Welcome (back) to AI Minds, a newsletter about the brainy and sometimes zany world of AI, brought to you by the Deepgram editorial team.

In this edition:

• 🎥 Grok 3 Strikes: Is it the new AI King?

• 🔍 Using Deep Learning to Detect AI Images

• 🔊 VocalCrypy: A Novel Active Defense Against Deepfake Voices

• 🗣️How Synthetic Voice Works - Voice Cloning and more

• 🚗 The 7 Best arXiv Papers to Learn how LLMs Work

• 🐦 Social Media Buzz: Training AI on Zuckerberg’s Facebook Posts (and more!)

• 📲 Three new, trending AI Apps for you!

• 🎙️ AI Minds Podcast with Vaibhav Saxena, Co-Founder at Infer

• 🔭 Steve Jobs’ Vision of AI (1983 Talk)

• 🍷 Why Can’t ChatGPT Draw a Full Glass of Wine?

• 🐇 The tests that tricked GPT-4o

• 📚 Glossary Entry: All You Need to Know about Augmented Intelligence

Thanks for letting us crash your inbox; let’s party. 🎉

🎥 Is Elon’s Grok 3 the new AI King?

Last week’s edition came out a few hours after Grok 3. That is, we weren’t able to give it a proper review, since the world was still testing it out by the time readers received our newsletter. However, the world has now had Grok 3 in its hands for seven full days. Here’s a brief review from none other than the legendary YouTuber, Fireship.

Description: “Take a first look at Elon Musk's Grok 3 deep thinking AI model. Let's find out how Grok 3 compares to OpenAI o3, Google Gemini, and Claude.”

🔍  Using Deep Learning to Detect AI Images and Prevent AI Voice Deepfakes

DejAIvu: Identifying and Explaining AI Art on the Web in Real-Time with Saliency Maps - This paper introduces DejAIvu. Using an ONNX-optimized deep learning model, DejAIvu automatically analyzes images on websites, identifies AI-generated content using model inference, and overlays a saliency heatmap to highlight AI-related artifacts.

VocalCrypt: Novel Active Defense Against Deepfake Voice Based on Masking Effect - The authors of this paper propose a novel active defense method, VocalCrypt, which embeds pseudo-timbre (jamming information) based on SFS into audio segments that are imperceptible to the human ear, thereby forming systematic fragments to prevent voice cloning.

⚡ Exposing Voice Cloning and the 7 best arXiv papers to learn how LLMs work

Exposing Voice Cloning: How Synthetic Voices Shape Futures - In this article, we will delve into the fascinating world of voice cloning, exploring the cutting-edge techniques and model architectures that make it possible.

The 7 best arXiv papers to learn how LLMs work - These papers cover various aspects of LLMs, including their architectures, pre-training methods, scaling properties, few-shot learning capabilities, and applications in tasks like reasoning and transfer learning.

🐝 Social Media Buzz: Grok 3 as a Coding Assistant and training an AI on Zuckerberg’s Facebook Posts

📲 Three new, trending AI Apps for you!

SongR is an AI-powered songwriting assistant that helps you create original songs in various genres. It allows you to enter lyrics and then generates accompanying music to match the lyrics and genre you select.

Free Use Score analyzes customer service calls using AI to provide actionable insights in just 15 minutes. The platform simplifies call evaluation into three steps: uploading recordings, AI-driven analysis, and receiving detailed reports.

Hello History enables realistic AI-driven conversations with over 400 historical figures. The app serves educational purposes by providing interactive tools for teachers and students. Users can explore diverse topics like science and philosophy through engaging dialogues.

🎤 The AI Minds Podcast

Vaibhav Saxena, Co-Founder at Infer, is our first guest to make a second appearance on the AI Minds Podcast. Infer are a company which accelerates revenue growth with AI voice bots automating phone conversations and executing tasks like humans for insurance and lending.

Vaibhav explains how tailoring voice agents to user preferences—such as tone, pitch, and cadence—can significantly enhance engagement and conversion rates.

🤖 Bonus Bits and Bytes

If you’ve scrolled down this far, here’s some extra bonus content for you!

• 🧠 Steve Jobs Vision of AI - Prepping for AI Bootcamp for UX Teams, the people at STRAT came across this 5 min video that proves that Steve Jobs envisioned AI experiences in 1983 that we just now beginning to realize in tangible ways.

• 🎞️ Why Can’t ChatGPT Draw a Full Glass of Wine? - While the answer to this question may be obvious to those of us who work in AI, it’s quite interesting to see ChatGPT’s shortcomings in action.

• 🐇 The Tests that Tricked GPT-4o - While GPT-4o excels in some areas, it continues to face challenges in others. However, these are not just hurdles but opportunities.

• 📚 Glossary Entry: Augmented Intelligence - This article dives deep into the realm of augmented intelligence, offering a comprehensive exploration of its foundational principles, practical applications, and its profound impact on enhancing human decision-making and productivity.