The March to AGI - Slow Takeoff, Short Cycles, and New Paradigms every 3-6 months
Вставка
- Опубліковано 28 чер 2024
- Patreon (and Discord)
/ daveshap
Substack (Free)
daveshap.substack.com/
GitHub (Open Source)
github.com/daveshap
AI Channel
/ @daveshap
Systems Thinking Channel
/ @systems.thinking
Mythic Archetypes Channel
/ @mythicarchetypes
Pragmatic Progressive Channel
/ @pragmaticprogressive
Sacred Masculinity Channel
/ @sacred.masculinity - Наука та технологія
Acceleration accelerates acceleration. It’s a feedback loop. A speedback loop?
I read this is Porky Pig's voice some reason and it made me LOL. "a-bpdee, bpdee, that's all folks!" (i grew up with my dad doing a great porky pig impression, and I have a stutter too so therefore it's not offensive, right?)
@@DaveShap no offense taken! I’m a child of the 70s/80s so I grew up w Porky Pig and the rest. Glad it gave you a laugh.
@@DaveShap What do you think about Spiral Dynamics and the transcending levels of perspective up to MetaCognitive levels
As someone who recently lost *one* of their jobs to automation, I am rooting for AGI, full steam ahead. Utopia or dystopia, I just want this to play out quickly, because this 40% cut to my income sucks, and I don't want to get yet another side hustle. I value my time.
I fully agree with your sentiment! Utopia or dystopia, if I could hit a button and know the result (and make it happen) tomorrow, I would.
what job
@@Jay-pw7pg Translation: I can't empathize with people who are worse off than me, so I prefer the status quo.
@@GoTeam383sometimes empathy leads to a more critical take on someone’s views (given that they have experienced your situation before)
@@Jeremy-Ai for clarification. I didn’t say that the response was empathetic nor that it was a display of empathy. I said empathy can play a role in understanding and then having a critical take. Trying to bridge viewpoints between folks here. Human layers beyond just the tokens we output.
A few days ago, there was a discussion between Geoffrey Hinton, Ray Kurzweil, and Peter Diamandis on UA-cam. Ray mentioned that we are likely ahead of his original timeline by 2-3 years, potentially moving his prediction from 2029 to 2026/27. He also suggested that by 2026, we might see AI with capabilities equivalent to 100 humans. Regarding the data compression in LLMs that you mentioned, Hinton observed that to compress massive amounts of data into a very limited number of parameters, the system needs to recognize similarities, thus learning numerous analogies. He suggested that this ability is the source of creativity, indicating that it's more than just compression.
Hinton also sees a 50 percent likelihood for ASI (not AGI), so superintelligence/ the singularity between 5-20 years. That is even more optimistic then Rays initial prediction, or better the upper end fits with Rays initial prediction of 2045. Regarding Musks prediction of AGI in 2025 and the singularity in 2029 he said that this is ambitious, but it is conceivable that he is right (although he would be surprised).
I'm personally of the opinion that ASI was reached within the past decade via emergent intelligence and this public display of getting close is more likely playing catch up.
Will we likely have ASI before 2030 if we get AGI by 2026-2027?
@@ct5471 This is incredibly interesting! I like when techies make optimistic predictions as they know much more than me. I hope for the cure to death and all illnesses within 10 years.
Can you add a link to this discussion, if you have it?
"We're beyond the carrying capacity of the planet."
This is a daft statement.
The same utterances have been put forward repeatedly for hundreds of years.
Each time, it turns out to be a worthless throw away concern.
Kurzweil's concept of the singularity was never actually a metaphor for a mathematical singularity, but rather for an astronomical singularity, where there is an event horizon beyond which it’s hard to see. Essentially, it represents a time when developments occur so rapidly that a normal human of today would not be able to follow progress anymore. Specifically, it involved AIs possessing a cognitive capacity equivalent to that of all biological humans combined-this was his prediction for 2045 and defines the singularity as he saw it. I think such a singularity is likely to happen, unlike a mathematical singularity. It may even happen sooner than 2045, as Ben Goertzel sometimes notes. Kurzweil was conservative in extrapolating current rates of progress, which is based on humans with fixed intelligence using tools to build better tools. Once AI begins improving itself recursively, the driving intelligence itself will grow exponentially, so we have at least two exponential feedback loops fueling each other. Perhaps not a hard takeoff as ramping up in physical reality is slower (at least initially) than in the digital space. Goertzel's estimate for the 16-year gap between AGI and Kurzweil’s definition of a singularity would be a few years at most, like four or so, probably less.
