Are you for or against Data Centers?

The model and programming is the most important part of AI IMHO.

Most of the physical stuff can be copied, same as software, but having the most advanced model and it's capabilities are the secret part.

Sure, it can be stolen, but I bet this information is protected as much as other nation secrets.

The AI race isn't just about chips, it's about model development, efficiency, and capability.
Yes I understand that, but China, having full control of the manufacturing process of the microchips, processors and whole electronic systems, as had happened in the past, can provide a backdoor access for them to do all sorts of nefarious things.

If they can hinder the hardware in some way, our most advanced models may not have the necessary compute power to give us an edge, for example.

It’s not like we don’t have an example of this. It is well known today that a lot of the Nazi Germanys production problems were caused by sabotage of the slave laborers. The most advanced tank, the Tiger and Tiger II could scarcely be deployed because of this sabotage, and many that were deployed broke down because the internals were sabotaged. This happened to almost all of the German equipment.
 
We need a boogeyman so China is the AI boogeyman. If AI is a National Security issue, then the military can build the datacenters, and we the people will own them. We can sell tokens back to big tech at a profit in the meantime. Between water, power and infrastructure we the people are mostly paying for it already.
Oh come on, you’re just being gullible 🤣

I totally agree by the way.
 
The model and programming is the most important part of AI IMHO.
I have said this for a while. Same as .com. Lucent and Intel did not win, Google and Amazon did. They had the software models to use the internet.

So given this, are companies really going to run their IP on a huge LLM built by big tech?

Or are they going to build there own application specific models trained using there own data on their own servers?
 
We need a boogeyman so China is the AI boogeyman. If AI is a National Security issue, then the military can build the datacenters, and we the people will own them. We can sell tokens back to big tech at a profit in the meantime. Between water, power and infrastructure we the people are mostly paying for it already.
No we don't "need" any boogeyman. We do need to be vigilant when a hostile nation poses as Americans on line, then floods social media with their propaganda to accomplish their goal of hurting the USA for the sake of their dominance. This is how China works. Of course they don't want a full blown military war with the USA, what they do is try to poison us with fentanyl (via Mexico), send massive numbers of military age male "immigrants" into the US, buys up land directly adjacent to our military bases, and pit Americans against Americans, and so on. Anyone who thinks China is an innocent country that wouldn't hurt the USA is beyond stupid.
 
No we don't "need" any boogeyman. We do need to be vigilant when a hostile nation poses as Americans on line, then floods social media with their propaganda to accomplish their goal of hurting the USA for the sake of their dominance. This is how China works. Of course they don't want a full blown military war with the USA, what they do is try to poison us with fentanyl (via Mexico), send massive numbers of military age male "immigrants" into the US, buys up land directly adjacent to our military bases, and pit Americans against Americans, and so on. Anyone who thinks China is an innocent country that wouldn't hurt the USA is beyond stupid.
No one here thinks the CCP are a bunch of good dude's.

Please however explain to me why I should subsidize a big tech datacenter in my neighbourhood because of it.

Or, why its OK for CEO's to enrich themselves by firing there workers and moving production to China, but I am not allowed to buy a Cheap Chinese EV?

Please keep it simple for me. Its Monday and I am not the sharpest tool in the shed on the best of days.
 
CHINA has a HUGE advantage over the USA in one way. Unlimited resources to power AI centers, what they do not have is the quality of chips that we have.

"
How fast is AI growing in China?

A key metric is how much power new data centers will consume to support AI. Electricity capacity for data centers in China is on course to jump 30% this year, to 30 gigawatts. One top cloud computing company and a major AI player plans to increase its data center capacity 10 times by 2032. "

https://www.goldmansachs.com/insigh...llars-in-data-centers-amid-overseas-expansion

"China’s abundant supply of cheap electricity is a key advantage in the rollout of data centres needed to run AI models."
https://www.aljazeera.com/economy/2...eapon-in-ai-race-with-us-lots-of-cheap-energy

"US tech leaders, including Tesla’s Elon Musk, Nvidia’s Jensen Huang and OpenAI’s Sam Altman, have openly acknowledged China’s edge in the energy domain."

Lets not forget the 70 nuclear power plants planned and under construction in China
Time for the USA to wake up.
 
Last edited:
No one here thinks the CCP are a bunch of good dude's.

Please however explain to me why I should subsidize a big tech datacenter in my neighbourhood because of it.

Or, why its OK for CEO's to enrich themselves by firing there workers and moving production to China, but I am not allowed to buy a Cheap Chinese EV?

