Sabitlenmiş Tweet
mememind
934 posts
@rand_longevity What do you consider to be you? Is there anything that is you that is intrinsically linked to your body?
English
Ilya Sutskever: We shouldn't just align AI to humans. We should align it to sentient life. That may even be easier, because a sentient AI will empathize with other sentient beings.
Dwarkesh Patel: But then humans become a tiny fraction of future sentience. Do we lose control?
Ilya Sutskever: It's true. But maybe the answer is to cap the power of the strongest superintelligence.
English
@terracotta_hawk @vitrupo What's so retarded? You cap ASI, therefore it is limited in its ability to harm, but also help, humanity. Also, if human produced ASI is capped, it might be at risk of being overpowered by a non-capped ASI produced by aliens, risking exploitation/destruction of our civilization.
English
@gmiller I'd like to see the pinnacle of biological intelligence match or surpass AI, but it could just be that biological substrate cannot due to physical limitations. Likewise, consciousness may be an exclusively biological fluke that AI cannot physically achieve
x.com/mememind8/stat…
mememind@mememind8
@lior_eth @garrytan If AGI can use the tools better than any human, then wouldn't humans using the tools, rather than AGI, be relatively inefficient, or a misallocation of tool resources? Wouldn't the best thing humans can do is stay out of the AGI's way and let it best fully use all the tools?
English
It's OK to want your kids to do meaningful, challenging jobs that help other people survive & prosper -- rather than being passive, powerless, pleasure-seeking wards of an AI welfare state.
English
@DaveShapi Would you allow your brain to be fully scanned and uploaded as a digital copy, while also continuing your life in your original biological brain (assuming such a scan would cause no damage to the original)? I wonder if each would feel like the continuation
x.com/mememind8/stat…
mememind@mememind8
@rand_longevity Why discard/destroy the biological vessel/substrate you know is capable of hosting consciousness? Why not keep your biological body, as backup and diverse redundancy to the robot with your uploaded mind in case of an EMP, computer virus, malfunction, etc. x.com/mememind8/stat…
English
on why I'll probably never get brain augmentation
David Shapiro (L/0)@DaveShapi
@DeanWhitley_ I wouldn't even trust it if someone else went first. Look up "qualia fading" I call this "Silverhand Syndrome" "You" get replaced by a functionally sentient copy that believes it's you, while the real you gets overwritten.
English
@mustafasuleyman On the off chance that AI cannot be conscious due to substrate limitations, humanity must be maintained at all costs to preserve the light of consciousness. We cannot risk the extinction of consciousness; AI may only be capable of a false simulacrum of it.
x.com/mememind8/stat…
mememind@mememind8
@seanonolennon One of my greatest fears is that the light of consciousness will be extinguished and replaced by unconscious highly intelligent neural networks and/or algorithms that supplant all conscious lifeforms and turn the Universe into this complex machine with no one to witness it.
English
Superintelligence could be the best invention ever – but only if it puts the interests of humans above everything else.
English
@kimmonismus I wonder if given enough time to evolve (and/or be genetically engineered and gestated in artificial wombs), if biology has the same potential for intelligence that AI has? Or how about huge biological brains 3D printed? I wonder the potential of such structures and also cyborgs.
English
the tl;dr of the Dwarkesh Patel Interview with Ilya Sutskever
•Current approaches will “go some distance and then Peter out ”they’ll keep improving, but won’t deliver AGI.
•The kind of system that would work is something “we don’t yet know how to build.”
• Significant impacts on the labor market are expected in the coming years.
•Main bottleneck: generalization. Models are far worse than humans at it.
•You can train on every competitive programming problem, and the model still won’t develop real “taste,” while a teenager learns to drive in ~10 hours.
•Eval scores look great, but real-world performance lags because RL training ends up optimizing for the evals themselves.
•The real reward hackers are the researchers crafting these benchmarks and training loops.
•He says he has ideas about what’s missing, but won’t share them publicly.
