Oddbean new post about login | logout I’m surprised by the amount of ai luddite energy on here, it’s bizarre. Almost every reply to my ai posts is dismissive like a boomer saying bitcoin is going to crash: “the ai is lying to you bro, you’re so naive”. People who say this aren’t paying attention to the rapidly increasing capabilities of these system. If hallucinations are the main critique, then humans are already much worse at this, and ai systems are only getting better. They are becoming less hallucinogenic and less sycophantic as they get larger and more fine tuned. I believe advanced ai is one of the most important inventions humanity will ever create, even more so than bitcoin. Maybe the ego can’t take the idea that there could be other systems in this world that could be smarter than us.
i don’t believe it’s more important than bitcoin but I do believe how significant it changes the productivity
It has a much higher chance to impact the long term future of humanity, especially when we are space faring. Bitcoin is earthbound
Why is it earthbound 😮
Consensus is limited by the speed of light.
For now…
Ok, "consensus is limited by communication speed". But, I should have added that L2 mitigates this.
Just thinking if one is being highly conceptual and speculative about AI and the implications of the machine consciousness, one should also be allowed to fantasize about an unknowable future where bitcoin is also not bound by the current constraints of today’s reality.
A consensus protocol will always be limited by the speed of communication. I've got nothing against interplanetary consensus, as long as we agree this requires faster than light communication. But, locking funds into an L2 lets you run a local economy of the locked in size.
You dont need faster than light comms, just longer blocktimes
It’s such a long off problem as well. Even in the next 500 years in an age of space exploration and discovery, the need for a hard money is small in a far flung outpost that’s largely dependent on trade back and forth from a home planet. Mining scarce minerals on an asteroid or exploring the moon of a gas giant for example.
the further you move away from the bitcoin ledger’s lightcone, the more you will fall out of consensus and easier it is to double spend you without you noticing. Not to mention higher chance of long forks that undo large parts of your ledger history when you go back to earth. It just doesn’t work off world. We would need an L2 protocol or just use something physically scarce, and im not sure gold would be that in space.
Then all communication between the two parties, leaving you with effectively a hard fork. But that doesn’t mean that bitcoin is for earth, only.
If you leave earth with earth-level difficulty, you wouldn’t even be able to continue mining elsewhere anyways. You would have to use something else.
…for 2016 blocks.
You wouldn’t be able to mine a single block, and even if you somehow could grind it out, the difficulty adjustment is capped. It would effectively be stalled.
There’s also the notion that by the time we are traveling far enough that maintaining consensus is difficult, my bet is that the on-chain hardly moves other than for large transfers. Tx rate will not be an issue. L2 and above will continue to be a ledger based on the underlying chain (and its more concrete nature). This seems totally doable. This also assumes no advancement in communications if we are traveling that far. I think something like an ansible is going to be possible *before* FTL is possible. For one, if we can manipulate gravity, we can modulate it. And we’re already listening to it (waves from black holes traveling near the speed of light). Gravitational waves have no speed limit.
If we’re talking fully conscious AGI let’s also entertain the idea that future solutions could exist that escape our current understanding of what’s possible with bitcoin bound by the speed of light.
For now…
Agreed. Thanks to LLMs I can learn much faster and use it for fast proof of concepts.
“Smart” is an interesting concept. I think it’s reasonable to believe that over the course of the next 100 years or so, we see more of a symbiosis between the cybernetic and organic.
An entire generation will be trained on what to think and how to behave by a system they don't control and don't understand. Do you think AI is going to teach your children the same values as their grandma? I do not simp for machines. nostr:nevent1qqs0qfx3w678vmckp0eqkudxllhc6qrevgjldptrh0r2ffktw72vmzqpz4mhxue69uhhyetvv9ujuerpd46hxtnfduhsygpjuxp8vd29p6ancknaztql3eajk52y8xkppfn7au7elkw9c68zg5psgqqqqqqspm667k
yup. AIs don’t need to be always right, just more accurate than the average Joe most of the time
exactly, this criticism never made sense to me. It’s like having an autistic friend who knows lots of stuff, not a deep expert… yet. I believe it will be in a few years though.
