Well shit… it’s already coming up to the end of Feb. It was a quiet January in the world of AI with no major announcements to the public (outside of a few releases here and there) but February really decided to be the month to kick things off for 2024 but that’s not why we are here today.
Back in December last year I had the opportunity to create a presentation on a topic of my choice for my colleagues, and of course, you know me, I chose AI. In particular, I wanted to cover a basic understanding of the current advancements and also have look back at 2023 to highlight the milestones of the big players and what AI trends I expect to continue into 2024 and beyond.
I would like to take the time to go over this presentation for those who missed out but to also provide my updated thoughts. Yes I know… it’s only been two months, but this blog isn’t called fast paced for nothing (also two months is a very long time for this goldfish of a brain, so there may be sections of this presentation where I genuinely have no idea of what I was trying to get at).
Quick disclaimer: Majority of the content presented below is conjectural to my observations from the past year, so please take everything with a very large grain of salt and if the topic interests you please do your own research!
My apologies for the long winded intro but this post will be a long one, so I’ve broken it up into two parts. Grab a drink, sit back, relax, and welcome to:
Contents:
- Introduction
- What is AI?
- AI Recap 2023 (Covered in Part 2)
- What now? (Covered in Part 2)
Introduction (Come on man… I just sat through one…)
2023 was a wild year when it came to the rise of foundational models especially from the major tech competitors in the scene. One thing you may notice, all these tech competitors gather incredibly vast amounts of data from their user base. I believe this data was the key to the successful explosiveness in training these models to mimic how you and I perceive and interact with the world around us. You may be asking “Evan, what is a foundational model”, please hold your theoretical horses I will cover this more in depth further down.
Update for 2024: We have since now seen the announcement and release of two new versions of foundational models with OpenAI’s GPT4 Turbo and Google’s Gemini Pro 1.5. I will cover these updates in a later post so stay tuned!
Look, I will be honest with you. This is one of those not really well thought out slides… I think I was trying to voice that we are starting to see the applications of this technology. With this in mind, the public start to get the initial taste of how AI can be packaged as a product for consumption. Even as I type this post up I’m presented with an option to have an “AI” assistant help me with typing. It is evident that this helped path the way for the incredible hype (and/or discourse) the public had in 2023 towards AI.
Awoooga indeed (this one is a bit of a shit post, but let’s start normalising shit posting at work). I want to highlight the copious amount of money that is currently being invested and funding AI development, which I believe shows the faith that both mega-corporations and the market have for this technology. This leads me to believe that this isn’t just another fad, looking at you blockchain, and that the few people pulling the strings to this whole operation actually see gold at the end of the tunnel. BUT… we are still a long ways away from said gold so it’ll be interesting to see a. if the hype train slows down, b. how useful the money is vs time, c. how it will be used effectively and d. how greedy people can get *cough cough* Sam Altman being removed from the board and as CEO *cough*. As my man Biggie used to always say:
Update for 2024: OpenAI has just been valued at $80 Billion USD (just a shy $10B of what they were asking for and up 275% compared to Jan 2023) and NVIDIA currently overtaking both Alphabet and Amazon in market cap with a stock price of $694USD as of 21st Feb which is a staggering 38% increase in value from the 18th Dec 2023.
What is AI?
Most of this content is typed out on the slides, so I’ll stay quiet for as much as possible.
AI is already an incredibly vast and fast growing field that I struggle to comprehend on the daily. So let me break down some key topics I’ve covered so far.
Update for 2024: Honestly GenAI sorta died off as a buzzword and became more of a terminology for what foundational models achieve. We might see this word being thrown around less, but only time will tell.
Think of foundational models as a kid in high school, someone that has a base understanding of the world. Users can then take these models and put them through university to create more specialised solution/knowledge to their niche problem. This analogy may be warped over time as foundational models are continued to be developed on.
I would like to try visualising the scale and complexity of what’s going on behind the scenes of these models to help you understand what is truly happening. https://bbycroft.net/llm is a excellent site for demonstrating this fact. It highlights the machine learning techniques (such as transformers, the T in GPT) used initially by OpenAI which helped demonstrate to the world the true potential of machine learning. Even if you don’t understand the computer science behind it all, I urge you to explore this site if you have any inkling of wanting to look behind the curtain to see the reality of what these models are doing.
As OpenAI as well as other companies making foundational models continue to develop on the tech, we start to see new techniques emerge from the field. Mixture of experts is one such technique we are currently seeing mass adoption in the field starting with GPT-4. I will be covering this topic further in depth at a later date.
This wraps up the first half of my presentation. Observe how fast development has been over the last decade, now last 5 years and now 1 year. I want to leave you with a couple questions.
- How fast are we really going here?
- What do you think could be next?
- Now think long term, what will 5 years from now look like? 10 years? in our lifetime?
Stay tuned for part 2!