saying to make both personal and business emails a breeze.

3. Adsgency.ai: Generating creative ads has historically required talent and creative vision in no small measure, but AdsGency puts the power in your hands. This AI suite allows you to create everything up to brief commercials using AI without needing a team to back you up.

4. Sttabot: Sttabot does away with the need for humans to create AI in the first place… almost. Users enter a prompt for an AI-based app of your own and Sttabot will create it for you. Sttabot can build you anything from a chatbot to a data gatherer.

5. GuideAI: GuideAI seeks to use the power of AI to ease the generation and distribution of audio guides for everything ranging from museums to tourist attractions. This admirable effort seeks to make places more inclusive for the visually-impaired and even supports automatic translation in up to 11 languages.

Top 5 AI Research Papers This Week

Note: arXiv is a free online library where scientists share their research papers before they are published. These are the 5 most viewed papers related to AI in the last week.

Assessing Whether We Can Trust AI Systems, and if so, How Do We Quantifiably Measure It? 

Determining whether or not we can trust AI is one thing, but how do we quantity our trust for it?

The writers of this paper set out to do so by coming up with a spectrum of trustworthy qualities that an AI can exhibit. They seek to quantify : functionality, reliability, security, robustness, usability, compatibility, ethical, and regulatability.

Of the aforementioned areas, the researchers found that the fields of reliability, functionality, and ethics had the largest relative effects on the trustworthiness of an AI. Other factors that the researchers looked into included the level of scope in the system’s automation as well as the degree of application criticality.

Context Improvement for Unannotated Data Sources to Remove Disambiguation in Associations, Basically Improved Connection-Making When AI “Thinks”

This paper discusses a method of improving context gathering for AI in terms of its ability to understand language. This method seeks to improve AI’s performance in the absence of annotated data, making it less reliant on external influence to understand language.

Over the course of the study, the researchers found that they were making headway, producing results that were superior to existing techniques. Previously, AI has had trouble recognizing that words’ meanings can change based on the word’s context - but this new model promises to change that.

Angle-Based Persuasion/Framing Detection Using AI 

This paper posits that AI can be used to determine whether media is biased or if a piece is framed in a certain way.

The researchers argue that the ever-increasing number of news sources calls for an AI-based method of separating fake news and propaganda from more reputable articles.

Self-Improving Prompting Using Gradients

One of the most frustrating aspects of LLMs is the prompt reliance on human generation to enact functions. Using gradients, self-improving prompting is possible which can also help account for and prevent blatant and subtle attempted jailbreaks by users and by autonomous models themselves.

The paper proposes a solution to this problem via APO or Automatic Prompt Organization which is in part based on using numerical gradient descent to automatically improve prompts. Initial tests have led to 31% improvements over other selected solutions. This will make it easy for LLMs to train on non-annotated data sets and not have to be reliant on vague prompts, instead self optimizing for specific instructions.

Implementing LLM in Classrooms (Read this if you’re a teacher)

Today’s final paper broaches the potential usefulness of implementing AI in classrooms.

The applications explored in the paper include the examination of content generated by AI, dialogue auto-completion, and other techniques to help students increase knowledge retention.

3x the information, for less than $2/week

Stay informed, stay ahead: Your premium AI resource.

Thank you for reading today’s edition.

Your feedback is valuable.
Respond to this email and tell us how you think we could add more value to this newsletter.

Read by employees from