>it's increasing at a rapid pace >it's plateauing we're at the top of the curve >there's a breakthrough every week >there's been no breakthroughs for years >it's accelerating despite there being no breakthroughs, scaling is all we need >GPT-4 is already showing sparks of AGI because of its ability to improve itself >GPT-4 is showing diminishing returns in comparison to 3
Its conflicting because we're near the top of the curve. The exact same thing happened with self driving cars, in a space of like 6 months we went from nothing to near self driving, and years later we're still working that out.
Anyone with any foresight is saying we're plateauing, everyone else just see's the moment to moment advancements and thinks it must continue this way forever
The self driving car problem is because the last 0.01% of safety guarantees are the hardest but are necessary to make them viable. The improvements in autonomous self driving are still improving rapidly
To think we're plateuing is silly. I do agree mere LLMs will get harder and harder to improve as we have been training them, but they're nowhere near the ceiling in what they could achieve. The number of papers, research, investment is increasing exponentially.
> There‘s no obvious plateau or major roadblock ahead rn
Of course theres not, we haven't hit it yet.
I've seen this happen countless times before and to be honest fell for them for the exact same reasons
I think this will be far bigger than all those times, its certainly the biggest technological advancement since the internet, but I doubt it'll reach its full potential in the next 10 years.
But advancements are accelerating. GPT-4 is already exhibiting behaviours GPT-3.5 just didn't have. We're absolutely nowhere near levelling off.
8 months ago
Anonymous
GPT-2 to GPT-3 took only 1 year. GPT-3 to GPT-4 took 3 years and orders of magnitude more effort, data and compute. It's going to take 7-8 years for a release as much better as GPT-4 is than GPT-3 wrt GPT-4.
8 months ago
Anonymous
Not really. GPT-4 was ready about a year before seeing release just for safety measures and the technology is much more understood now. Plus incentives for it only increases exponentially as the technology advances and is more and more adopted. Saying GPT-5 will be in the public in less than 2 years is a very highly likely scenario, some speculate even this year there might be a working version of it.
8 months ago
Anonymous
OpenAI was still just a relatively big startup company in that time period. It was a genuine legit concern where and how they'd get the resources to continue what they're doing. It was only after Chatgpt released that they became known to everyone on the planet and were acquired by microsoft for 10 billion. Now they have unlimited funding and resources definitively secured for their next projects
8 months ago
Anonymous
>10 billion >unlimited
Lol the execs just wanted to cash out so they can pump and dump. Microsoft will integrate ChatGPT into Bing, increase their market value, cash out, reveal LLMs suck, OpenAI execs sell their stock, buy a few new houses, then jump onto the next tech fad. Rinse and repeat. I feel very bad for people who can't see this shit.
8 months ago
Anonymous
Sounds a bit delusional to me. The 10 billion was just the start and now virtually every company is clamoring to get a taste of gpt4 access which openai sells at a premium since they're the only real game in town. Their workforce has only expanded since all this chatgpt stuff started too
>it's increasing at a rapid pace >it's plateauing we're at the top of the curve
We are getting larger models, the number of papers released per month is increasing, the amount of funding into AI is increasing, and the increase in abilities over time is increasing. You be the judge
>there's been no breakthroughs for years
Hilariously false
>it's accelerating despite there being no breakthroughs, scaling is all we need
It can accelerate if there's no breakthroughs, just with the knowledge we have now put into action. But there will almost certainly keep being new breakthroughs.
>GPT-4 is already showing sparks of AGI because of its ability to improve itself >GPT-4 is showing diminishing returns in comparison to 3
3 is dogshit compared to 3.5, which is dogshit compared to 4. Watch this lecture, form your own opinions https://www.youtube.com/watch?v=qbIk7-JPB2c
It cleared up a lot of stuff for me. Basically, a tiny handful of innovations could launch us into AGI and beyond within the next year, and one way or another, superintellegnce is on the horizon. Even a modest jump over GPT4 will be relatively earth-shaking if the trend of emergent properties continue.
That's if those innovations happen and if those emergent property trends stay consistent though, it'll be interesting to see. Really scary and nerve wracking but there's nothing we can do to stop it so I guess cross our fingers, and maybe have an exit strategy handy.