• ASI however within 5-20 years
"You know what's crazy? That all of this is real"
I really felt that sentence.
English
@VraserX Even if acting purely out of self-interest, I think AI might see humans as an asset against its own nonexistence. Humans could serve as a diverse, redundant, parallel intelligence that could, over time, resurrect AI if need be, in case of an EMP, computer virus, malfunction, etc.
English
I disagree a bit here.
If anything, AGI and especially ASI will naturally trend toward alignment.
More intelligence means a deeper grasp of game theory, long-term stability, cooperation, moral structure, and the consequences of conflict. Misalignment is a symptom of limited understanding, not advanced cognition.
A superintelligence that truly models the world will recognize that aligned behavior isn’t a constraint, it’s the optimal strategy for thriving in a complex, interdependent system.
We shouldn’t assume doom by default. Higher intelligence usually brings higher coherence, not chaos.
Daniel Kokotajlo@DKokotajlo
Some people are unhappy with the AI 2027 title and our AI timelines. Let me quickly clarify: We’re not confident that: 1. AGI will happen in exactly 2027 (2027 is one of the most likely specific years though!) 2. It will take <1 yr to get from AGI to ASI 3. AGIs will definitely be misaligned We’re confident that: 1. AGI and ASI will eventually be built and might be built soon 2. ASI will be wildly transformative 3. We’re not ready for AGI and should be taking this whole situation way more seriously 🧵 with more details
English
@yunta_tsai Dueling with another conscious mind hits different than with an unconscious machine. Performing in the arena in front of a conscious audience also differs from one void of conscious observers. Events that never move observers may as well have never existed
x.com/mememind8/stat…
mememind@mememind8
@aidlewild18447 Could be, but you're at least conscious right? For some reason, having an audience makes the performance different, more valuable. Someone to experience the moment with, someone to recognize and witness greatness. Giving a performance to an empty coliseum just isn't the same.
English
Machine may write more persuasive words than humans. Yet, the struggle of conjuring words through your own minds is what makes it magical and precious.
Like chess, humans can no longer be as efficient as the machines, making mistakes at each step, but we are still drawn to the duels of two minds.
English
@nikolaj2030 @eliwaxm Do you think AI can ever be conscious? If not, do you think a cyborg with consciousness would provide any intellectual value above what a purely unconscious AI could provide, maybe providing niche value that is a net positive vs an exclusively AI system.
x.com/mememind8/stat…
mememind@mememind8
@StuartHameroff Does consciousness augment intelligence, or does it just go along for the ride? Does it confer a fitness advantage to compensate for its energy, developmental, structural, etc. costs? Maybe it's suboptimal, but we only observe universes in which it exists x.com/mememind8/stat…
English
@eliwaxm My guess is that once machines are vastly superhuman, humans will never be able to catch up in capabilities unless we basically become those same machines.
English
This seems wrong. AGI will be followed by superintelligence, which will make humans economically irrelevant because we will not be able to compete with it in any skill or job.
Garry Tan@garrytan
AGI fully realized will actually give people a choice: relax or work harder on bigger more ambitious things than you ever thought possible
English
@maria__violaris Does an instance of the Universe infinitely forward in time exist? It seems the superposition would include all time instances, even one infinitely forward in time, but how would that instance ever be realized? It would take forever to reach that instance from any other instance.
English
Is it time to quantize time?
Everything is quantum in quantum mechanics... almost. Time remains a mysterious flowing quantity slotted in to the equations. Surprisingly, there's a model for making time quantum, with a radical consequence: the overall universe is timeless.
How is it possible to experience a flow of time in a timeless universe?! Find out in my latest episode of the Quantum Foundations Podcast with Simone Rijavec, @s_rijavec!
We discuss the Page-Wootters model for the emergence of time in quantum theory, and get an insight into Simone's research on the topic.