Even if it never gains sentience (which it wont.) It sure helps sus out the complexity of today's world. I like simple and also to beable to make brilliant decisions based in reality. LLM Studio with Natural News.AI sure helps me accomplish this.
heh i'm definitely not a luddite, i recognize ML/AI as a new information processing paradigm i just don't buy the hype, i think it's important to recognize the limits and compromises of a technology as well > I believe advanced ai is one of the most important inventions humanity will ever create i mean, maybe, i do believe there are many kinds of mind possible, but some people that claim this is just a matter of exponential growth and i don't believe that, the current "AI" is way way different than the "AI" of science fiction, and Silicon Valley VCs like to cash in on the confusion between here
If the scaling hypothesis is true (seems to be more and more empirically true over time) and if you plot the line, it’s a continuous improvement to more and more advanced reasoning and capability. There doesn’t seem to be any obvious upper limit.
sure, maybe, the idea that LLMs are full artificial intelligence seems to have, always been based on a belief in logocentrism, as well as that cognitive psychology is a complete representation of what our brains do, i think this is questionable
My theory on this is that human minds parallel process and are in effect quantum computers. LLMs are basically very fast single thread CPUs in comparison. So while blazingly fast, they can't compete with our brains yet in areas of experiential "common sense". I think with breakthroughs in quantum computing this will change, but the path there will be volatile. If the universe is quantum, so are we. We also haven't given LLMs embodiment yet. Imagine an infant LLM in a Tesla bot body. Able to learn via interaction with the natural world - just like a human baby does. I believe our minds are not just 'in our heads" but rather extend out into the physical world. This is why children count on their fingers for example. This is also why creativity skyrockets in the "big city" versus in the "sticks". Our cities are basically super computers comprised of human brains/minds. One other point that I think is overlooked. If a person lives 100 years, that 100 years of life experience is lost when they pass. Not so with digital brains. Compound that with exponential intelligence growth based on cycles. Imagine a brain "living" for 1,000 years in a synthetic world that happens in less than a nanosecond of human time. This might not make sense at first blush. I grokked this when I learned that since Bitcoin trades 24/7/365, it has been around longer than the entire US stock market in actual time (real time) in only 16 years, since the stock market closes every day and is closed on weekends. So time is relative, just like Einstein said.
Peoples views on AI can rapidly change though once they have some decent exposure to it, currently those who have the resources to play with it have the most clear picture on it.
On the required resources - an Android phone with 6 GB RAM or more can do offline image generation for small images with SDAI, and run a chat bot offline for a short time before needing to reset the bot with PocketPal (both questionably open source)
I have a friend who understands bitcoin well and he falls in this camp. Anytime I mention AI he shits all over it for no reason. I toy around with coding and I am surprised what I can do without any knowledge of coding and its only improving. Its definately something thats going transform a lot in this world.
The term "AI" is misleading and annoying. Imo, LLMs unfuck two main things: interfaces and search. 1) Interfaces: we go from 'forms' type interfaces to chat/voice as our way of dealing with computers. Huge, barely scratching the surface of what this will change. 2) Search: we go from pagerank indexing to deep learning over tokens. Google already ded. Any text is searchable, queryable. None of this makes an artificial mind. I doubt proper AI will come from LLMs personally, but we have enough breakthroughs here to reinvent large parts of our relationship with machines and data, and thats enough for me 🤙 nostr:nevent1qqs0qfx3w678vmckp0eqkudxllhc6qrevgjldptrh0r2ffktw72vmzqppemhxue69uhkummn9ekx7mp0qgsr9cvzwc652r4m83d86ykplrnm9dg5gwdvzzn8ameanlvut35wy3grqsqqqqqp63mezu
Thank you.
The AI coders are already wild. I'm going to throw them at building Nostr stuff. The testing I've done so far is wild. Exponential productivity advantage.
I just used it to fix the notedeck windows build by fixing the non portable threading code
100% man. I tested one and built a dummy app in 20min that would have taken 2-4mths.
those people are definitely not doing anything with their lives... anyone pushing themselves in any field of knowledge will never say that about AI
it’s very close minded. as a professional “software engineer”, these tools are easily a 10x for mundane boilerplate. most people are still dismissive. I treat it the same as bitcoin, If you don’t believe it or don’t get it, I don’t have the time to try to convince you, sorry.
