🤖 Apple GPT to fuel Siri & iPhones

PLUS: How to use Meta’s Llama-2?

Author

The Intelligo AI
July 20, 2023

Hello Intelligos!

We’ve got some quite impactful headlines in the AI world today, with key developments from Apple and Meta, a study on ChatGPT’s performance, and other news from OpenAI and NVIDIA.

☕️ Grab your coffee and let's get started.

Here is what we are covering today:

  • 🤖 Apple GPT to fuel Siri & iPhones

  • 🔻 ChatGPT is losing quality over time

  • 🤖 How to use Meta’s open-source ChatGPT competitor?

  • 🏆 5 useful AI Tools

  • 🗞️ 2 AI stories from ChatGPT and NVIDIA

  • 💡 AI Tutorial: Convert YouTube Videos to Blogs & Audios with ChatGPT

Read time: 4 minutes

Let’s get to it! 👇🏼

🗞️ FRESH OFF THE PRESS

APPLE

Summary: According to an Bloomberg, Apple is quietly building its own AI chatbot, also known as "Apple GPT", that could be integrated into Siri & Apple devices.

Key Points:

  • Apple is using its own system, "Ajax", to make the new tool.

  • The chatbot was stopped for a bit because of safety worries, but more Apple employees are getting to use it.

  • They don't seem to be interested in competing with ChatGPT. Instead, Apple wants to find a consumer angle for their AI.

Why it matters? With 1.5 billion active iPhones out there, Apple can change the LLM landscape overnight.

CHATGPT

Summary: Users and researchers are raising alarm bells, saying the quality of GPT-4's responses are dropping off - and there's evidence to back them up.

Key Points:

  • A study by Stanford University suggests that the June version of GPT-4 performs worse than the March version on tasks.

  • Chain-of-Thought, a method to improve AI reasoning, didn't help the latest GPT-4 version.

  • Code generation performance also decreased from 52% to 10% in the same period.

  • The cause isn't clear, but there are theories it might be because OpenAI is using smaller, specialized models that are cheaper to run.

Why it matters? These issues matter because they could mean trouble for people who are making apps with GPT-4. If AI's behavior changes over time, that could spell bad news for reliability and trust in AI systems.

META

Summary: Meta has unveiled Llama 2, an open-source LLM that can be used commercially! Here are the key features and how could you use it

Key Points:

  • In partnership with Microsoft (Azure), and will soon be on AWS, Hugging Face and more.

  • Qualcomm will work with Meta to put Llama 2 into devices starting in 2024.

  • Pre-trained on 2 trillion tokens and has double the context length of LLama.

  • The models are available with 7B, 13B and 70B parameters. (paper)

How to use it?

🏆 AI CHAMPIONS

USEFUL AI TOOLS

  • Folk: A customizable CRM system designed to automate tasks and enhance relationship management for teams.

  • Get Magical: Utilize AI to accelerate repetitive tasks in the workplace.

  • Bardeen: Automate manual workflows with the power of AI.

  • Hify: Create personalized and engaging sales videos directly from your browser.

  • AI Tools Database: Explore a comprehensive Notion database featuring 1350 useful AI tools curated by The Intelligo.

📖 CATCH UP: AI STORIES & READS

ChatGPT Plus subscribers now have an increased messaging limit of 50 messages in three hours with the introduction of GPT-4. Previously, the limit was set at 25 messages in two hours due to computational and cost considerations.

Source: OpenAI

NVIDIA has taken the lead in the AI boom, with its chips in high demand for AI models like ChatGPT. Their special AI H100 chip, priced at $25K, holds a dominant position in the market. Not only are they selling chips, but they are also shaping the future of AI through investments in startups.

SPONSOR US

📢 Partner with The Intelligo

Interested in supporting the next edition? Our newsletter is read by tech professionals, investors, engineers, managers, and business owners globally. Reach out today

💡 AI TUTORIAL

Convert YouTube Videos to Blogs & Audios with ChatGPT

Ever wished you could repurpose your YouTube content into blog posts and audios? In this tutorial, we'll show you how to convert YouTube videos into written and audio content using ChatGPT and a few helpful plugins.

Step 1: Install Necessary Plugins

You'll need three plugins for this task:

  • Video Insights: Extracts key information from videos.

  • ImageSearch: Finds relevant images to enrich your blog post.

  • Speechki: Converts your blog text into voiceover audio.

You can install these plugins from the plugin store.

Step 2: Enter the Prompt

Once you have the plugins installed, paste the following prompt into ChatGPT:

Perform the following tasks based on YouTube video below:

[URL]

1. Take the captions of the video and convert it into a blog

2. Add required images for the blog

3. Create a voiceover for the blog

Replace "[URL]" with the URL of your YouTube video.

Step 3: Get the blog and the voiceover

After entering the prompt, ChatGPT will create a blog post based on the video's content. It will also suggest suitable images from Unsplash and generate a voiceover for the entire blog.

Expected Outcome

The output should be a well-structured blog post, complete with images and a voiceover. This way, you can extend your reach beyond YouTube and cater to audiences who prefer reading or listening to content.

Did this tutorial help you? Have any questions or suggestions? Let us know by replying to this email or reaching out on Twitter @mustafarrag.

We're always looking to improve and provide more valuable content.

🐦️ Tweet of the day

How to use an AI (ChatGPT Code Interpreter) to train your AI

That’s all!

Thank you for reading.

We're eager to hear from you! Have feedback, thoughts, or questions? Just reply to this email or reach out via Twitter: @mustafarrag

Until next time,