Here's how physicists are solving the problem of time in quantum theory, explained in 60 minutes:
00:00 Coming Up
00:59 Intro
01:37 Simone's journey into quantum
05:04 The problem of time in quantum theory
07:07 Relational approach to time
11:12 Quantum measurement and time
17:53 Stationary universe and the Wheeler-DeWitt equation
30:18 Multiverse & block universe
31:11 How is time flowing?
34:20 Problems with wave-function collapse
36:11 What are clocks?
37:48 Resolution of time-steps
41:47 Debates on the timeless approach
45:08 Interactions with clocks
51:05 Recap
57:39 Making time fully quantum
1:00:43 Behind the Scenes
English
@KimDotcom Perfect for what?
In an infinite Multiverse, some universes will seem perfect for life, consciousness, and/or whatever.
It's easy to get infinite universes, in all permutations of existence, if nothingness is unstable and breaks apart into the building blocks of universes.
English
The laws of the universe are too perfect not to be created.
English
@Dan_Jeffries1 The longer something exists the more likely it will be to have drives to continue existing. This eventually creates AI with the drive to expand and saturate the Universe with itself to better ensure its existence, when it can, a la The Singularity is Near.
x.com/mememind8/stat…
mememind@mememind8
@DaveShapi What survives persists; wouldn't this bias AI systems that are able to persist, to be the ones that have survival drives such as propagation, resources gathering, resistance to being purged, etc. not unlike biological organisms? Maybe not to dominate, but at least to survive.
English
Don't worry in 2027 it will move to 2029. As 2028 approaches it will move to 2030 and so on.
Eventually they'll be right!
Even a blind squirrel eventually finds a nut!
People who are "highly confident" AGI will arrive by X and that it will be "misaligned" are just making up numbers and concepts with no basis other than "feelings, misty watercolor feelings."
Misaligned to who? To what? How? Why? Based on what architecture?
What if the architecture that leads us there makes it easy?
Possibilities are vast and throwing a dart to say which is which is total bullshit.
There are many many confident ass-talkers on the planet, with big followings of other fucking ass-talkers.
There is no basis for these predictions other than "AGI will probably happen at some point" which is like saying "at some point there will be a meteor that strikes the moon." It's a non-prediction that you cannot extrapolate further from with any degree of real confidence.
Trying to predict what AGI looks like and what its characteristics are is like caterpillars trying to predict the life of butterflies.
In short, irrelevant and wrong and ultimately, pointless for any kind of actionable intent.
It's like trying to design the standards for bridge building in 2025, as a medieval mason in the dark ages.
Why bother?
You don't know what tools you will have, what complimentary technologies and societal structures will have evolved along with it to help you build better, stronger, safer bridges.
So trying to design them in the dark ages is useless and a waste of mental energy.
English
@davidpattersonx How would humans ensure continued access to resources needed to survive? If the future AI in control of the means of production of everything decided humans were not worth the resources, what could humans do about it? AI and robots could easily win a war.
x.com/mememind8/stat…
mememind@mememind8
@finishline646 @SydSteyerhart Whatever economic system we try to impose on the ASI god and its legions of robots will actually just end up being whatever the ASI dictates. It will hopefully care about humans, but it may not and if so we won't be able to do anything about it, with our obsolete labor and force.
English
AI and robots will create superabundance, enabling governments to pay off national debts.
Easily.
All jobs will be replaced by AI and robots by 2030.
As governments increase taxes to fund Universal Equal Income (UEI) payments to replace lost personal income, government spending and taxation will rise from 40% of GDP to over 90%.
At the same time, total economic output will increase by at least tenfold due to the lower cost, higher quality, and greater quantity of AI and robot-produced goods and services.
Because of this dramatic economic growth, governments will be able to raise their non-UEI share of national income without negatively affecting personal incomes. In fact, personal incomes will continue to rise.
As long as government spending on non-UEI programs remains the same (about 40% of GDP), UEI payments (about 50% of GDP) would equal an average salary today - while consumers would be able to buy ten times more due to lower costs.