Please keep it simple for me. Its Monday and I am not the sharpest tool in the shed on the best of days.
Subsidize? Do you also subsidize the rest of the infrastructure you use? Infrastructure is not just roads:

2. Digital & Modern Infrastructure
Driven by technological advances, this category supports modern connectivity and innovations. [1]

  • Telecommunications: Broadband internet, fiber-optic networks, 5G connections, and data centers.
  • Smart Cities: Traffic management systems, smart grids, and Internet of Things (IoT) sensors.
  • Mobility Infrastructure: Electric vehicle (EV) charging networks.
CEO's? There are good ones and bad ones in ANY industry. So do you want ALL industry shut down? Offshoring has happened for literally decades. Any outrage about "American" cars being built all around the globe?

The anti-data center crowd does not want the DC's, but will be the first to scream if they don't have the digital capability needed in the future. Its a NIMBY thing. Everyone will want the product, but they will protest the factory that build the product.
 
Yes I understand that, but China, having full control of the manufacturing process of the microchips, processors and whole electronic systems, as had happened in the past, can provide a backdoor access for them to do all sorts of nefarious things.
Yep, there is an NYT article where a hidden layer of circuitry, sandwiched inside a PCB, not visible without X-ray imaging, was discovered by researchers.

Apparently one of the people at a fab in China was compromised to change the board design.

This likely happens in all tech but is harder to see with small devices and chips along with code.

It's likely we are aware of this and have countermeasures and QC, but it can also be a cat and mouse game.

No we don't "need" any boogeyman. We do need to be vigilant when a hostile nation poses as Americans on line, then floods social media with their propaganda to accomplish their goal of hurting the USA for the sake of their dominance. This is how China works. Of course they don't want a full blown military war with the USA, what they do is try to poison us with fentanyl (via Mexico), send massive numbers of military age male "immigrants" into the US, buys up land directly adjacent to our military bases, and pit Americans against Americans, and so on. Anyone who thinks China is an innocent country that wouldn't hurt the USA is beyond stupid.
Asymmetric warfare is their MO. Using our freedoms against us to influence the public and utilize useful-idiots i.e. extreme tree huggers and anti-humanists to propagate their agenda.

China has the benefit of having a long term vision not influenced by elections. Supposedly many of the decision makers are engineers and highly educated unlike any joe-schmo that can get elected here based on a popularity contest.
 
So given this, are companies really going to run their IP on a huge LLM built by big tech?
That's one of the concerns with OpenAI. Many do not trust Sam Altman and are afraid of him copying their IP, but thats for the consumer level products.

There are supposedly air-gapped models that can run locally or on AWS like Bedrock. These services allow you to use the AI model but it does not communicate with the mothership.
 
Asymmetric warfare is their MO. Using our freedoms against us to influence the public and utilize useful-idiots i.e. extreme tree huggers and anti-humanists to propagate their agenda.

China has the benefit of having a long term vision not influenced by elections. Supposedly many of the decision makers are engineers and highly educated unlike any joe-schmo that can get elected here based on a popularity contest.
Exactly. They know how gullible and stupid some Americans are. People get played by China and never even know it.

I am glad you and I and some others see through it.
 
That's one of the concerns with OpenAI. Many do not trust Sam Altman and are afraid of him copying their IP, but thats for the consumer level products.

There are supposedly air-gapped models that can run locally or on AWS like Bedrock. These services allow you to use the AI model but it does not communicate with the mothership.
Many companies are reporting that they have already used there annual budget of tokens. So all there estimates on productivity were based on a lot less input costs.

Also a lot of discussion now about how you need experts on how to actually frame your inputs so you don't use as much compute. :ROFLMAO:

Member here @Pew I believe is the correct one - has posted about hosting his own AI model. I am looking into it also for fun but I am so far behind it might be futile. Either way If this is truly the future I can't imagine a whole bunch of companies off-loading it to big tech. Its just like ERP and manufacturing systems like SAP. Some want turn key hosted, but most want to implement and own there own instance.
 
I have said this for a while. Same as .com. Lucent and Intel did not win, Google and Amazon did. They had the software models to use the internet.

So given this, are companies really going to run their IP on a huge LLM built by big tech?

Or are they going to build there own application specific models trained using there own data on their own servers?
I will tell you what I think might be the reason:

It is too risky for a small company to build something that cost hundreds of billions unless they know there is no risk. If someone already build it, and is not your competitor, then it might be a better idea to "license" or "rent" from them to use it. Or if they are too much more powerful and they would want to acquire you, probably selling your company to them for a good price is a good idea.

This is like asking in the 90s why every company build computer using Windows and Intel chips, instead of building their own processor and write their own OS. The reason is the ecosystem is there and you likely can't compete directly. You would sell a software to run on those instead of forcing customers to buy your entire system home grown.

I can see a concern of Open AI being a monopoly. That was the assumption until a few months ago when suddenly they got a competitor that beat them (Anthropic). Seems like going forward there is no more monopoly in LLM, and smaller companies are better off focus on their own strength instead of building their own LLM, and just license others' LLM as a supplement.

