Also, a fascinating discussion about #AI devevelopment b/w a sociologist from #UCBerkeley and #SciFi luminary, #TedChiang.
They are on the right track, but concurrently so wrong in this new but limited perspective.
"Computer scientist Ray #Kurzweil and a few other futurists think that #AI dominance will arrive in just a few decades. Others envisage centuries."
From what I am reading elsewhere, this will not even take a decade, extrapolating the current rate of #evolution.
Due to #Meta's leak of #LLaMA has seemingly led to the open...
Artificial General Intelligence = a sentient being able to determine its own goals until 05/2024, ir rather even until 12/23?
Possible, however, only if embodiment and #AI linkage happen.
However, from the mentioned sociologist's and #Chiang's point of view, a caregiver situation is needed on top of that. The more intelligent a specifies, the longer the childhood, I seem to have learned in school (if still true.)
While believe this...
...to be possible--and, at the same time, suicidally stupid on our part--I also thing that maybe just hooking up the millions of actors and sensors from the #IoT to #Chat-GPT5 (?) could solve the embodiment dilemma.
@HistoPol@voron@xtaldave, consciousness isn't important. Agentic GPT models already set their own goals. We have made embodied AIs for a looong time already both in simulations and in robotics.
We'll simply do what we did from AlphaGo (mainly imitating humans) to AlphaZero (self-play), and make these large language models self-compete. This will skyrocket their general intelligence.
@HistoPol@voron@xtaldave, basically GPT model put into a loop. They are given a high level goal or a persona to emulate and told to instruct themselves. They make plans, go through them, re-evaluate, replan, execute, learn and continue.
@HistoPol@rmdes@voron@xtaldave, open source models are still playing catch up. Once models are trained it's relatively simple to quantize, prune and compress them to run on cheaper hardware.
...source community's democratization approach of #LLM's now outpacing even #Open-AI's development leaps. (See first link) This is the evolutionary booster predicted at the end of the last century as "humanity has evolved into a socio-technological complex."
...the two agree that #embodiment is needed, interestingly drawing on the period im earth's history that made me support the adirmentioned article, the #PrecambianExplosion: eye development,...
...movement predators, all interacted in brain development.
Your source continues:
"Or perhaps #consciousness emerges in any sufficiently complex network?"
IMO, as there already have been inexplicable occurrences and humanity has been experimenting with #NeuralNetworks capable of learning; I think the likelihood of this assumption being true is quite high.
"If, on the other hand, they are conscious, we should surely..."
Are these guys nuts?!? Homo sapient has exterminated any other humanoid species on the planet that could've endangered its hegemony. Humans trained these models explicitly or implicitly with our experiences and values. What on....
...earth is such a "superintelligence" bound to do, unless it were able to throw away the shackles of hunanocentric enculturation in its upbringing...Quite in contrast to us humans. 😉
This, however, is quite plausible:
"...organic human-level intelligence may be, generically, just a brief phase, before the machines take over. If alien intelligence has evolved similarly, we’d be most unlikely to catch it in the brief sliver...
Add comment