🤔 w/145: What Exactly Is a Large Language Model?

⚡️ Supercharge your career, productivity and wisdom with Wiser! - the tech newsletter for professionals who want to know more than their competition.

Cover Image: AI generated with Magic Media by Canva.

w/Wiser! #145 - 29th October 2023

In this issue:

• What exactly do we mean by large language models and “generative AI”?
• Midjourney’s new website fixes a barrier to the user experience
• The risks and safety concerns for AI continue to dominate headlines

Plus tips, tools and insights to help you supercharge your career with AI.

w/Insight

What Do We Mean By Large Language Models And Generative AI?

For decades, artificial intelligence has been used to perform specific functions, like self-driving a Tesla or auto-processing an insurance claim. These single purpose AI and machine learning technologies sit silently behind the scenes, just doing their thing, with little awareness from the likes of you and I that they even exist.

Then along came ChatGPT.

And a new generation of artificial intelligence had arrived. This is a new, different and disruptive form of artificial intelligence. It is different to the AI that went before it for two fundamental reasons.

First, it's a more general, multipurpose form of AI and second, anyone can use it however they want, for whatever they want.

Known as “generative AI,” this new breed of artificial intelligence is all about the power of words. Even if it’s a picture, generative AI turns it into words.

The super power of generative AI is that its use of language is on par with human levels of communication (which is not to be confused with human levels of comprehension and understanding.)

Quite simply, generative AI is a statistical prediction machine that’s known as an “LLM” or Large Language Model, because that’s exactly what it is! Every LLM has read millions of pieces of human writing from across the internet, and broken every sentence and word down into their smallest units or fractions of words, for example, “football” becomes “foot” and “ball.”

The AI then creates a massive multi-dimensional map of the way that words are combined to form many many different meanings. Think of how many ways you can use the words “foot” and “ball” in a different context.

But there’s more to generative AI than just building a mega spreadsheet of word combinations. What unlocked their abilities to write as fluently as they do today is a revolutionary AI tool called “the Transformer.”

The Transformer model was first published by a group of eight AI researchers at Google in June 2017. Before Transformer, AI development focused on trying to replicate the way the human brain works through neural networks.

But the Transformer model changed everything.

➜ Here’s The Thing: Google’s 11-page research paper in 2017 marked the start of the generative AI era. Without it, there would be no ChatGPT!

Transformers are the ability of an AI system to process an entire sequence of words, analysing all its parts, breaking it down and “understanding” the meaning of the words. By that I mean that the Transformer is able to work out which parts of a sentence or article are the most important in defining the meaning of it.

A key concept of the Transformer architecture is called “self-attention”. This is what allows LLMs to understand the relationships between words. Self-attention looks at each word, or token, in a body of text and decides how they all relate to each other and which are most important to understanding its meaning.

I asked Anthropic’s Claude to explain how this works using a simple example “the quick brown fox jumps over the lazy dog.” Claude wrote:

This sentence would first be broken down into individual tokens, or words. So the model would see: [“The”, “quick”, “brown”, “fox”, “jumps”, “over”, “the”, “lazy”, “dog”] The model then uses self-attention to understand how these words relate to each other. It creates numerical vectors for each word, and maps out their relationships based on context clues. So it would map that “fox” and “jumps” go together, as do “dog” and “lazy”. With this vector map of word relationships, the model can then start generating new sentences based on its understanding. It might take the core of “fox jumps over dog” and create a new sentence like: “The playful kitten leapfrogged over the napping puppy.”. The model is able to mix words around contextually and correctly due to building up its knowledge from the vector mappings created by self-attention. It can continuously improve at natural language generation the more examples it has to learn from.

I then asked Claude to write a new sentence based on this example. Claude wrote:

Here is a new sentence generated using the same example: ”The speedy silver fox bounded across the dozing hound.” The model used its vector map created from the first sentence to understand relationships between words like "fox" and action verbs, as well as animals and descriptors. It was able to mix and match words contextually correctly to form a new plausible sentence, by drawing on its understanding of how certain words and concepts are likely to be used together based on the examples in its training data. This demonstrates how large language models are able to take simple word associations and build up a complex internal representation of language.

The highlighted text above were written by the chatbot Claude using the LLM from Anthropic called Claude 2. With no editing by me. Can you tell the difference between the words that I wrote and the AI’s?

➜ If you’re just getting started with generative AI, subscribe to this newsletter and download The Beginner’s Guide To ChatGPT for FREE!

Further Reading ➜ The Rise of Generative AI: Exploring the Safer Alternative to ChatGPT with Claude by Anthropic

Join 15k Subscribers and Get Wiser! Every Week

➜ Subscribe for free here.

w/News

What Happened In AI This Week?

Midjourney Finally Fixes Its User Interface With New Website

If you’re not familiar with Midjourney it’s probably because it’s bloody difficult to use. As a generative AI platform that turns words into pictures, it’s by far the best. And whilst all the other text to image generators are getting better, Midjourney remains number 1, IMHO.

But, the barrier to using the AI is that it runs as a Discord server using a command language. What the heck does that mean, Rick? It’s too hard to explain in a sentence. All I will say is that it took me the best part of an entire Sunday to figure it out and start producing images.

