Forum Replies Created
-
Posts
-
So what are you’re thoughts on those?
I think only you can effectively analyze the meaning of your own dreams.
The specifics of the dream in my writeup weren’t important, and I handwaved the finer details purposefully. It was a trigger, a launching pad of sorts that made me contemplate how dreams contribute to memory, what their potential purpose is as an evolutionary adaptation. It did so because my brain understood that suppressed movement was non-ideal, so it made me watch the action play out instead of performing the actions first-person.
If dreams are a memory consolidation system and mental model testing ground, what does that say about our efforts to build an intelligence? What more than inert knowledge-base do we need to construct to take those next steps?
It’s not about what happened in the dream, but why the dreaming world exists at all, how it adapts and incorporates current and past memories, how lucid dreaming, daydreaming, and dozing in that in-between space may play into everything.
The contents of dreams vary wildly from person to person and so hold no interest to me as they are intensely personal and mean little to outsiders, but the why? I think it’s hallucinations much like what we see from Stable Diffusion when it’s just mashing things together to see what works. If we’re essentially organic AIs, we can repeat the process.
Maybe we already have but haven’t realized it yet.
Thanks @fiesty!
Ah, yeah. Telegram is also a great resource. The link you have there seems to redirect to a private group, so I can’t join.
Hi All,
My name is Shaun, and I’m a Postgres guru / zealot who’s recently also been dabbling in AI. I’ve been working in IT since 1999, and keep an overloaded Dell R730 in my basement for various experiments. I’ve run the gamut from dev to DBA, conference speaker, author, blogger, and the usual litany. Really, I’m just a tinkerer.
Given how the pgvector Postgres extension works, I’ve seen projects where sections documentation are transformed into embeddings and then searched by comparing a vectorized question into a similarity for near matches and then fed into a RAG for further feedback. I’m thinking of using a similar technique on my own local files and other data, so I’m always scouring everywhere for Python API examples before I start breaking things. (If you have one, feel free to share!)
My current favorite local LLM is OpenChat-3.5-0106, but I’ll admit I haven’t tried Mixtral-8x7B yet. It’s crazy how fast all of this is currently moving.
Anyway, I’ll be around!
Edit: Is there a Slack, Discord, or some other more interactive medium for members? Old PHPBB forums are nice, but they’re excruciatingly slow and not exactly conducive to active collaboration.