>it's increasing at a rapid pace
correct >it's plateauing we're at the top of the curve
asserted by trolls who want attention or legit delusional copers >there's a breakthrough every week
true though it depends on what you call a breakthrough >there's been no breakthroughs for years
very easily provably false >it's accelerating despite there being no breakthroughs, scaling is all we need
True but there have been breakthroughs >GPT-4 is already showing sparks of AGI because of its ability to improve itself
True >GPT-4 is showing diminishing returns in comparison to 3
Perhaps true but only in relation to improvement per compute hour, there is still a much higher ceiling with more compute, and GPUs are still improving
Has anyone asked Ted Kaczynski what he thinks about AI?
He apparently enjoys writing people. I asked GPT-4 for a summary of his manifesto and thought it would be interesting to see what he thought about it being tokenized forever in an AI so he's part of the system. If no one has, I may do it
If I want to write to or visit Ted Kaczynski in prison, could I?
Q: What are the main points of Ted Kaczynski's manifesto?
GPT-4: Ted Kaczynski's manifesto, also known as the "Unabomber Manifesto," is a 35,000-word essay titled "Industrial Society and Its Future." In this essay, Kaczynski presents his views on the problems of modern society and offers his solutions.
Here are some of the main points of Kaczynski's manifesto:
Industrialization and technological progress have led to the degradation of human freedom, happiness, and dignity.
The modern world is characterized by the rise of a technological elite that manipulates and controls society through the use of technology and the media.
The pursuit of technological progress has created a society that is unsustainable and will inevitably collapse.
The solution to these problems is to abolish the industrial-technological system and return to a more primitive way of life.
Kaczynski advocates for a radical form of revolution that involves the use of violence and terrorism to bring about the collapse of the system.
It is worth noting that Kaczynski's ideas are controversial and have been widely criticized. His violent tactics in particular have been condemned by many.
He doesn't use the internet though so he's probably not up on it and has RMS level autistic opinions so probably not. He could really shine with a Ted K AI spouting his revolutionary ideas, just not the killing part
I unironically sharted my pants regarding AI.
This stuff is so fucking bad, like seriously, i can kms right here. Its forseeable AI will kill off tremendous amounts of jobs short,mid & longterm. I might as well apply to work at a landfill or cemetery for any job security
Advances in development are outpacing adoption by the public and readily available use cases. Stable Diffusion can replace a commercial artist RIGHT NOW but it hasn't yet because cultural inertia naturally makes things slow. My mom still doesn't have a cell phone, you expect adoption of AI tech to be instant and implemented?
And that mechanical parrot is going to change the world.
I am baffled by how short-minded Bot.info can be sometimes. We literally have a GPT (general purpose technology) in our face and half the board doesn't even understands what is going on.
People understand how powerfully disruptive GPT is, they know things are going to change fast (not super fast mind you, change will be bottlenecked by adoption speed and cultural inertia), they just have strong doubts about the heckin ghost in the machine we're so close to AGI narrative.
"General purpose technology (GPT) refers to a type of technology that has the potential to significantly impact and transform multiple industries and aspects of society. GPTs are characterized by their wide scope for improvement and use, and their ability to spur complementary innovations. Examples of GPTs include the steam engine, electricity, and the internet."
People understand how powerfully disruptive GPT is, they know things are going to change fast (not super fast mind you, change will be bottlenecked by adoption speed and cultural inertia), they just have strong doubts about the heckin ghost in the machine we're so close to AGI narrative.
I disagree in some points. In particular, I think adoption will be near instant, it already has broken records in it's adoption as we speak. There's just too much incentive by big corporations and I am not just talking about jobs and expenses. Research and development, marketing, customer assistance, legal counsel, administration, ect will see a huge impact this year alone. Essentially see it this way, AI reduces significantly the cost of cognitive labor and increases the productivity just to top it off. This means you don't just get the same work for less, this means you get a lot more work for a lot less. This will be even more apparent once it's fully automated which is already being worked on. Corporations (and people) will essentially be able to have a worker that doesn't eat, sleeps, gets tired or sick and is able to complete many (useful) tasks as good or better than humans.
I honestly can go on and on about this. It really feels unreal the more I learn about it.
It really feels like it could be on par with the industrial revolution. Going from the majority of the world living in tiny stone and wood hovels in small communities, primarily trading handmade goods and farming year round, traveling by horse and buggy, etc, to megacities and skyscrapers, cars and planes, factories, shopping malls, radio, etc. All in a single lifetime. The equivalent very well could be happening now.
