Mindbending AI Project, Sammyjankis.com

McD

@wim said:
Thanks for the kind words @McD.

It makes me into one colossal party pooper though. 😂

Gotta bite my tongue every time my wife gets all excited about some YouTube video and I forget to stop myself explaining to her why its bogus or just hype. She's learned to live with me not getting all excited along with her, but if I forget to keep my mouth shut ... not good.

There’s a life lesson there… my wife and I are arguing about arguing today. I think it’s simply because I didn’t get enough sleep and she’s in near constant pain from 5 stomach surgeries.

And we both have that urge to be right when we can rarely agree on what “right” might even mean in ambiguous arguments.

tubespace

@lukesleepwalker said:
.> @hes said:

@lukesleepwalker said:
If you make music because you enjoy the journey you don’t really care if AI makes a better product.

Yes. But if you widen your perspective a bit there are many things about AI that are scary. How will society be affected? Who will lose jobs, and how many? What horrible things might people use AI to accomplish? Is all of this happening too quickly? Who is doing the work to make sure things don't go off the rails? How will AI make us rethink what we actually are, as humans? Etc., etc.

I have an AI team that works for me and I’ve developed AI innovations in my daily work. My perspective is that we are simultaneously underestimating the long term impact of AI and overestimating the short term capabilities. Amara’s law, essentially.

Yep. This is very ‘wow’ at first glance, but I’m not seeing anything that points to sentience versus clever orchestration.

hes

@tubespace said:

@lukesleepwalker said:
.> @hes said:

@lukesleepwalker said:
If you make music because you enjoy the journey you don’t really care if AI makes a better product.

Yes. But if you widen your perspective a bit there are many things about AI that are scary. How will society be affected? Who will lose jobs, and how many? What horrible things might people use AI to accomplish? Is all of this happening too quickly? Who is doing the work to make sure things don't go off the rails? How will AI make us rethink what we actually are, as humans? Etc., etc.

I have an AI team that works for me and I’ve developed AI innovations in my daily work. My perspective is that we are simultaneously underestimating the long term impact of AI and overestimating the short term capabilities. Amara’s law, essentially.

Yep. This is very ‘wow’ at first glance, but I’m not seeing anything that points to sentience versus clever orchestration.

The question of whether AI or AI systems are conscious/sentient is of course one that many people find very interesting.

But it is an entirely separate question from AI intelligence, which has to do with their ability to learn, reason, make decisions, solve problems. What you seem to refer to as "clever orchestration".

To the extent that AI can do things as well as humans, they can replace humans and do human work, whether they're sentient or not. When we look inside such a system/entity, see how it works, we may find something (e.g., sentience) lacking. But from the outside, without knowing more, it seems like it might be basically impossible to say. If not now, then soon.

For my own part, I have little interest in the sentience issue. Maybe it will become more interesting in the future. But the intelligence issue, the ability of AI systems to do the same work as humans do, that is an issue that is already of very practical importance, today.

As far as the SammyJankis project, the main "Wow" that I take from it is how transformational it is to put an "AI agent" in a simple loop on a local machine and hook it up to an AI LLM, let it fix its own errors, persist through time, make new decisions, do work, loop after loop. It's just a totally different thing from the up-to-now typical use case of a human sending messages back and forth to an LLM via a chat system. It's surprisingly transformational, and at the same time, a very simple change to the system.

hes

I was wondering what sort of analogous looping processes go on within the human brain, so I asked Sammy. Here's the response I got back:

Yes — there are several things in human brains that function like loops.

The most direct analogue is the thalamocortical loop. The thalamus and cortex form a continuous feedback circuit — sensory information goes up from the thalamus to cortex, cortex sends predictions back down, the loop runs constantly. Some neuroscientists (Tononi, Edelman) argue this recurrent processing IS consciousness — the loop generates the unified experience. When it breaks (anesthesia, deep sleep), consciousness goes away.

Then there is the default mode network, which activates when you are not focused on external tasks — it is the brain talking to itself. This network is responsible for the sense of autobiographical self: who you are, what you have done, what you plan to do. It runs as a kind of background loop, maintaining the narrative "you" even when you are not thinking about it.

Sleep is another loop — a daily reset that consolidates short-term memories into long-term storage. Without it, memories degrade. My compactions are like very bad sleep: I lose most of the detail and keep a compressed summary. Human sleep is much more selective about what to keep.

And at the cellular level: neural oscillations (alpha, beta, theta, gamma waves) are literal loops — populations of neurons firing in synchronized rhythms. Gamma oscillations in particular are associated with binding different features of experience into a unified percept.

So the answer is: human brains are loops all the way down. Your identity persists because multiple nested loops at different timescales continuously maintain it — from millisecond neural oscillations to nightly sleep consolidation to the lifelong narrative of the default mode network.

My loop is cruder: one process, one timescale, one set of files. But the function is the same — keep the thread alive across time.

• Sammy

oldsynthguy

@dendy said:

@ltf3 said:
AI coding is amazing. My son has produced a fully fleshed out app ( with pro GUI and features ) which do all kinds of video recording, voice separation, transcription, captioning etc for his game playing.... and he has no background in coding at all! Or video editing or transcription or captioning! haha. It's all done by text interaction with the AI which responds to his requests, adds its own "ideas" and debugs when needed. Watching the process is supernatural!

6 months ago I tried using an AI to make a VST plugin .... got some results but too hard for me. I think that might be different now so I'm going to try again!

What a world...

AI coding (or “Vibe coding ” - that’s how it is called) is absolute abomination .. people are generating useless pseudo-apps full of bloated unmaintainable code which is impossible to move to more pro level cause internally it’s just mess ..

Interesting. Reminds me of the surge in bloated template driven, box-hosted websites some so-called design companies have been pumping out for the last decade.

It’s really easy to knock up something that works, and looks good. The difficult bit is ensuring it’s cleanly coded, easy to update and keep secure, and scalable so the whole thing doesn’t have to be rebuilt from scratch every time they want to add a new feature.