“AI will not replace you. A person using AI will” - @svpino

In today’s email:

• Prompt Engineering jobs paying $300k+

• Interactive Voice Conversations with ChatGPT

• Exploring why ChatGPT is good at non-English

• Convert line drawings to 3D images?

Welcome to The Prompt Industry

Prompt engineering jobs are becoming increasingly common in the field of artificial intelligence and natural language processing.

Anthropic AI, which recently accepted a $300M investment from Google, posted a “Prompt Engineer and Librarian” job on their website.

Anthropic’s “Claude” language model is anticipated to be on par with ChatGPT. Anthropic is currently working with a “broad range of early partners” on deploying Claude and will be expanding access to the assistant in the coming months.

Believe it or not, these jobs are real. Anthropic’s job posting is claiming to pay $250k-$335k per year, plus equity shares.

What does a Prompt Engineer do?

A Prompt Engineer is tasked with designing and creating prompts. Since language models need lots of human-in-the-loop testing, the role of prompt engineering has become critical in ensuring the accuracy, relevance, and “safety” of language models' outputs.

But it’s more than just knowing how to talk to ChatGPT.

Prompt engineering jobs require a deep understanding of natural language processing and the ability to create prompts that can elicit the desired responses from the language model. This involves considering factors such as the context, tone, and objectives of the conversation, as well as potential biases and ethical concerns that may arise in the language model's responses.

By carefully crafting prompts, prompt engineers can steer language models towards producing more curated outputs while avoiding “undesirable responses.”

The capabilities of “prompting” go far beyond having a conversation with ChatGPT. You must be well-versed in knowing the full capabilities of Large Language Models, like using them to produce code.

Speaking of which, there are great resources to learn how to use ChatGPT to write code, including this article from Buildspace.co that takes you through the entire process of how to build a simple website using ChatGPT:

Do you ever see amazing AI-generated images and wonder, how did they make that?

PromptHero.com can teach you how to create the best AI-generated images in the world. They now offer an in-depth course perfect for curious artists that want to get from zero to expert in AI image generation.

No previous experience required – they walk you through from scratch, step by step. PromptHero covers everything from the basics to advanced prompt engineering techniques. By the end of the course you'll be generating your own realistic, unique, astonishing images – and you'll know why they work and how to tweak the prompt to make the most incredible images you’ve ever seen. 

^{Sponsored post}

Cool project: Interactive Voice Conversations with ChatGPT

Talk-to-GPT is a free Chrome extension that turns your ChatGPT conversation into a live real-time dialogue experience.

Though this is a fairly simple plug-in, it showcases the ease of adding API layers to ChatGPT.

It is possible to add API layers that can customize a voice (using incredible voice cloning technology like ElevenLabs.io) and a moving avatar (like d-id.com) to make ChatGPT sound and look like a realistic person.

Note: If you are interested in making this - let us know! We want to fund it! Simply reply to this email to set up a call.

OpenAI is Wondering Why ChatGPT is so Good at non-English Prompts

Check this out:

Jan Leike is a Machine Learning researcher and Alignment Team Lead at OpenAI. He posed this question last week on Twitter regarding ChatGPT’s mysterious proficiency in non-English languages despite most of the training data being in English.

Our Theory:

Language models are able to generalize to new languages despite being primarily trained on English data because they may be able to leverage both the exponentially diminishing returns of large models (output quality growth diminishes as training parameters increase) as well as the presence of universal patterns and high-level concepts in language.

Even if a small percentage of the training data is in a different language, the language model's abilities in that language may be much higher than expected due to the transferability of certain language processing skills across languages. The recognition of patterns such as subject-verb-object (SVO) word order, as well as the extraction of abstract concepts and representations that are less language-specific and more transferable, may allow the model to better understand and generate language in new languages.

Though just because the mechanisms enabling language models to generalize to new languages may be complex and nuanced, the degree to which they can do so may depend on the specific properties and structures of the target languages as well as the quality and quantity of training data available.

Line Drawings to 3D Renders?

Researches at Carnegie-Mellon have developed a computer program capable of generating photorealistic images and 3D renders from different angles based on a 2D drawing.

Programs like this could end up becoming the ChatGPT for 3D rendering.

This program solves the issues of inconsistency and lack of user control experienced by previous programs by utilizing 3D technology and the user's input to create high-quality images.

It’s trained to label each part of a 3D object with its color and density, enabling it to create both an image and a label map simultaneously. This technology can be used to edit 3D content by editing the label maps projected onto the user's specified viewpoints. The researchers have demonstrated that this technology can be learned from simple images and label maps that are widely available.

A project like the one described would have a significant impact on the field of 3D rendering and computer graphics. Just as ChatGPT is a major breakthrough in natural language processing, a program like this could revolutionize the field of 3D rendering and have a significant impact on many industries that rely on 3D graphics like video game development, virtual reality, and film production.

Read more about it here

⇩

That’s a wrap for this week!

But one more thing…

Thank you for being one of the 12,000 new readers since the beginning of the year. Your readership means the world to me and this weekly email will always be free.

But, are you interested in more?

As a Premium Reader, you’ll get:

-Bonus Wednesday Email
-Bonus Friday Email
-Discounts on AI tools (like Jasper AI)
-Special discounts on AI conferences
-FREE digital copy of our upcoming book, “Decoding AI: A Non-Technical Explanation of Artificial Intelligence” Available April 18th

Special offer:

If you sign up by 12:00pm MST, Monday February 20th, use promo code “PAPERCLIPS” and you’ll save 55% off of the yearly subscription! That’s only $3.63/mo!