A new, upgraded version of ChatGPT has arrived!

- Simplified AI -
Your gateway to discovering the world of Artificial Intelligence!

To celebrate the unofficial kickoff of summer next week, we're stepping away from the usual Simplified AI newsletter to introduce our brand-new edition: Amplified AI - a monthly deep dive into a singular AI topic. We'll explore use-cases and real-world application of AI, answer burning AI questions, incorporate expert research and insights, and make predictions to ignite conversations on how AI can revolutionize both business and daily life.

Moving forward, this will bring us to 5 newsletter releases per month - a weekly Simplified AI, and an Amplified AI at the end of each month.

A sneak peek of what’s below:

• AI terminology refresher

• Companies like Slack are taking your info and training AI with it

• The new GPT-4o is here - watch what it can do

• AI Pictures - bloopers!

Pre-Summer AI refresher!

Turning AI topics into a walk in the park, not a mind-boggling maze

We've covered a lot of terms in the past editions of Simplified AI. With summer coming up, I thought it would be a good idea to get a quick, easy refresher on some top AI terms. Here’s 5 crucial terms to be familiar with:

Artificial Intelligence:

• Definition: Technology that enables computers and machines to simulate human intelligence and problem-solving capabilities. This is a very broad term that encompasses many other specific areas of the field.

• Something to Consider: While Generative AI (GenAI), such as tools like ChatGPT, aims to create content like text and images by learning patterns from existing data, there is also Artificial General Intelligence (AGI), which represents the future state that some large companies are striving to achieve. This involves building an intelligent AI model capable of rivaling human thought, rather than merely "simulating" intelligence.

• Read more: Previous Newsletter

Algorithm:

• Definition: A step-by-step procedure or set of rules meticulously crafted to execute a specific task or resolve a particular problem efficiently.

• Something to Consider: Algorithms are the foundation of most computer programs and AI systems, but their effectiveness heavily depends on the quality of the data they process and the logic they follow, which can lead to biased or inaccurate results if not carefully designed and tested.

Large Language Models (LLMs):

• Definition: Advanced AI systems that have been trained on vast amounts of text data to generate and manipulate human language, allowing them to perform tasks such as translation, summarization, and conversational responses.

• Something to Consider: While LLMs can produce human-like text and perform impressive language-related tasks, they do not possess true understanding and can sometimes generate plausible-sounding but incorrect or biased information (aka hallucinations).

Machine Learning (ML):

• Definition: A subset of AI that involves training algorithms on data so they can learn patterns and make decisions or predictions without being explicitly programmed for specific tasks.

• Something to Consider: ML models can improve over time with more data and better algorithms, but it's not a perfect process. They can suffer from issues like "overfitting", where they perform well on training data but poorly on new, unseen data.

• Read more: Previous Newsletter

Natural Language Processing (NLP):

• Definition: A field of AI that focuses on the interaction between computers and humans through natural language, enabling machines to interpret and respond to human language in a meaningful way.

• Something to Consider: While NLP enables machines to process and generate human language, it can struggle with understanding context, nuance, and ambiguity, leading to potential misunderstandings or inaccurate responses in conversations.

• Read more: Previous Newsletter

AI News Flash

Catch up on the latest AI news from the week!

• OpenAI unveiled and released GPT-4o - more thoughts on this below in the AI spotlight!

• Whoops…Slack failed to tell you that they're mining your conversations to train AI models

• Following several months of AI insight forums, four senators issued a "roadmap" outlining government-funded AI research and regulation

• Tune In: Microsoft's "Build 2024" event is scheduled this week on May 21-23, showcasing the future of AI integration into their tools and hardware

AI Spotlight

Explore how AI is actively shaping our digital future

OpenAI shared a live stream earlier this week to debut an upgraded, more powerful version of their ChatGPT-4 model, called ChatGPT-4o. The 'o' stands for "omni," referencing the Latin word, indicating a focus on this new model's ability to accept prompt inputs that are a mixture of text, audio and images.

While some individuals online expressed disappointment that we didn't get GPT-5 (or whatever the next model will be called), I view the release of GPT-4o as a huge win. Although upgrades were made to the model itself to generate more accurate responses, I believe this release primarily focused on enhancing the user experience during interactions with ChatGPT and expanding the user base.

The live demo showcased a continuous conversation between the OpenAI team and their chatbot, covering a variety of topics and questions, and it appeared to be a pretty seamless experience. One ongoing issue with AI interaction has been slow latency speeds - where you ask a question and have to wait for the response as it "thinks," and even though it's only an additional second or two, it can make the interaction feel less human. However, in the demo, the responses seemed snappy, highlighting how effectively a voice experience could function for those who prefer not to type in a prompt.

Why does this release matter?

An AI tool won't garner widespread adoption unless it's user-friendly, and OpenAI seems to be keenly aware of this. Alongside enhancements to voice interaction and updates to their app, the company also unveiled a new desktop app available for Mac (with a Windows version coming later this year.)

Perhaps the most significant aspect of the announcement is that OpenAI is making their GPT-4o model available for FREE. Previously, access to GPT-4, their more powerful model, was restricted to paid subscriptions, while the GPT-3.5 version remained free. Now, for a monthly fee of $20, users will enjoy five times higher message limits.

This move, coupled with their earlier decision to allow app and website usage without requiring someone to make an account, underscores their commitment to expanding their user base.

What are some examples of what this new model can do now?

The links below will bring you to the OpenAI YouTube videos that showcase what the new model can do:

Real time translation - Traveling in countries where you don't speak the language just got a lot easier with the low-latency speech capabilities of GPT-4o.

Math Tutor - Share your screen and let GPT-4o assist you in solving and understanding math problems.

Coding assistant - Utilize the desktop app to copy and paste code for insights on how to enhance it.

However, it's still essential to acknowledge that like all AI, there's still a chance of occasional errors with GPT-4o. I love OpenAI's transparency here as they openly admit this reality - they shared a blooper reel, showcasing instances where the chatbot made mistakes.

AI (Blooper) Pictures of the Week

Since OpenAI shared their AI bloopers, I decided to follow that trend and share some of my own!

Over the past few weeks I've demonstrated the remarkable realism of AI-generated images. However, creating an AI image that convincingly deceives the human eye is more challenging than you might assume. AI often makes mistakes, requiring numerous iterations to achieve what you're looking for.

When you try to create an image of someone running across a bridge, but instead you get this descendent of Jesus running on water. Hope she brought another pair of socks with her.

When you ask for a painting of gladiators fending off tigers with swords…but instead you're treated to the best boxing match of the 1st century. He must have pioneered southPAW stance.

New wedding trend incoming where grooms wear a piece of their bride's outfit?

I asked for a small pile of wood on the side of the house. Whoops. At least you know they're not running out of firewood anytime soon.