GPT-4 Vision: Seeing More Than What Meets the (Computers') Eye

Welcome to another edition of the best damn newsletter in AI.

This free newsletter is designed to keep you ahead of the curve and open your mind to using AI in your work and business.

Digging deep in AI for work or AI Operations? Take a look at our membership.

Our #1 goal is to be useful. So please shoot us an email 📩 if you have questions or feedback, and especially if you implement something we share!

Here's what we're covering today:

• Why GPT-4 Vision is a step up from normal computer vision

• Reframe negative thoughts with this free chatbot (oh and you can also build your own)

• Content and copyright keep making the news, so we’ll keep covering it and making sure you’re up to date!

Let’s get to it! 👇

TODAY'S PERSPECTIVE

GPT-4 Vision: Seeing More Than What Meets the (Computers') Eye

GPT-4 vision is out across all ChatGPT Plus users, and rumor on the street is that OpenAI is adding this capability to their API at their upcoming Dev Day.

We want to take today’s edition to explain why this will be a big deal, and how to be prepared:

First, “computer vision” isn’t new. In fact, self-driving cars are one of the most popular and advanced use cases of it. Those car computers can see so well, they can drive!

But computer vision models have traditionally required lots of data, very specific tasks, and if you wanted a computer vision model to understand something in an image - you had to specifically teach it to do so.

AI Models like GPT vision work differently. Instead of simply seeing the image; these models embed images to map its understanding, much like how it comprehends language. So instead of being taught explicitly what certain things look like in an image; it’s taught to understand.

And what gets AI people excited here, is that this capability can identify insights that humans may miss. And thus things we haven’t taught computers to “see” or understand yet.

You might not be impressed with GPT-4 Vision yet, but we’re basically on version 1 of these multimodal vision models.

The best thing you can do is begin testing out use cases in your ChatGPT Plus account so that you can start developing this side of your AI literacy and intuition.

Just like GPT, it’s likely we’ll see an explosion of study around how to prompt them, how to teach them new tasks - and most importantly for business, how to mitigate the risk of hallucinations.

And the more you experiment, the more you’ll be ready to take advantage of the new techniques and new products as they emerge.

USE CASE

Reframe those negative thoughts

In case you forgot, you can create your own free AI bots on Poe.com, the AI bot platform built by Q&A tech platform Quora.

One of our favorites was made by our friend Allie Miller, a prominent voice and thought leader in AI.

Meet Reframe AI, the bot designed to turn your negative complaint into a positive reframe.

Try it out for free here

LINKS

For your reading list 📚

The content and copyright wars continue...

• Universal Music Group is suing AI company Anthropic for distributing copyrighted song lyrics.

• Mike Huckabee and fellow authors are suing AI giants for allegedly pirating their books to train AI models.

• Reddit's denying reports of walling off content, but is negotiating with AI companies for data usage.

Big tech keeps marching forward…

• Apple's job listings hint at infusing generative AI into multiple products.

• OpenAI is expanding access to DALL-E 3, with safety measures to dodge past controversies.

• OpenAI won’t tell us the release date of its AI-generated image detector.

And if you’re really nerdy…

• DeepMind's ethics are under the microscope - can we trust them? Read this fairly opinionated piece on them.

That's all!

We'll see you again on Thursday. Thoughts, feedback and questions are much appreciated - respond here or shoot us a note at [email protected].

Cheers,

🪄 The AI Exchange Team