LLMs in 2024

What’s going on in the world of AI? It’s obviously a heavy question to answer, but here are some interesting trends to look out for:

1/ The rise of non-transformer models like Mamba?

What, no way- transformers aren’t the only thing happening in the world of AI?? This year, they’ve been the star of the show, but a paper called “Mamba: Linear Time Sequence Modeling with Selective State Spaces” shows there’s action outside of transformers like GPT.

Mamba’s been picking up momentum in the AI community– at the core of it, it can scale impressively to lengthy sequences. 

Now, why is this fascinating?

To know the context- let’s first look at State space models, which have promising properties and can capture long range dependencies very effectively, and are more memory efficient than transformer models like GPT.

Anyways…all that to say….AI mamba is coming in big time!

2/ AI Companies will focus more on UX

The infrastructure layer of AI was all the rage in 2023, and it will continue to be. However, now there will be a shift and more prominent focus on verticalized, AI companies that solve issues for a specific vertical. 

Verticalized AI companies would include things such as companies that apply AI to a specific niche- such as finance, education, robotics, or construction.

The thing is, vertical specific AI is the path forward because the big companies basically have an oligopoly over the infrastructure layer of AI- Google, OpenAI/Microsoft and others cover the infrastructure layer of AI, so you do have UX focused companies like Midjourney, Harvey, and others covering specific verticals and experiences like images and law.

3/ Do AI models know when they are hallucinating?

The thing is, this shows that probing is an effective method to evaluate hallucinations in LLMs. Good to know for quality check purposes!

To summarize- yep they know they’re hallucinating.

4/ Giga ML is helping enterprises adopt generative AI.

It’s the horseshoe effect- software started on-prem, moved to the cloud, and with some of the latest AI tech, on-premise seems to be the option of choice for many companies. 

5/ Midjourney hits $200 Million in Revenue with $0 in fundraising

Incredible outcome. This stat speaks for itself…One thing’s for sure- AI helps teams stay lean and produce more output.

6/ The rise of GenAI developer platforms

Remember what Canva did to design? The same is happening across the AI development stack and we’re here for it.

Big things are coming for sure.

Happy New Year, and peace out