Artificial Intelligence thread

iewgnem

Junior Member
Registered Member
its not sheer speed nor energy... more about cost...

if lawyers charge $500/hr of billable work, but I can go on hyperbolic.xyz and use Llama 3.3 model for 20 cents per million token, guess which one most people are going to use?
Lawyer is a very bad analogy. Most legal work are boilerplate, the money you pay is the accountability against screwing up.
The same is true in many industries, accounting, general aviation, medicine...
 

tphuang

Lieutenant General
Staff member
Super Moderator
VIP Professional
Registered Member
I agree partially, since qwen is pumping out weights left and right

but US other trick is to keep EUV "inhouse" as itself and vassals.

Elon musk is building million GPU AI supercluster

High NA EUV is now the standard, 1.5nm process nodes coming soon. if China cant break the 3nm wall, the rest of the world stand zero chance

Models need infra to run inference, if USA chokeholds the gpus then they can control the market and set the terms of access...

Iff the AI is good enough, some say o1 pro already might be, then any country that doesnt use OAI gets left behind...if they get enough of head start, set the standards, and get enough momentum of rest of world to convert their apps, processes etc to all be infused with US AI via API etc then thats vendor lockin the way andriod, windows, apple have held hostage large swabs of the planet.

so naturally the idea is to have the world flock to US , instead of oil, America is now gatekeeper of intelligence

This has been the plan since 2018...

China needs to counter and take this more urgently.
The low level quality in your posts really astonish me sometimes.

the h100 rental cost is lower in China than in America so they clearly have no problem smuggling in GPUs. You don’t need 3nm chips for AI. Anything else you wrote is pretty much just alarmism. You keep going by like this and I am going to ban you from this thread.
 

9dashline

Captain
Registered Member
Please, Log in or Register to view URLs content!

So "Open"AI, who collaborates with the NSA and DoD, is critical of competition who actually makes their models "open", simply because these alternatives orginate from a country that was able to successfully counter a CIA color revolution coup attempt many decades ago.....

"Open"AI, charging 300x the cost of Qwen QwQ ($60 per million tokens vs 20cents ) actively hides the reasoning tokens that customers paid for at a 30000.00% inflation over truly open (Qwen QwQ that is free, open source, open weights, open reasoning etc) alternatives yet somehow feel like they are entitled to a moral high ground... and now they wants suckers to pay up $200/mo and were even teasing a $2000/mo subscription...

Good thing China is developing out modelscope, as Huawei found out the hard way with Andriod and GooglePlay, the Huggingface wont be reliable for too much longer
 

iewgnem

Junior Member
Registered Member
Please, Log in or Register to view URLs content!

So "Open"AI, who collaborates with the NSA and DoD, is critical of competition who actually makes their models "open", simply because these alternatives orginate from a country that was able to successfully counter a CIA color revolution coup attempt many decades ago.....

"Open"AI, charging 300x the cost of Qwen QwQ ($60 per million tokens vs 20cents ) actively hides the reasoning tokens that customers paid for at a 30000.00% inflation over truly open (Qwen QwQ that is free, open source, open weights, open reasoning etc) alternatives yet somehow feel like they are entitled to a moral high ground... and now they wants suckers to pay up $200/mo and were even teasing a $2000/mo subscription...

Good thing China is developing out modelscope, as Huawei found out the hard way with Andriod and GooglePlay, the Huggingface wont be reliable for too much longer
$200/mo for 4o is basically an admission that they're in deep s* both financially and technologically.
 

luminary

Senior Member
Registered Member
The low level quality in your posts really astonish me sometimes.

the h100 rental cost is lower in China than in America so they clearly have no problem smuggling in GPUs. You don’t need 3nm chips for AI. Anything else you wrote is pretty much just alarmism. You keep going by like this and I am going to ban you from this thread.
I'm very sorry that you have to deal with these posts.

I've worked with a number of OpenAI "top devs" and they are absolutely EA rationalist nuts who spout AGI crap all day.

When you know that these people mostly had inhabited hackernews, 4chan and reddit before they became self-proclaimed AI god messihas, everything makes a lot more sense. Basically losers and basement nerds going on a massive power trip/ superiority complex, and not being able to handle it like adults since they're just manchildren.

What's worse is many of them believe other tech folk and people in general are useless chaff soon to be brushed aside by the AI god they will create. They are bizarrely hostile to anybody not part of the cult and the mission because they believe they are at odds with humankind and organic life itself.


...
On a more positive front, I'm happy to note that pretty much everything we predicted in this forum a year ago has come to pass.

People with level heads have made the right calls. Doomers and concern trolls have come and gone.

March 2023
Plateau of data:
I actually think the US is approaching the limit for what kinds of AI they can develop. The internet seems to be the single biggest source of data they can train their AI on. From here on for them, it's all about making AI work more efficiently with smaller datasets, self-classification and reinforcement, novel model architecture, and possibly generating "life-like" fake data. Will they be able to sustain a positive feedback cycle with these methods? Or will they hit the "simply not enough data to work" development wall? Or (imo likely) end up with a fake, marketing-gimmick "AGI"? It's anybody's guess.
Narrow AI (months before Ren Zhengfei comments):
artificial intelligence is not tied to software development at all, but rather hardware development: whoever can produce and connect the most IoT devices to attain valuable real life data.

When I frame the AI "race" in terms of production and structural implementation, who is leading becomes obvious. Making a neural network that analyzes daily consumption trends to optimize power grid operation for millions of people is useful. Making a chatbot that imitates redditors or used car salesmen is useful... for a nation of bellicose politicians.
Strategy:
As of now, China is a few years ahead of US in IoT tech and decades ahead in implementation. Now is the time to apply strong competition and financial pressure to Silicon Valley from all sides. Due to investor expectations, FAANG are backed into a corner and must do all sorts of unreasonable things to maintain a high profit level. If you squeeze their R&D budget, poach their talent, then this tech gap will widen.
 
Last edited:

caudaceus

Senior Member
Registered Member
I've worked with a number of OpenAI "top devs" and they are absolutely EA rationalist nuts who spout AGI crap all day
Yeah I can think as an analogy that we have Industrialists soy jacking over discovery of steam engine and have a wild dream of making God out of machine. You can't make the Deus Ex Machina but focusing development of Mechanical Engineering for productivity increase is much better goal.
 
Top