ChatGPT Rolls Out New Features!

Welcome back 😅!

ChatGPT has been all over the headlines recently, and deservedly so. OpenAI has released a bunch of feature updates and much-awaited usability tweaks to its premier AI-powered application.

And that’s not even half of it. NVIDIA has also captured the spotlight with plans to release its next-gen chip. What’s more, we’ve even taken the liberty of testing a pretty popular AI detector to ascertain whether these tools actually work. Suffice it to say, we have a bunch to share with you.

As usual, here’s a quick run-through of what we have in store. Enjoy 😉!

In this issue, we’ve got:

🤖 ChatGPT rolls out new features!

🏇 NVIDIA takes another stride with its new Superchip.

🤔 Experiments from the Neural Frontier: Do AI Detectors Really Work?

⚡ The Neural Frontier’s weekly picks: 3 AI tools you should be using.

🤖 ChatGPT rolls out new features!

As specified on OpenAI’s help page, the company recently rolled out a couple of much-needed updates. As far as we can see, these changes aim to bring in heavily requested updates and improve the overall user experience. Here’s a quick breakdown:

• ChatGPT users on the free plan can now set custom instructions to customize their experience and interactions with the tool.

• ChatGPT now suggests responses to help you continue your conversation with the AI.

• To help you ease into the process, ChatGPT now provides suggested prompts at the beginning of a new chat.

• ChatGPT now remembers the last model you used (GPT 3.5 or GPT-4) and will default to that whenever you start a new chat.

• ChatGPT can now interpret, analyze, and provide insights across multiple uploaded files.

• You will no longer be logged out every 2 weeks, as ChatGPT now allows you to stay logged in for longer.

• Shortcuts are finally here! Try Ctrl (⌘) + / to see a list of nifty shortcuts to try out.

🏇 NVIDIA takes another stride with its new Superchip.

At this point, the name “NVIDIA” is almost synonymous with AI technology advancements. And it’s well deserved too! Seeing as the company’s superchips are a crucial component in training LLMs—thereby pushing the boundaries of AI tech—it’s not surprising to see them making headlines once again.

As stated during the annual SIGGRAPH conference, the company plans to roll out an update to its GH200 Grace Hopper platform. Additionally, next-generation HBM3e memory technology will power this chip. To put it succinctly, Nvidia founder and CEO Jensen Huang stated:

“We announced Grace Hopper recently, several months ago, and today we’re announcing that we’re going to give it a boost.”

According to Nvidia, this next-gen HBM3e memory is about 50% faster than its counterpart in the current-gen GH200 chip. The company also stated that this boost would allow its “superchip” to run AI models 3.5 times faster than the current model—potentially revolutionizing the AI ecosystem.

🤔 Experiments from the Neural Frontier: Do AI Detectors Really Work?

This week, we decided to finally answer the question: do AI detectors really work? With so much controversy surrounding the topic, we had to test it out for ourselves. Here’s how we went about it:

We took the introduction of this newsletter, and passed it to ChatGPT (the GPT-4 model) to rewrite the text and add some unique elements. Here are the results:

Upon obtaining these results, we passed them to our AI detector for testing purposes. Primarily, we utilized AI Detector Pro for this experiment. Here’s what we got:

We also tested these results using Copyleaks’ AI detector and got a similar verdict.

Now, we have to admit that with the use of emojis and other storytelling elements, even a human would find it hard to detect that the above was AI content. However, it appears that most AI detectors struggle to identify AI content in cases where the AI tool has been tasked with rewriting human-generated content.

To confirm this, we tested another AI-generated introduction—this time, with no human text given to the AI to rewrite. This was the prompt and the corresponding result:

As usual, we plugged these results back into AI Detector Pro and confirmed our hypothesis!

Since we tasked GPT-4 with generating content from scratch, it was much easier for AI Detector Pro to identify the results as AI-generated content.

Unfortunately, though Copyleak’s AI detector suspected a 47.1% probability of human content, its results were still pretty far off.

And there you have it! Our first AI-centric experiment. As you can see, some AI detectors🤖 perform pretty well when it comes to identifying AI-generated content from scratch.

However, if there’s significant human input as regards the prompting process, these AI detectors begin to struggle 😓.

⚡ The Neural Frontier’s weekly picks: 3 AI tools you should be using.

True to our promise of equipping you with tools and insights to surf🏄 the wavy world of AI, we’ve curated another lineup of AI-centric productivity boosters. Have a go at some of these tools, and watch them transform your workflow!

1. 🧑‍💻Flo: Calling all programmers! With a promise to “never get stuck on errors again” this AI-powered command line interface tool catches errors in your code and shows you how to resolve them after scanning your codebase.
    

2. 🤖AI Detector Pro (sponsored): From our experiment above, you should be able to deduce what this tool does. Offering itself as a tool that helps you “find and wipe the traces of AI,” this tool performs pretty well at identifying AI content generated from scratch.
   
   Interested? Try AI Detector Pro today!
   
   

3. 🔮 AI Hubs: What’s better than one AI tool? How about a repository containing 5000+ AI-powered tools for you to explore? That’s exactly what AI Hubs delivers to your doorstep. So, if you’re an explorer, who likes testing random tools, you’ll find a home here.

That’s been it! We poured our hearts into this piece and truly hope you learned a thing or two. Remember, if it resonated with you, it might just make someone else's day too. So go ahead, spread the joy, and hit that subscribe button if you're not on board yet.

Can't wait to see you next week! 🚀🌟