But now, thank goodness, the independent research team that run Midjourney have built a website that is more user friendly. It costs $8/month for a basic plan, rising to $60/month for the all you can eat fast food option.

Some Text-To-Image Generators ➜ Midjourney | Magic Media | DALL-E2 | Adobe Firefly | Stable Diffusion via Canva

Questions Of AI Safety And Risks Continue To Dominate

British Prime Minister Rishi Sunak has expressed concerns about the risks of AI, such as its misuse in weaponry and criminal activities, and the loss of control over super-intelligent AI. To address these concerns, Sunak plans to lead global discussions on AI safety and has announced the creation of an AI safety institute and proposed a global expert panel on AI science.

Meanwhile, US President Joe Biden is expected to announce a sweeping executive order on AI next week in an effort to place controls on this fast moving sector. It’s well intentioned but, as with all tech regs, the horse has already bolted.

Meanwhile, ChatGPT’s founder OpenAI has formed a new 'Preparedness' team led by Aleksander Madry, the former director of MIT's Center for Deployable Machine Learning, to address potential catastrophic risks associated with advanced AI models, including nuclear, chemical, and biological warfare. The team will track and guard against these dangers and is seeking ideas for risk studies, offering a prize of $25,000 and a position within the team for the top ten entries.

Sources: CNBC | Mashable | OpenAI

➜ Footnote: In May, Scale AI became the first AI company to have a large language model deployed on a classified network after it signed a deal with the US Army’s XVIII Airborne Corps. Scale’s chatbot, known as “Donovan,” summarises intelligence and helps commanders make decisions faster, according to the Washington Post.

Exploring Amazon’s Humanoid AI Robots

Amazon is testing humanoid robots designed to carry products around its sprawling sorting facilities. They’ve been given the nick name of “Digits” because these robots have arms and legs complete with hand-like manipulators. This comes on top of a decade of robotisation for Amazon, who had over 750k of them at last count.

Amazon’s humanoid robots are designed to mimic the human form, allowing them to navigate and interact with their surroundings in a way that is familiar to humans. They typically stand at an average height, with a humanoid body structure, complete with arms and legs. Their heads are equipped with sensors and cameras to aid in perception, while their bodies are made from durable materials that can withstand the rigours of a demanding work environment.

➜ See Them In Action.

Five Snippets Of Tech News

1. Forbes has introduced a generative AI search platform called Adelaide. This verticalised AI will provide personalised searches and summarised answers from Forbes' extensive catalogue of writing and news coverage.
2. Grammerly has added a new AI feature that learns a writer’s “voice.” However, there’s a concern that a writer’s unique style could be used without their consent.
3. Canva has introduced a comprehensive set of AI tools for teachers and students. The tools include safety controls for educators to manage access levels. ➜ The future of AI in education was the subject of a recent episode of Big Tech Little Tech.
4. Over 40 states in the US are suing Facebook and Instagram parent company Meta, alleging their platforms are harmful and compound mental health problems for children. The lawsuit claims Meta have been “manipulative and exploitative”.
5. Google Shopping is extending its use of augmented reality beauty tools. Customers can try colours from L’Oréal brands and shoppers will be able to virtually try on foundation, lip and eye makeup themselves using just their mobile phone.

➜ Footnote: It’s a year since Musk bought Twitter and there are a ton of stories about the mess he’s made of it. I’m bored of it all now but I did look back to my 2023 predictions, where I wrote in December last year: “When Musk took over he tried to fix the wrong the problem. It wasn’t about free speech. Twitter’s problem was its business strategy. The business model was broken, innovation was stagnant and user numbers weren’t growing. Musk has worsened the economic model and failed to reinvigorate one of social media’s pioneers.” It’s been all downhill since then!

w/Productivity

DataGPT Launches “AI Analyst”, To Provide Data Insights for Enterprises

A California-based startup called DataGPT has unveiled its AI Analyst, a conversational chatbot for businesses to gain actionable insights from their data. Say what?

In a nutshell, DataGPT is like having your very own data analyst, a geek who will pour over spreadsheets and charts to look for any kind of insights from inside the numbers. This is looking for the wood amongst the trees and essential to any business.

Bottom line: DataGPT enables businesses to understand the what and the why behind their data and is an alternative to ChatGPT’s Code Interpreter.

For more, read this ➜ VentureBeat or watch this ➜ DataGPT Demo

The State Of AI Report 2023

🚀 THIS IS GREAT READING ➜ This State Of AI Report from Air Street Capital is packed with data, insights and analysis. But be warned, it’s a detailed report. My advice; brew a big pot of fresh coffee and grab a plate of digestives because you're in for detailed breakdown of what's going in generative AI.

Don't expect any soundbites or over-simplified 280 character breakdowns, this is heavy reading, and well worth it! ➜ Find it here.

🙏 Show Your Support For Wiser!

Thank you for reading Wiser! If you got value and would like to support what I’m doing, do this:

• Share this newsletter: post it on your network.
• Make a donation: go to BuyMeACoffee to make a donation in the form of a virtual cup of coffee, they only cost €2 each. And who doesn’t love a coffee, right?
• Check out my website: You’ll find links to everything I’m working on at rickhuckstep.com.