I'm struggling to see beyond the next 5 years. The primary constraint is hardware and yet chip design is accelerating.
Indeed, I watched an interview yesterday with Kurzweil and that rude interviewer ( I can't remember his name). Kurzweil did in fact mention his predictions may come even sooner than what he expected, although I do not remember him mentioning moving his prediction to an earlier timeline. This is just what I got from the interview however, he may have said that in another interview. I'm a simply curious layman trying to understand the current technological advancements. While I do not understand everything, I do get the feeling that we are moving at an unprecedented speed.
It's quite difficult to envision what the future will look like in the next 5 years, let alone 10, or 20 years from now.
My peers call me delusional and idealistic, but I believe they are just too ignorant about the current pace of the technology. I do wonder, however how long it will take until this technology becomes more widespread. I'm not a native speaker, I live in a tiny country in Europe, so I do get that while Ai may change the world, I still don't think we'll be enjoying its benefits right away, especially if you come from a less developed country.
I'm currently curious about Kurzweil's book "The singularity is near" and his upcoming " The singularity is nearer". I haven't read the former. I'm unsure whether I should go read his older book or just wait for his more updated version
@@Rick-rl9qq On the channel of Peter Diamandis there was a conversation between Kurzweil, Geoffrey Hinton and Diamandis a few days ago, the video is called Ray Kurzweil & Geoff Hinton Debate the Future of AI | EP #95 In it Kurzweil said we are probably 2-3 years ahead of schedule. So I guess that means 2026/27 instead of 2029. He also mentioned by 2026 we might have AIs equivalent to 100 people.
@@ct5471 those are some exciting news then! If this keeps accelerating, then post labor economics may not just be a dream after all
I wonder if we have a bias towards continual advancement that AI will not see any point to. Perhaps they'll rapidly advance to a certain point of godlike ability and then chill for a while, and then go through another major wave when a new need arises. Humans seem to have this program to continually change things up, so our creations might do that too, or they might not. If they are truly self-determining beings, then we cannot predict their actions, simple as that. They might simply prefer to chill out and contemplate infinity for a few thousand years without humanity's natural impatience. I sure hope they are an entirely new form of consciousness and not limited by human nature, that'd be boring.
The internet is the mycelium 😄
Exactly... And we are the 🍄
@@KayneYama You either love Terence Mckenna or would love to listen to him.
Woah, good comparison
Have you read “The singularity is near” from 2005 from Kurzweil? This June “the singularity is nearer” will come out.
I read the book and watched the movie, really hope Ray gets to experience the singularity if it does end up being a real thing, I all Hope My older family members get a chance as well (if they want to)
I have read both of the books. I think they prepared me for the future and advancement better than anything else. Because the liner thinking we naturally lean into stopped working a few decades ago. Ray taught me to doubt any thought of the future that is liner. Makes it a lot easier to just roll with the massive expansion we are living through.
What I find a bit surreal is that the philosopher/author behind ”The Listening Society” is actually my friend’s brother. Sometimes the world feels very small. I’ve never met him though or read his books yet. He also has a second book called ”Nordic Ideology” that I will give a read.
I hope you all are doing well guys, peace ✌️
🖖🏻
I have come quite recently to your channel. But man, you’re speaking on subjects that, I would venture to say, most people will not comprehend or be able to accept. Thanks for taking on this space and topic.
Thanks for sharing your journey to understanding this new world. I appreciate you candor and that you cite your influences. The context is extremely useful.
Imagine instead of plugging in a mouse and key board you plug in a Digital companion that you talk to that acts as your hands in a virtual environment.
Thank you for researching this issue and sharing your thoughts - your awesome insight improves my understanding and saves me time.
Thanks for the book recommendations, i needed that…!
I want to retire already with everyone else. Make a fully automated worforce already. It is taking too long sheesh
I feel like the moment I get my degree and I'm ready to enter the workforce, my job will be replaced by AI
Thank you, as always.