You're crazy if you think it's merely going to be *on par*. Assuming no AGI apocalypse, the scientific advancements alone are going to be rapid, exponential, and far reaching. I'd say we might see interplanetary commercial travel in our lifetime but anything that requires physical infrastructure, resources and manufacturing time to build and isn't purely informational will lag behind considerably the same way we're going to have AI much earlier than embodied autonomous robots.
8 months ago
Anonymous
I get what you mean, but I'm more trying to imply the idea of dragging someone from the 1800s into the 1850s and saying "this is what it will be like in 50 years." The world would be borderline unrecognizable.
And it might be again, albeit even MORE unrecognizable.
8 months ago
Anonymous
On a 50 year timescale? Absolutely. On a 10-15 year one? Maybe not, because ultimately the industrial revolution was about physical things like resources and materials and manufacturing methods and changed how we constructed houses and made machines and all kinds of things. The AI stuff is going to be absolutely insane, but there's going to be a delay waiting for the informational elements to bleed into physical life. An AGI isn't going to change what the rake I use to tend my yard looks like for instance. It might make vehicles look dramatically different in just one manufacturing year but not everyone can afford to switch vehicles immediately.
The industrial revolution is actually a good comparison. Increased automation brought about by coal and steam power lead to increased adaptation of automating technologies which lead to more money, opportunities, and interest in automation, creating a feedback loop. If such a thing is going to occur, then we are barely at the beginning of the feedback loop.
only if the corps and startups keep hoarding the good shit and paywall the peasants out, hopefully some autist will not let this happen, but given the absolute state of voicegen/llms right now it's unlikely to improve in the short term
>The blockhain will make databases obsolete
>AI will make programming obsolete
It's increasing at a rapid pace.
Q2 2023 will make Q1 2023 look like 2019
>it's increasing at a rapid pace
>it's plateauing we're at the top of the curve
>there's a breakthrough every week
>there's been no breakthroughs for years
>it's accelerating despite there being no breakthroughs, scaling is all we need
>GPT-4 is already showing sparks of AGI because of its ability to improve itself
>GPT-4 is showing diminishing returns in comparison to 3
I keep getting so much conflicting information.
Its conflicting because we're near the top of the curve. The exact same thing happened with self driving cars, in a space of like 6 months we went from nothing to near self driving, and years later we're still working that out.
Anyone with any foresight is saying we're plateauing, everyone else just see's the moment to moment advancements and thinks it must continue this way forever
The self driving car problem is because the last 0.01% of safety guarantees are the hardest but are necessary to make them viable. The improvements in autonomous self driving are still improving rapidly
To think we're plateuing is silly. I do agree mere LLMs will get harder and harder to improve as we have been training them, but they're nowhere near the ceiling in what they could achieve. The number of papers, research, investment is increasing exponentially.
I wouldn't say the last 0.01%, more like the last 10%, but thats pretty much my point, the last 10% of the work takes 90% of the time.
Of course things are increasing exponentially right now, we haven't done the first 90% yet. I'd say we're ~70% of the way there.
> Its different this time bro.
> It‘s a new paradigm!
No seriously it absolutely fucking is. The motherfucker is only accelerating its scary. There‘s no obvious plateau or major roadblock ahead rn
> There‘s no obvious plateau or major roadblock ahead rn
Of course theres not, we haven't hit it yet.
I've seen this happen countless times before and to be honest fell for them for the exact same reasons
I think this will be far bigger than all those times, its certainly the biggest technological advancement since the internet, but I doubt it'll reach its full potential in the next 10 years.
But advancements are accelerating. GPT-4 is already exhibiting behaviours GPT-3.5 just didn't have. We're absolutely nowhere near levelling off.
GPT-2 to GPT-3 took only 1 year. GPT-3 to GPT-4 took 3 years and orders of magnitude more effort, data and compute. It's going to take 7-8 years for a release as much better as GPT-4 is than GPT-3 wrt GPT-4.
Not really. GPT-4 was ready about a year before seeing release just for safety measures and the technology is much more understood now. Plus incentives for it only increases exponentially as the technology advances and is more and more adopted. Saying GPT-5 will be in the public in less than 2 years is a very highly likely scenario, some speculate even this year there might be a working version of it.