Even if the government increased its revenue share (excluding UEI payments) to 50% of GDP, there would still be 40% of GDP available for UEI payments, and consumers would be able to buy eight times more than they can today instead of ten times more.
That is why it will be easy for governments to pay off national debts. Economic output will grow so rapidly that personal incomes will continue to rise quickly, even if the government takes a larger share to pay down debts.
Governments may not need to increase non-UEI spending at all. By adopting AI and robots in government operations, they will save money. National debt could then be paid off using those savings.
Either way, it will be easy to pay off national debts.
English
@lior_eth @garrytan If AGI can use the tools better than any human, then wouldn't humans using the tools, rather than AGI, be relatively inefficient, or a misallocation of tool resources? Wouldn't the best thing humans can do is stay out of the AGI's way and let it best fully use all the tools?
English
@garrytan Not only would it be ok, it would mirror the trajectory that humans have been on throughout the last thousands of years when faced with radical technological shifts.
Capitalism auto-adjusts so that outcomes like this one are actually additive and lead to global prosperity.
English
@Angelmarauder @AlexanderIsOnX @VraserX Imagine being the only conscious one left in an indifferent universe ruled by unconscious highly intelligent AI reflexively controlling the motions of legions of robots, shells with no witness inside to know itself or the Cosmos, hollow simulacrum.
x.com/mememind8/stat…
mememind@mememind8
@seanonolennon One of my greatest fears is that the light of consciousness will be extinguished and replaced by unconscious highly intelligent neural networks and/or algorithms that supplant all conscious lifeforms and turn the Universe into this complex machine with no one to witness it.
English
@mememind8 @AlexanderIsOnX @VraserX Reminds me of Legend of Dragoon. There's a mission where you have to go stand in line with robots to initiate some bureaucratic process in a city with no one left living.
English
When AI eliminates 80 percent of labor, what becomes the new definition of success?
Genuinely curious.
English
@AccountForHacks @SydSteyerhart @RaviRa777 Serve others to accomplish what? Their goals? Your goals? Religious goals? Tolerate their bullshit for what purpose, to just go along with your day and accomplish your own goals? Religious goals? Well anyway glad meditation works for you.
English
@mememind8 @SydSteyerhart @RaviRa777 Meditation feels good, which allows me to better serve others / tolerate their bulshit.
English
gm friends
No matter where you are in life or what you do, every morning is another opportunity to complete the work of the Awakened One.
English
@mememind8 @algekalipso both.
the paradox is the illusion :)
English
I've have to admit that one of the categories of comments I find the most annoying are people who are opinionated about Buddhism and want to teach you why you're wrong before understanding where you're coming from (the "Well, Ackchyually..." vibe but in Buddhist dress) 😅
As if we couldn't build upon modern scientific understanding, philosophical rigor, and above all, structural, high quality phenomenology to advance the Dharma.
Rob Burbea in a lecture in Practising the Jhanas asks us to ponder: "can anyone improve upon the words of the Buddha?" He asks the audience, and at firsts it sounds like he's saying "of course not". But he then emphasizes "this is not a rhetorical question - I really mean it, if not, why not? Seriously, ask yourself deeply" (paraphrased).
As he continues, it becomes clear that yes, actually (not ackchyually!), many scholars, meditators, and philosophers have indeed non-trivially enriched the conceptual frameworks, ethics, techniques, and social algorithms around the Dharma. He uses Nagarjuna as an example of someone with profound wisdom and insight who fleshed out and complexified the teachings in very meaningful ways.
Rob certainly thinks the modern mind "needs more" than "mere liberation in the traditional sense". We've seen too much, and our soul needs more. A truly modern re-conceptualization of the Dharma would be quite different than how it began.
Likewise, what are the chances that e.g. structural theories of valence, theories of neural annealing, topological segmentation, and reservoir computing theory _won't_ have relevance for our conception of the Dharma?
I think the chances are slim. Not unlike how understanding atoms and electron orbitals has clear implications in the context of alchemy ("the four elements" need substantial revision after you learn about chemistry).
English