So, it seems like the future is here and rapidly unfolding! Ironically I've been struggling to even mention anything about recent machine learning events, here on this blog. Despite AI being such a big feature of its theme on exponential tech changes...

New Scientist style chatbot illustration generated using StableDiffusion

5 days to 1M users! The fastest growing web   service, ever. From Twitter via YouTube.

To recap: from the middle of 2022, there was simmering online interest in the spectacle of AI generated art via text prompts: DALL-E vs Stable Diffusion.

Also, Google's firing of a senior software engineer who was claiming sentience and personhood for their LaMDA chatbot (e.g. Guardian). Which kicked off nervous discussion.

Then, at the end of the year, OpenAI unleashed ChatGPT-3 to the public, making a huge splash in the process! Machine learning has sustained main stream awareness through 2023, so far, with news hype hitting fever pitch in tech circles, over the last week of announcements. E.g. on YouTube LTT wan show from here and Matt Wolfe.

As talked about in many sources, machine learning has been advancing up a much steeper exponential curve than Moore's law, both GPU hardware advancing faster and ever bigger datasets being trained on.

Of course, the main reason I've struggled to keep fully abreast of these exciting developments: I'm still struggling with the ME/CFS. As I've lamented on Twitter, this blocked my desired career path into academic AI/robotics. Being forced to give up on 3rd year studies just over 15 short years ago. And in my small amounts of productive time, during the last year, I've been working on informational resources (e.g. Steam guides) and other things in the Core Keeper community. Which is sometimes within my cognitively impaired abstraction limit.

Anyway, I'm aiming for this post to be ongoing, like the previous Covid posts. Updated with my thoughts and new events as they unfold. Because I don't have the executive function to write a big piece with an overarching narrative. If I've ever really achieved that here before, heh.

I've no hope of capturing any of the soap opera of corporate and big-player drama, or even most of the notable offerings. Let alone the technical details of how the systems opera. So really, this is largely an exercise (as always) in showing that I was here and I maybe understood some fraction of what was going on. Maybe an aide memoire for me, like taking notes in class. 

I'm also explicitly aware that the ever-growing LLM training data with likely include this text at some point, too! Which is very meta and a little thrilling to see happening already.

I'll start by catching up on some backdated bits and pieces...