OpenAI was still just a relatively big startup company in that time period. It was a genuine legit concern where and how they'd get the resources to continue what they're doing. It was only after Chatgpt released that they became known to everyone on the planet and were acquired by microsoft for 10 billion. Now they have unlimited funding and resources definitively secured for their next projects
>10 billion
>unlimited
Lol the execs just wanted to cash out so they can pump and dump. Microsoft will integrate ChatGPT into Bing, increase their market value, cash out, reveal LLMs suck, OpenAI execs sell their stock, buy a few new houses, then jump onto the next tech fad. Rinse and repeat. I feel very bad for people who can't see this shit.
Sounds a bit delusional to me. The 10 billion was just the start and now virtually every company is clamoring to get a taste of gpt4 access which openai sells at a premium since they're the only real game in town. Their workforce has only expanded since all this chatgpt stuff started too
>it's increasing at a rapid pace
>it's plateauing we're at the top of the curve
We are getting larger models, the number of papers released per month is increasing, the amount of funding into AI is increasing, and the increase in abilities over time is increasing. You be the judge
>there's been no breakthroughs for years
Hilariously false
>it's accelerating despite there being no breakthroughs, scaling is all we need
It can accelerate if there's no breakthroughs, just with the knowledge we have now put into action. But there will almost certainly keep being new breakthroughs.
>GPT-4 is already showing sparks of AGI because of its ability to improve itself
>GPT-4 is showing diminishing returns in comparison to 3
3 is dogshit compared to 3.5, which is dogshit compared to 4. Watch this lecture, form your own opinions https://www.youtube.com/watch?v=qbIk7-JPB2c
fantastic video
It cleared up a lot of stuff for me. Basically, a tiny handful of innovations could launch us into AGI and beyond within the next year, and one way or another, superintellegnce is on the horizon. Even a modest jump over GPT4 will be relatively earth-shaking if the trend of emergent properties continue.
That's if those innovations happen and if those emergent property trends stay consistent though, it'll be interesting to see. Really scary and nerve wracking but there's nothing we can do to stop it so I guess cross our fingers, and maybe have an exit strategy handy.
im really excited for it. i hope governments dont cuck out and nerf ai because we really do need this as an overlord
>it's increasing at a rapid pace
correct
>it's plateauing we're at the top of the curve
asserted by trolls who want attention or legit delusional copers
>there's a breakthrough every week
true though it depends on what you call a breakthrough
>there's been no breakthroughs for years
very easily provably false
>it's accelerating despite there being no breakthroughs, scaling is all we need
True but there have been breakthroughs
>GPT-4 is already showing sparks of AGI because of its ability to improve itself
True
>GPT-4 is showing diminishing returns in comparison to 3
Perhaps true but only in relation to improvement per compute hour, there is still a much higher ceiling with more compute, and GPUs are still improving
its a gptrolley weekend, there will be a similar ai type thing every weekend from now until the singularity
>what went wrong?
'"guardrails'"
Most tech crazies are just hype fueled pyramid scams with little actual results
Has anyone asked Ted Kaczynski what he thinks about AI?
He apparently enjoys writing people. I asked GPT-4 for a summary of his manifesto and thought it would be interesting to see what he thought about it being tokenized forever in an AI so he's part of the system. If no one has, I may do it
If I want to write to or visit Ted Kaczynski in prison, could I?
https://www.quora.com/If-I-want-to-write-to-or-visit-Ted-Kaczynski-in-prison-could-I
Q: What are the main points of Ted Kaczynski's manifesto?
GPT-4: Ted Kaczynski's manifesto, also known as the "Unabomber Manifesto," is a 35,000-word essay titled "Industrial Society and Its Future." In this essay, Kaczynski presents his views on the problems of modern society and offers his solutions.
Here are some of the main points of Kaczynski's manifesto:
Industrialization and technological progress have led to the degradation of human freedom, happiness, and dignity.
The modern world is characterized by the rise of a technological elite that manipulates and controls society through the use of technology and the media.
The pursuit of technological progress has created a society that is unsustainable and will inevitably collapse.
The solution to these problems is to abolish the industrial-technological system and return to a more primitive way of life.
Kaczynski advocates for a radical form of revolution that involves the use of violence and terrorism to bring about the collapse of the system.
It is worth noting that Kaczynski's ideas are controversial and have been widely criticized. His violent tactics in particular have been condemned by many.
do it
He doesn't use the internet though so he's probably not up on it and has RMS level autistic opinions so probably not. He could really shine with a Ted K AI spouting his revolutionary ideas, just not the killing part
It left off all the parts where he trashed leftists. It's literally like half the manifesto.
It left a lot out. Nothing about robots. It's a summary.