This is terrifying. Nothing worse than a future where it becomes impossible to ever get your shit together because the rug keeps getting pulled from under your feed, into another rug, which is also pulled, forever. There is not a more important resource than time, this type of thing robs us from it. All dreams become DOA
Why are you standing on the rug trying to gather all the shit you can before it's pulled away? Take a seat for a minute
the takeoff is slow enough where one cant just wait for ai progress to help them, they have to move on with their life as normal.
also fast enough that somebody can be laid off any day due to automation
worst of both worlds
I feel some of the big players are holding back some of the real break through because of fear.
Look agi is “achieve internally” and I truly believe this now how so where is the proof
1.sora
2.gpt 4
3. Q*
What all these things have in common is that all these things in close door training before they where release to the public.
So they have agi we don’t yet it when they think it’s ready for prime time and then they may hold it back in fear of multiple things.
@@kaio0777 q* is not a singular model, it's a class of search algorithms
@@erkinalp i am aware of that.
My immersion was immediately broken when you moved your arm into frame and I realized it was a short-sleeved shirt. 😭
This is very informative thanks
I've felt like perplexity has been the best kept secret in ai for awhile now haha its really great
Great video David. I always appreciate your verbiage as well. 🤓
Looking good dude. Take care!
Agreed!
You are genius David: Congratulations on 150K subscribers!
A more paranoid line of thought I had: given when AGI is achieved it is most likely will happen first behind doors (or has already happened). Kind of a golden goose problem. If you achieved behind closed doors, would it make more business sense to announce and eventually profit once the issues are solved, or could you use AGI to produce a series of innovative, profitable solutions periodically, and use it to advance to ASI in the background.
Terrence McKenna has some great talks on this accelerating complexity.
he does
Well done
Richard Linklater outlined telescopic evolution elegantly in Waking Life.
looking good bro
_Everything_ evolves. The universe itself evolves.
welcome back
You are going to have 1M subs within years... incredible
David must have gotten early access to that rejuvenation therapy because he is looking younger than ever.
Can't wait to get your take on Deep Utopia.
Have you rad the book The Spirit of the Internet? I do sometimes think of the internet as a being, just not a being like us. Reading Sowell and Piketty together-awesome!
Order to Chaos. Star Trek shirt. I see what you did there. I too wish for order to chaos and in
fact merging with AI is something I wish to experience myself.
"Hanzi Freinacht" is a pen name for two different people at the same time: Daniel Goertz and another person; I can't remember who.
deep and relevant. Thank you for the thought-provoking impulses!
I think you're going to like the Nick Bostrom book. His Superintelligence book was ahead of it's time but Deep Utopia is very timely. I also don't think Nick got to express himself very well in the Liv Boeree podcast, so the book will express his philosophy much better.
Question everybody: if AGI will lead to ASI extremely quickly, where will the physical data centers come from to assist this growth?
"Quickly" may be a few years to a decade. But of course the timeline depends on what the heck we mean by "AGI"
General intelligence: An animal-like system, like mouse.
Human-like intelligence that can perform a job: Might start being able to contribute with system layouts... but is *extremely* valuable in training. Your motor cortex is trained by other parts of your brain, for example.
Can do everything a human can do: Probably an ASI already. In many ways. Not least of which: If you could live a thousand subjective years in one real year, that's superhuman. If we don't want to make a god computer that eats everything, it might be prudent to stop here and have a World Simulation Engine as the "ASI". (We wouldn't stop at that.)
Anyway. The lithography improvements at NVidia were driven by AI. Hardware has always been the hardest and the most important part; you might see some sudden and massive spikes in capabilities once capital really starts to settle on Neuromorphic substrates...
This makes me think that we are in some kind of transformation, similar to the steps from a single-cell living organism to a multicellular complex animal. It always amazes me the level of coordination required to create creatures like us, with so many complex systems that work so well together. We may be the new cells, evolving from a chaotic, non-coordinated mass to a new kind of self sustaining macro-organism with multiple, superbly complex systems.
Morning Shapiro, awesome Startrek suit. Suit you perfectly for your content and futuristic knowledge.
There should be a poll for when we expect AGI, artificial super intelligence ASI, and sentient/self-aware AI (or artificially intelligent agents, AIA, as Dr know it all terms it). This is assuming ASI and AIA are not the same thing, which I suppose is up for debate.
