I recall a basic philosophical issue from a girlfriend who was majoring in philosophy. Either reality can be observed or it can’t. If it can’t there is no point in exploring that, so the alternative not only provides results, but insights into mechanisms.
I see many discussions of nerves acting like wires which they do not. The data is transferred by reversing cascades of sodium and potassium ions. The rate and pattern of those cascades vary a lot but do not cease and when it comes to a synapse the effect is also quite variable and influenced by what is going on at all the other synapses. That is modified and changes what it sends out continuously. Multiply that by a few billion cells and the simplicity and uniform result (either on or off) of digital computers cannot begin to catch the nuances of the way nerves work. One does not have to follow this to the specifics of the end results to see that the differences increase as the complexity increases.
That is why I put in what I did. Now if some basic piece is invented that has the nuances of a nerve cell then the question would need to be revisited, but the basic binary every bit equal to all others but for address cannot get there.
I am of a mind that computional power won’t be a limiting factor. Binary computers are already the past. If you think of quantum computers compared to traditional computers as one single technological advance, in 100+ years there are likely to be dozens of such evolutionary advances. Not just increases in processing power but how we define a computer. Moore’s law will be irrelevant. The most visionary of computer engineers today will be as prescient as a Neanderthal. I believe this will only take 100-200 years but lets say it is 1000 years.
But all that aside… If I am a simulation, my perception of reality is what the program tells me reality is. Wouldn’t even have to be very complicated or computationall intensive. I would have no way to tell the difference.
If I am actually a biological being and all my external stimulus is coming from a simulation then how my brain works doesn’t have to be replicated or even understood. It just needs to be fooled.
As if civilization much less the species is still around to be involved. That is the problem I see with all the magical thinking. Having no idea how a problem can be solved they assume it will be, even as those things where the steps are known the process to accomplish them is whistling past the graveyard while the specific steps are optimistic at best. I am wildly in favor of positive progress and will do what I can but the math does not work in the end.
i hadn’t, that’s cool, too. i’d done some lego custom minifig research a couple of years ago for a work project, but we ended up not using my (really effing cool, if i do say so myself) design for our internal awards wall.
I’m a software guy and only peripherally involved in any “AI”, a term that’s popular on Wall Street and thus applied with a broad brush. So this is opinion, but that’s all the experts can give you anyway.
The human brain is not a single system, rather it’s an assembly of many systems that cooperate. This village of systems supports our sentience. Any one such system might be an idiot on its own, but the whole village is elevated to higher function.
Currently we’re in the business of building idiot savants. They can train themselves to exceed human expertise in precise fields; they explore success and failure for “years” in a handful of seconds. But they are generally static, unable to learn once trained (no memory of post-training failure). In the future we’ll build villages and give them the gift of continuous training. They will make smart decisions with a memory of failure. Is that consciousness? We’ll have to ask.
Keeping with the visual theme – more “AI” art than faked video: you have likely seen DALL-E 2 output all over the place lately but, have you seen Google’s Imagen?
I mean, what can you say about midjourney. It’s just astounding. I just did one for a realistic carved wooden figurine of a dog on a desktop, and it did this:
I’m in the stable diffusion beta, it’s pretty cool. It’s ridiculously fast compared to MidJourney, though I like the aesthetic of MJ’s output a bit better.
For comparison: MidJourney develops 4 concepts for a prompt in about 60 seconds. Stable diffusion just did 9 variations for me in under 14 seconds.
Video of someone putting together a laser-cut hurdy gurdy kit. There is a great view of lovely interlocking pegs at the 9-minute mark, and a hand crank with rotating knob right after it. Lots of good ideas for mechanical joins, and LOTS of gears.
I haven’t finished watching, so i don’t know if they play. [edit: it is played at the end]