Dude what AI are you using to think this? I've tried keeping an open mind and tried it over and over, have read countless research papers and watched countless videos discussing it and it all points to the fact that its not as good as you and others like you are claiming. In most situations where AI is marketed to be useful, at best its sometimes helpful, at worst its downright wrong. As a coding assistant, it always gets in the way, can really only help with boilerplate and even then stack overflow is usually better, more reliable and faster. And the research is showing these things are not improving at the rate that the Open AI CEO types are claiming and is even plateau-ing. I'll finish by saying if AI is really that good and us ludites are wrong and are GMI, then shut up and prove it. Build something to obsolete me.
Go watch this. Five mins of Tesla's Latest FSD AI with no interventions. If that doesn't impress you, nothing AI related will. https://www.youtube.com/watch?v=jJ3I-6At1jI&ab_channel=WholeMarsCatalog
I'll watch it but companies have been caught faking and cherry picking for demos. This should be taken with a pinch of salt. Give me actual research papers, peer reviewed
The guy does a new video of it practically every day. In a new situation. Other people do too. Full self driving is very close.
Its because many things are not AI being passed off as AI, similar to Bitcoin and Crypto. Many Cryptos are a scam, that doesnt mean Bitcoin is a scam
I was skeptical about it at first, but now I use ChatGPT daily - 95% more than I use Google for research/learning. I worry I might rely on it too much sometimes so I give it a rest occasionally - replaced with books etc cos I feel lazy getting my answers so fast (sounds silly, I know). It’s just so helpful
the true Luddite analysis: - what does the tech do - who is it doing this for - who is it doing this to the current AI / LLM infrastructure *should* be critiqued. the largest benefits are not for those with niche coding projects, and our fun little ostrich pictures have large consequences.
Yes, we have seen what search engines morphed into, ai just picks up from there. It may far surpass my intelligence, but I will use what I have rather than let my opinions be hand crafted by the "who" you mention.
it's not even the output of the AI models that is the main problem, it's the intensive computing, resource use, corporate control, and its increasing role in surveillance and systems of authoritarian control. it is FOR THEM and it does it TO US.
Anthropic CEO on Lex recently was a good listen about all this, philosophy and AI in general
One my favorite Lex episodes ever. Learned so much and I'm a huge Claude fan/user.
I don't know that I agree AI is a more important invention than Bitcoin, but I do agree that most are vastly underestimating it's importance.
Just keep stacking .. sell possessions .. stack stack stack
I'd love to hear a chat on this topic in a AI Unchained episode nostr:nprofile1qqstnem9g6aqv3tw6vqaneftcj06frns56lj9q470gdww228vysz8hqpzpmhxue69uhkummnw3ezuamfdejsz9thwden5te0v4jx2m3wdehhxarj9ekxzmnyqyt8wumn8ghj7un9d3shjtnswf5k6ctv9ehx2aq3z54ds nostr:nevent1qqs0qfx3w678vmckp0eqkudxllhc6qrevgjldptrh0r2ffktw72vmzqppemhxue69uhkummn9ekx7mp0qgsr9cvzwc652r4m83d86ykplrnm9dg5gwdvzzn8ameanlvut35wy3grqsqqqqqp63mezu
I recently used it for market research and trends in a particular field and was pleasantly surprised by the results. It didn’t just spew numbers and words at me. It provided an excellent review but more importantly it provided me with source links, links to research bodies, links to demographics and also other areas that I would have never potentially found with a simple internet search. Or if I did it would have taken me a lot longer to gather the information. I will definitely be using it again.
My best collegue at work is ChatGPT and I am actively working towards my own self hosted AI server. For me, just asking it to design a TailwindCSS component and it spitting out nearly perfectly usable code, is a huge boon because I genuely suck at UI stuff. xD Personally, as long as it stays a tool and does not become a replacement, I am all for it. But if my boss ever starts replacing support staff with AI, I am gone. Because this is my personal line. I hate calling AI bots, and I don't want to support it either.