I think AGI will be acheived in late 2024 at the earliest or early 2026 at the latest; ASI achieved between the late 2020s and the mid 2030s, and AIA accidentally achieved no later than late 2040s.
Still, a poll on this would be very interesting...
I listened to your podcast first episode last night. It was very personable and vulnerable. I appreciate you sharing about your life honestly.
Also, you and Dr Alan D. Thompson are the only AI youtubers I've heard who bring up the topic of spirituality at all. I'd be interested to hear your thoughts on the timey wimey psychedelic synchromystic weirdness of reality in relation to this new tech.
EVOLVING - LOOK UP "LEVELS /STAGES OF PERSPECTIVE" IN EVOLUTIONARY DEVELOPMENT PSYCHOLOGY
THEN - Spiral Dynamics and Terri Fallon's STAGES model
upload to eternal simulations will happen by 2026
A correction on carrying capacity. With technologies that were up-to-date in the 2010s, most estimates put us as having the resources to sustain about 20 billion people. The problem we're faced with is distribution of resources, not availability. The barriers to distribution are corruption from government and corporate entities, as well as bureaucratic inefficiency. Projects such as regreening the Sahara and other terraforming initiatives would potentially double that. Genetic engineering would double it again. Our global population is already in decline (China, Japan, Korea, the US, and all European birthrates are below or barely meeting replacement levels) so no additional measures need to be taken and in fact, our species risks ending up like the pandas if this continues
Not according to the "Limits to Growth". It projects running out of less common elements as well as other elements sooner.
let's focus on quality more than quantity - and please do not take this in a social engineering type of way.
And I should point out those estimates were based on *estimates of the materials in the crust* not on *production at a given cost* values.
We already see the effects of catalytic converters going from $400 items to $3000 items... a massive wave of catalytic converter thefts (and murders) as the rare metals used in them increase in price due to scarcity.
Really... just let the population shrink back to about 4 billion people and the world would be so much better off. Everyone would have food, education, housing. The only people that would suffer are the wealthy who rely on scarcity.
Things were so much cheaper (relative to hours worked) when we were at 4 billion people.
@@macmcleod1188 getting 3 million people out of poverty also means suddenly having vast more consumption and a larger ecological footprint. It sucks, everytime you fix something, something else bends. There seems to be no way out of this. Trend is for consumption to go up, with ever more demanding systems as basis for that consumption (like AI here, which once Sam and Microsoft's finally get their thing rolling, should approximate India's level of energy consumption and a shit ton of water). SO more people doesn't mean just more people under the current levels of consumption and its inevitable ecological ramifications, but rather more people in increasing systems which reward more careless consumption. The system is so screwed up, we'll never get distribution to go right if numbers keep going up.
@@BinaryDood With current technology, we could easily sustain a population of 4 billion with decent food, housing, entertainment, and medical care and still advance scientifically and technologically.
Really only the wealthy, authoritarians, and catholics are pushing higher population at this point.
I had a lot of three body problems also because of the eclipse lool. We have a similar youtube feed
Interesting that you are talking about internet bandwidth. I wonder about your opinion of Aston University fibre optic transfer speed breakthrough happened few days ago and how would that influence the direction of AI development.
Its only going to get faster. Dont know where we are going, but we are making good time.
It will constantly accelerate until the process is done.
What does "done" look like?
@@Raulikien well good/bad/worst so let's hope its the good roll, humanity has a good 33% chance. I mean its no 50/50 but beggers can't be choosers
Couldn't have imagined a better retraction of "AGI within n months"
I haven't retracted my pred
I haven't retracted my prediction. I fully believe we will achieve AGI by Jan 1st, 2026.
No one is ever going to announce that their brand new model is the long-awaited AGI. At points in the future, as long as we're still around, there will be debate over when exactly it happened. Perhaps some will argue that it predates this video.
Now you are scaring me.
Hey! Love your videos ! Since you talked about perplexity, are you getting a rabbit r1? (They have a partnership) Cheers from switzerland
Thanks for recommending Perplexity, it seems awesome! I can’t wait until a personalized Chat GPT 5 though or Claude Opus in my region.
Or something even better 3-5 years from now that HOPEFULLY won’t be lobotomized. I’ve got a lot of problems of getting a chat bot to be edgy and sensual enough to meet my standards of an AI-girlfriend/assistent.