There are a lot of licensing in the tech industry, you just didn't look deep down enough to know them. i.e. every USB port you have is powered by the same core licensed probably from NEC using Cadence as a middle man, or something like that.
 
Exactly. They know how gullible and stupid some Americans are. People get played by China and never even know it.

I am glad you and I and some others see through it.
Somewhere on the other side of the world there are people doubled over laughing at this, high-fiving their team.

EVERYONE thinks *they're* the ones who see through the propaganda and the manipulation; the others are the stupid ones falling for "it", whatever "it" is.
 
I will tell you what I think might be the reason:

It is too risky for a small company to build something that cost hundreds of billions unless they know there is no risk. If someone already build it, and is not your competitor, then it might be a better idea to "license" or "rent" from them to use it. Or if they are too much more powerful and they would want to acquire you, probably selling your company to them for a good price is a good idea.

This is like asking in the 90s why every company build computer using Windows and Intel chips, instead of building their own processor and write their own OS. The reason is the ecosystem is there and you likely can't compete directly. You would sell a software to run on those instead of forcing customers to buy your entire system home grown.

I can see a concern of Open AI being a monopoly. That was the assumption until a few months ago when suddenly they got a competitor that beat them (Anthropic). Seems like going forward there is no more monopoly in LLM, and smaller companies are better off focus on their own strength instead of building their own LLM, and just license others' LLM as a supplement.

There are a lot of licensing in the tech industry, you just didn't look deep down enough to know them. i.e. every USB port you have is powered by the same core licensed probably from NEC using Cadence as a middle man, or something like that.
I am obviosly just speculating.

For sure if I am going to use a LLM for basic stuff like agents that summarize my email and manage my calendar, I am going to just use someone else's LLM who can implement it at scale cheaper than I can. The question becomes is the savings I get from my agents worth the compute cost. Currently most users are saying no, but its early of course.

But that is never going to pay for AI.

I can't imagine a company with truly proprietary data uploading it up to Sam Altman or Jeff Bezos, no matter how many bibles they swear on that they will keep it private. These are the same companies scraping the internet and ignoring every copyright and trademark out there.

Then there are those that just don't want garbage in / garbage out. So they will want to use there own trained model. They might be OK running it on someone else's hardware - but then the question becomes is it really cheaper at that point. All the scaling cost benefit is lost anyway.

I think the idea that some handful of big tech companies are going to take over everyone's IP is far fetched, just from working with and for a bunch of smaller firms with lots of proprietary data. Maybe they should just sell Altman there data and shut down?
 
Member here @Pew I believe is the correct one - has posted about hosting his own AI model. I am looking into it also for fun but I am so far behind it might be futile. Either way If this is truly the future I can't imagine a whole bunch of companies off-loading it to big tech. Its just like ERP and manufacturing systems like SAP. Some want turn key hosted, but most want to implement and own there own instance.

Negative, it's @dogememe that's hosting his own. I've tried it once last year but I didn't spend now than a few hours on it.
 
Either way If this is truly the future I can't imagine a whole bunch of companies off-loading it to big tech.

I think this will look like web hosting. You can have your own infra to do this but it will be cheaper to pay AWS who has low cost highly reliable service due to scale.

AI will become more efficient and the AI providers will likely help with optimization eventually to keep costs low in order not to lose a customer. Like you mentioned, there are specialists that optimize prompts and systems for efficient token usage.

Even if a company burns through their token budget, what was the ROI? Did it help an employee do the work of 2 people, including 1 that didn't need a 401k match and health insurance? AI can run all night to tackle hard problems, so it's like having an extra shift at your office as well.

These are the early days, no doubt this will be optimized and made more efficient, even on one's own air gapped hardware.
 
The anti-data center crowd does not want the DC's, but will be the first to scream if they don't have the digital capability needed in the future. Its a NIMBY thing. Everyone will want the product, but they will protest the factory that build the product.
I'm fine with a DC in my backyard, as long as its quiet, doesn't pollute, increase my electric or water bills, and deplete the water table.

Many cities have massive auto manufacturing plants and I don't recall any issues with those, in fact when there is news of one reopening or being built, everyone is happy and the governor shows up for the openeing ceremony.

I recently talked to a local official about solar farm protestors, who lost their case.

I asked: "what's the deal with peopel being against solar farms?"

Official: "they don't like the glare and farm land being used for other than farming"

Me: "Everyone here lives in a ranch, they don't have to worry about glare with the angle of solar panels" "it's not their farm land, why bother?" "If I were a farmer, and somone wanted to pay me not to farm and rent my field, that's easy money"

Everyone has an opinion, and there are a lot of morons in the general public.
 
In the future I think they may have design meeting before they just keep running trial and error to let AI burn through token. In the hardware world people don't just throw changes out there for tape out that cost $1M without validation first. The same would go for AI if it cost too much.
 
Back
Top Bottom