I unironically sharted my pants regarding AI.
This stuff is so fucking bad, like seriously, i can kms right here. Its forseeable AI will kill off tremendous amounts of jobs short,mid & longterm. I might as well apply to work at a landfill or cemetery for any job security
>AI will kill off tremendous amounts of jobs
If you only knew how bad it was
Explain more.
it has literally just started
Advances in development are outpacing adoption by the public and readily available use cases. Stable Diffusion can replace a commercial artist RIGHT NOW but it hasn't yet because cultural inertia naturally makes things slow. My mom still doesn't have a cell phone, you expect adoption of AI tech to be instant and implemented?
Ai is literally a mechanical parrot
And that mechanical parrot is going to change the world.
I am baffled by how short-minded Bot.info can be sometimes. We literally have a GPT (general purpose technology) in our face and half the board doesn't even understands what is going on.
People understand how powerfully disruptive GPT is, they know things are going to change fast (not super fast mind you, change will be bottlenecked by adoption speed and cultural inertia), they just have strong doubts about the heckin ghost in the machine we're so close to AGI narrative.
how can you expect people to listen to you when you don't even know what GPT stands for
"General purpose technology (GPT) refers to a type of technology that has the potential to significantly impact and transform multiple industries and aspects of society. GPTs are characterized by their wide scope for improvement and use, and their ability to spur complementary innovations. Examples of GPTs include the steam engine, electricity, and the internet."
I disagree in some points. In particular, I think adoption will be near instant, it already has broken records in it's adoption as we speak. There's just too much incentive by big corporations and I am not just talking about jobs and expenses. Research and development, marketing, customer assistance, legal counsel, administration, ect will see a huge impact this year alone. Essentially see it this way, AI reduces significantly the cost of cognitive labor and increases the productivity just to top it off. This means you don't just get the same work for less, this means you get a lot more work for a lot less. This will be even more apparent once it's fully automated which is already being worked on. Corporations (and people) will essentially be able to have a worker that doesn't eat, sleeps, gets tired or sick and is able to complete many (useful) tasks as good or better than humans.
I honestly can go on and on about this. It really feels unreal the more I learn about it.
It really feels like it could be on par with the industrial revolution. Going from the majority of the world living in tiny stone and wood hovels in small communities, primarily trading handmade goods and farming year round, traveling by horse and buggy, etc, to megacities and skyscrapers, cars and planes, factories, shopping malls, radio, etc. All in a single lifetime. The equivalent very well could be happening now.
You're crazy if you think it's merely going to be *on par*. Assuming no AGI apocalypse, the scientific advancements alone are going to be rapid, exponential, and far reaching. I'd say we might see interplanetary commercial travel in our lifetime but anything that requires physical infrastructure, resources and manufacturing time to build and isn't purely informational will lag behind considerably the same way we're going to have AI much earlier than embodied autonomous robots.
I get what you mean, but I'm more trying to imply the idea of dragging someone from the 1800s into the 1850s and saying "this is what it will be like in 50 years." The world would be borderline unrecognizable.
And it might be again, albeit even MORE unrecognizable.
On a 50 year timescale? Absolutely. On a 10-15 year one? Maybe not, because ultimately the industrial revolution was about physical things like resources and materials and manufacturing methods and changed how we constructed houses and made machines and all kinds of things. The AI stuff is going to be absolutely insane, but there's going to be a delay waiting for the informational elements to bleed into physical life. An AGI isn't going to change what the rake I use to tend my yard looks like for instance. It might make vehicles look dramatically different in just one manufacturing year but not everyone can afford to switch vehicles immediately.
AI parrots globohomo
Wtf would i give a shit about the opinion of globohomo?
it isn't
wise man
there is no hype because ai is everywhere now.
The industrial revolution is actually a good comparison. Increased automation brought about by coal and steam power lead to increased adaptation of automating technologies which lead to more money, opportunities, and interest in automation, creating a feedback loop. If such a thing is going to occur, then we are barely at the beginning of the feedback loop.
Fuck you talking about. Every AI related service is constantly inundated with traffic
Just wait for the next breakthrough
its only beginning peasant
only if the corps and startups keep hoarding the good shit and paywall the peasants out, hopefully some autist will not let this happen, but given the absolute state of voicegen/llms right now it's unlikely to improve in the short term
its the fault of alttroon endless shilling
AGI in two weeks anon!
crypto>AR>VR>NFT>AI