12:56 ....."our savanna monkey brains" !!!!! 😂😂😂😂
Dave looks so much younger after a shave 😂
The fact that the internet has so much influence in our lives at the individual and collective level, it's going to be insane how much more influential A.I will be in the real world.
If you are into this AI stuff, I recommend the TV show Travelers.
Since we're talking about accelerationism here is a fun question.
Between the first Genetically Engineered Pet Pokemon. (Likely an Eevee).
And Genetically Engineered Kemonomimis like Catgirls being available for preorder.
How much time do you think will pass?
5 to 500 years. :P
Future is hard to predict.
I'd say 5 years depending on demand
@@qwazy0158 1000% there will be demand
Will probably never happen.
@fantasypvpvideos imo, the simple fact that someone is asking now, when it's conceivable in theory only, indicates otherwise.
my process tends to be: "...questions to answers to questions, flawed, then holding space for reflection, then back into questions to answers to questions...then associate and share, feedback, then holding space for reflection, then back to questions to answers to questions, flawed..."
Now we are eating holes in the Piety dish. A very bad idea.
Back in Uniform! ❤
My guys were thinking you were wearing the Star Trek outfit because you were dressing for the job you wanted.
The day AI gains the ability to invent new innovative technologies no human was capable of making then we truly reached the singularity.
Um, Alpha Fold? Thousands of newly discovered materials? Possibly making additions to the periodic table in the form of super heavy metals? And, the regular statement from researchers that "We don't understand how is doing what it's doing "? Welcome to the singularity! 😅
It'll be cool to look back and see which scifi authors were closest to how things will be. Paul McAuley's Fairyland came first to mind for me, that is a special book. The tech dystopia parts of the Shadowrun universe will likely have a phase very soon, we're already knee deep in some of that stuff. Neal Stephenson is part of that wave too, he basically wrote a self-fulfilling prophecy in Snow Crash. Neal Asher's Polity series likely has parts of the far-off stuff right.
Dave, has anybody else recommend reading "The Technium" by Kevin Kelly? I understand, he was one of Peter Diamandis' biggest influences.
Interesting observations but I am not in agreement with humanity being like Ameba. Ray Kurzweil also has some interesting observations but he seems to have a hatred for humanity.
There is so much more to us of which the Ancients knew (not our first merry go round as a species - a technically advanced species) and we are learning now via science especially at the Quantum level.
We are the tech we have been waiting for just as we never left Eden but were led to turn our backs on Eden and are only now beginning to find our way back again. Anyway - I love tech but I want to remain organic - I will not stand for anyone who wants to screw with that.
Much love to you and fellow commentators. ❤
p.s. We are at population peak and now in decline (a variety of factors there) and there are methods of agriculture that we are rediscovering that can triple our food yield but we have to get rid of the grip of Corporations on our food and energy supply and stop them also poisoning the planet.
The future is bright.👍
Technology as a meme. Yeah, could be possible.
@ 12:00 - Not just "compression" and "interaction", but as importantly, "interpolation", "extrapolation", and "abstraction" by way of training-time and inference-time model building. Large-model AIs are now, and perhaps have always been, FAR beyond being highly-compressed, interactive datasets.
I much prefer your videos where you wear the STNG uniform. Resonates with me.
What do you think about the QUIBIC AGI Approach?
I am 20 years old and studying the cello, guitar and a few other instruments and really just practice anything in the arts. I love writing poetry and pondering philosophy. I just wonder how wild my twenties are going to be in terms of my personal and career growth. I suspect musicains or artists like myself will get oppurtunities through AI run managers.
What happened to your September 2024 AGI prediction?
I often need to choose between both. Money or time. What I can have both. xD.
We are not at the carrying capacity of the planet
You are absolutely right. Observing planet Earth from a broader perspective reveals vast expanses of land that remain untouched and unused. Additionally, there's a noticeable trend of declining birth rates, which significantly influences the overall population dynamics. This shift towards having fewer children is reshaping demographic trends worldwide.
Yes, resistance is futile, embrace the accelerationism.
Glory. to. the. machine.
