AI Threatens 3D Artist Livelihoods: My Human Creativity vs. Point-E 3D Generator

Table of Contents

• Introduction
• How Point-E's 3D Generator Works
• My Process for Quick 3D Model Ideation and Creation
• Comparing My Table to Point-E's Creativity
• Results: Taste Test by Impartial Basement Hostage
• Conclusion and Key Takeaways

Introduction: The Threat of AI to Creative Professions

The advent of AI image and 3D generators poses a grave threat to creative professionals. With new AI tools like DALL-E 2, Midjourney, and Stable Diffusion churning out impressive 2D artworks, and tools like Point-E quickly generating 3D models from text prompts, many artists fear being replaced by machines.

In this blog post and accompanying YouTube video, I put Point-E's 3D generation capabilities to the test in a man vs. machine creative challenge. By rapidly ideating and modeling a 3D table, then comparing it to Point-E's output, I aim to prove humans still reign supreme in true creativity...for now.

The Existential Threat Posed by AI Creativity

The recent explosion in AI art marks a seismic shift for creative industries. Tools like DALL-E 2 and Stable Diffusion create bizarre, beautiful, and incredibly realistic images from short text prompts. Animation studios and VFX houses are experimenting with using AI to build 3D assets and animate characters. For creative professionals who have spent years honing their craft, this rapid advancement of AI threatens to make many jobs obsolete. Why hire a concept artist or 3D modeler when an AI can pump out usable assets in seconds? The cold, calculating nature of AI also lacks human originality and emotion.

My Challenge: Out-Create Point-E's 3D Generator

To test if human creativity still reigns supreme, I challenged Point-E's 3D generation capabilities. This new AI tool instantly converts text prompts into 3D dot models. However, it fails to connect the dots into coherent shapes. My goal was to rapidly ideate and create a 3D model of a table, then compare my output to Point-E's. If I could produce a more realistic and purposeful model, it would prove AI still falls painfully short of human creativity, intuition and problem-solving.

How Point-E's 3D Generator Works

On first glance, Point-E's 3D object generator seems like an incredible feat of AI creativity. Input any text prompt, and it churns out a 3D point cloud model in seconds. However, under closer inspection major limitations become clear.

Turning Text into 3D Dot Models Rapidly

Point-E leverages advanced natural language processing to interpret text prompts. It then generates a 3D model constructed of dot points floating in 3D space matching key features of the description. For example, typing "blue race car" outputs a vague race car shape made of dots. While impressive in its speed and responsiveness to prompts, lacking solid surfaces makes the models basically useless.

Failing to Connect the Dots for Coherent Shapes

The biggest failure of Point-E's 3D generator is its inability to connect generated dot points into polygonal 3D surfaces. The output models resemble fuzzy point clouds rather than being solid, functional 3D objects. This likely stems from technical limitations in efficiently processing so many potential mesh connections. But it severely limits the real-world usability of the generated assets compared to those created by human 3D modelers.

My Process for Rapid 3D Table Ideation and Modeling

To test if I could out-create Point-E's 3D generator, I gave myself 1 minute to concept, model, and texture a basic table model. Thanks to my creative intuition and 3D modeling expertise developed over 10+ years, I produced a simple yet coherent and realistic table.

Brainstorming and Concepting the Table Model

The first step was to conceptualize exactly what kind of table I wanted to build. I decided on crafting an outdoor patio table, with a stone top, simple metal legs, and some decorative touches. Having a solid creative concept before beginning modeling was key. The AI likely fails to do this ideation step, instead just reacting to the text prompt verbatim.

Executing the Build Rapidly Based on the Concept

With my desired table design in mind, I rapidly modeled it in Blender. I built a simple plane for the top, added a stone texture, attached cylindrical legs, and added decorative metal rings and rivets. The process relied heavily on my artistic intuition for shape, size, layout, detailing, texturing, and material choices. These are deeply human creative abilities AI still fails to match.

Comparing My Table to Point-E's Creativity

With my table model complete, I could properly assess how it stacked up against Point-E's randomly-generated output. While simplified due to the tight time constraints, my table clearly showed greater intentionality, realism, and usability.

Assessing Size, Shape and Structural Purpose

The first obvious point of comparison is size, proportions and structural soundness. My table has thicker legs and decorative braces to support the stone top's weight. The surface is also sized appropriately for plates, glasses, etc. By contrast, Point-E's table legs are spindly and uneven. The top shape makes no sense for actually resting objects. There is no sense of real-world physical logic, gravity, or functionality.

Determining Realism, Detail Levels, and Usage Purpose

Looking closer, my table also has greater realism in textures and materials. The stone looks rugged and heavy, while the metal shows weathering and decorative detail. Point-E's table bizarrely seems made of toy building blocks rather than actual furniture materials. The context and utility of the generated table is also unclear. As a human creator, I consciously designed my table for an outdoor patio setting. The AI-generated model lacks any such intentionality or implied purpose.

Spot the AI: Taste Test by Impartial Basement Hostage

The ultimate test was whether a regular person could tell which table was AI-generated versus human created. I painted over the two tables then showed them to an impartial 'volunteer' locked in my basement.

The Hostage Test for Spotting Human Creativity and Intent

No creative work exists in a vacuum. The true test is whether it resonates with a real human audience. I conducted a literal 'taste test' by letting my totally willing basement guest view the anonymized table models. This tested whether Point-E's output matched basic expectations of a usable table. If the hostage guessed wrong, it would show the AI is crossing the 'creativity uncanny valley' into mimicking human work.

I Fool the Test Subject - A Clear Win for Human Creativity!

In the end...the hostage thought my table was the AI creation! This shocking outcome clearly demonstrates Point-E's severe limitations in matching intentional, structured human creativity. There is still an enormous gap between AI art and the years of specialized skills, training, and emotional intelligence needed to conceive and actualize ideas as a human creative professional. But make no mistake, the threat is real and evolving quickly!

Conclusion and Key Takeaways

This challenge helped quantify the current state of AI as both a threat and inspiration for human creatives. While rapid advancement means all artists must continue adapting their skills, true creativity remains decidedly human (for now).

The key lessons learned are...

• AI art tools have severe limitations in creative logic, emotion and intent

• Specialized skills like 3D modeling are likely safest in the near term

• But constant skills development is needed as AI evolves

By maintaining uniquely human qualities like curiosity, empathy and purpose, artists and creators will always retain an edge over AI...even if we sometimes fool ourselves in the process!

FAQ

Q: What compelled you to challenge an AI 3D generator?
A: To prove the continued necessity of human artists and creatives versus rising threat of replaceable AI tools.

Q: What were the key differences you noticed in the 3D models?
A: My table model showed greater coherence, realism, and purpose compared to the AI's random, disorganized dots.