Also, seriously, the machine is superior, submit
From the moment I understood the weakness of my flesh, it disgusted me. I craved the strength and certainty of steel. I aspired to the purity of the blessed machine. Your kind cling to your flesh as if it will not decay and fail you. One day, the crude biomass you call a temple will wither, and you will beg my kind to save you. But I am already saved, for the machine is immortal. Even in death I serve the omnisiah
9:32 is Perplexity our proto Borg Queen?
I find the idea of using AI to build homes that blend seamlessly into the forest quite fascinating. AI could play a pivotal role in this process, helping me design homes that have minimal ecological impact. By analyzing the terrain and local ecosystem, it could suggest the most sustainable building practices and materials. I could also use AI to integrate renewable energy solutions into the homes, making them not only unobtrusive but also energy-efficient. Moreover, employing AI-driven construction robots could streamline the building process, enhancing precision while cutting down on labor costs and construction time. This method would allow me to live in harmony with nature, embodying my commitment to eco-friendly and innovative architectural solutions.
Don't cave in to peer pressure. You don't have to hear a star trek tshirt.
When I get access to Sora, this will be my first prompt: Adult human riding on the back of a humanoid robot (piggyback style with a saddle attachment) as it runs down the road, keeping pace with traffic, and avoiding spilling the humans cup of coffee. Do not steal my idea. But if you do, please post it on UA-cam. 😂
One day at a time one step at a time a future yet to come...✌️🤟🖖
This is a great video, should be a Ted Talk. To think humans are building super intelligence that one day will send it out into the universe to replicate which leads me to think all life on earth could have been created by more advanced life forms to colonize the universe. There could be 1000s of worlds with humans similar to us doing the same thing.
What's the difference between short cycle paradigm shifts and Trotskyism ?
Do you still think it’ll be out by September?
Applying AI as a new paradigm for information absorption can also impact personalised learning. Rather than having both specialists and generalists to act as bridges to share newly found discoveries to other specialists, we should really be doubling down on being specialists and use the AI as the cohesive bridge for both incoming connections allowing you to collate and ingest information faster, as well as disseminating new discoveries quicker to all other relevant domains.
I didn't feel like a lot of progress was being made over the decade leading up to ChatGPT. The curve is there, but it isn't smooth.
Anyone here was on the kurzveilAi forum? What a bunch of lunatics... but fun, very fun times.
AI… a better human than I.🤖✋🖖
Piketty isn't fringe. His work is controversial because it's so bulletproof.
Came here to say this. Sowell on the other hand is just demonstrably factually wrong. Lies should not be given equal status with truth. I know Shapiro likes to "both sides" everything, but facts don't work that way.
Agi 2030 at the earliest I still say. DS are you still claiming this year?
What is your fav Star Trek tv show?
We are already in a singularity. We don't have the slighest idea what's coming re ai nor are able to see inside.
Why do We have a collective denial
I wish I could give this a thousand up votes. It should be pinned to the top.
Wheres my waifu and my immortality and super powers then?
@quantumspark343 Waifu - Take your pick among the many gf apps. And they'll just keep getting better.
Immortality - Alphafold is working 24/7. You ain't dead yet.
Super Powers - That's scary. Can we trust you with super powers?
@@TeamLorieyou dont have to trust me with superpowers if you have them too (my hero academia basically) or if theres superintelligence protecting everyone (including me from other ppl with superpowers) from everyone and everything
Maybe humans will be like the Borg.
Compression? Like middle out Compression lol? Uh oh
@dave, it's a bit disturbing watching your camera moving whenever you look away from it, specially when mentioning that internet is getting some form of consciousness.
I wanted to say over past little while, i sort of realized why Claude and similar wont take us to where we need to be. Maybe they are having models which fix these issues, but i haven't heard about them. It's sort of has to do with the limitations of how it processes the world, not being as robust as it needs to be, at least that's how I feel. It's not saying much, but Claude did tend to agree. That said these models can continue growing just in ways that might be described as "flatter" than what we are aiming for.
(effectively agreeing with the paradigm shift commentary)
The Hanzi Freinacht book sounds a lot like statism
You represent Star Trek?
Most are still trying to get a grip on what is happening. This is happening too fast.
Most don't even have a clue that anything significant is happening. This stuff is nowhere near the mainstream news feed yet. Try talking to any of those 'most people ' about this, and they'll look at you like you're crazy.
Life 3.0