I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 US.
1,000,000 x $500,000 = $500 billion US of capital.
Lynn McGuire <lynnmcguire5@gmail.com> writes:US.
I am on the periphery of the ongoing blanketing of the USA with AI=20 >>servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a=20 >>million (SWAG). The current cost for a single AI server is $500,000 =
1,000,000 x $500,000 =3D $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
On Thu, 14 Nov 2024 15:00:01 GMT, scott@slp53.sl.home (Scott Lurndal)
wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:US.
I am on the periphery of the ongoing blanketing of the USA with AI=20 >>>servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a=20 >>>million (SWAG). The current cost for a single AI server is $500,000 =
1,000,000 x $500,000 =3D $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
I suggest two reasons why defining "AI server" precisely is not
feasible:
1. Nobody knows -- an AI Server is whatever the makers make it be.
2. The tech will change, so, even if a "precise definition" existed
today, next week it might well be outmoded.
Bit-coin mining is meeting with opposition from local communities with >stressed electric systems. AI Servers will as well, I expect.
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 US.
1,000,000 x $500,000 = $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
On Thu, 14 Nov 2024, Scott Lurndal wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 US. >>> 1,000,000 x $500,000 = $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
My guess would be a server stuffed with GPU:s.
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 US.
1,000,000 x $500,000 = $500 billion US of capital.
Each AI server idles at 5,000 watts and uses 15,000 watts at full
compute level.
15,000 watts x 1,000,000 = 15,000,000,000 watts = 15,000 MW = 15 GW Current peak usage in Texas is 85,000 MW = 85 GW.
Where is all of this electric power going to come from, especially since
we are shutting down 100+ coal power plants per year in the USA ?
On 14/11/24 12:43, Lynn McGuire wrote:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 US.
1,000,000 x $500,000 = $500 billion US of capital.
Each AI server idles at 5,000 watts and uses 15,000 watts at full
compute level.
15,000 watts x 1,000,000 = 15,000,000,000 watts = 15,000 MW = 15 GW
Current peak usage in Texas is 85,000 MW = 85 GW.
Where is all of this electric power going to come from, especially since
we are shutting down 100+ coal power plants per year in the USA ?
"So how is this pertinent to Science Fiction and Fantasy, aka
Speculative Fiction ?" Lynn
(AGW. LNG Worse Than Coal. October 2024)
I don't have a problem with your off topic posting because it usually
results in interesting discussion but wished to remind you of your
hypocrisy, (again).
On 11/14/2024 11:17 PM, Titus G wrote:
On 14/11/24 12:43, Lynn McGuire wrote:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 US. >>> 1,000,000 x $500,000 = $500 billion US of capital.
Each AI server idles at 5,000 watts and uses 15,000 watts at full
compute level.
15,000 watts x 1,000,000 = 15,000,000,000 watts = 15,000 MW = 15 GW
Current peak usage in Texas is 85,000 MW = 85 GW.
Where is all of this electric power going to come from, especially since >>> we are shutting down 100+ coal power plants per year in the USA ?
"So how is this pertinent to Science Fiction and Fantasy, aka
Speculative Fiction ?" Lynn
(AGW. LNG Worse Than Coal. October 2024)
I don't have a problem with your off topic posting because it usually
results in interesting discussion but wished to remind you of your
hypocrisy, (again).
I figured that you were smart enough to figure that this posting had to
do with excess heat and climate change. Maybe all of the science
fiction stories about the Earth burning up due to excess heat might come true.
Lynn
On 11/14/2024 2:18 PM, D wrote:
On Thu, 14 Nov 2024, Scott Lurndal wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 US. >>>> 1,000,000 x $500,000 = $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
My guess would be a server stuffed with GPU:s.
1. A huge rackmount frame the size of a refrigerator
2. Two Intel Zeons with 64 or 128 cpus each
3. Many terabytes of ram in 20 to 40 memory slots
4. 10 to 40 SSD hard drives of 1 TB to 8 TB
5. four to twelve Nvidea H100 or H200 boards
6. multiple power supplies
7. etc, etc, etc, etc, etc, etc
In other words, it looks like a 1980s / 1990s mini computer in a rack configuration with many swapable / upgradable modules.
Here is just one variant: https://www.dell.com/en-us/dt/servers/specialty-servers/poweredge-xe-servers.htm#scroll=off&tab0=0&accordion0
Lynn
Paul S Person <psperson@old.netcom.invalid> writes:AI=3D20
On Thu, 14 Nov 2024 15:00:01 GMT, scott@slp53.sl.home (Scott Lurndal) >>wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with =
a=3D20servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently =
=3Dmillion (SWAG). The current cost for a single AI server is $500,000 =
fluff)US.
1,000,000 x $500,000 =3D3D $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
I suggest two reasons why defining "AI server" precisely is not
feasible:
1. Nobody knows -- an AI Server is whatever the makers make it be.
2. The tech will change, so, even if a "precise definition" existed
today, next week it might well be outmoded.
Machine Learning (which is the proper term, "AI" is just marketing =
has two components:
- Feeding the model data (Training)
- Applying the model (Inference)
The former is computationally intensive (and thus requires massive
amounts of energy). The later uses the data produced from the training >activity and is much less energy intensive.
=20
Bit-coin mining is meeting with opposition from local communities with >>stressed electric systems. AI Servers will as well, I expect.
Bitcoin is a complete waste of energy. =20
Where is all of this electric power going to come from, especially since
we are shutting down 100+ coal power plants per year in the USA ?
On Thu, 14 Nov 2024, Scott Lurndal wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 US. >>> 1,000,000 x $500,000 = $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
My guess would be a server stuffed with GPU:s.
Lynn McGuire wrote:
Where is all of this electric power going to come from, especially
since we are shutting down 100+ coal power plants per year in the USA ?
As I understand it about 170 are due to be shut down between now and
2030. And given that delays in this field are not uncommon, I suspect
the actual number will be less.
Since we shut down our coal plants fifteen years ago, the number of respiratory alerts in Toronto has gone from thirty per summer to one or less.
That alone made it worth the cost.
William Hyde
On Thu, 14 Nov 2024, Lynn McGuire wrote:
On 11/14/2024 2:18 PM, D wrote:
On Thu, 14 Nov 2024, Scott Lurndal wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is
$500,000 US.
1,000,000 x $500,000 = $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
My guess would be a server stuffed with GPU:s.
1. A huge rackmount frame the size of a refrigerator
2. Two Intel Zeons with 64 or 128 cpus each
3. Many terabytes of ram in 20 to 40 memory slots
4. 10 to 40 SSD hard drives of 1 TB to 8 TB
5. four to twelve Nvidea H100 or H200 boards
6. multiple power supplies
7. etc, etc, etc, etc, etc, etc
In other words, it looks like a 1980s / 1990s mini computer in a rack
configuration with many swapable / upgradable modules.
Here is just one variant:
https://www.dell.com/en-us/dt/servers/specialty-servers/poweredge-xe-
servers.htm#scroll=off&tab0=0&accordion0
Lynn
Just what I would have expected! It will be fun if there's another AI- winter. Then we'll be eating GPU:s for breakfast given all the
oversupply that will exist! When do you think disillusionment will set
in (if at all)?
On 11/15/2024 3:23 AM, D wrote:
On Thu, 14 Nov 2024, Lynn McGuire wrote:
On 11/14/2024 2:18 PM, D wrote:
On Thu, 14 Nov 2024, Scott Lurndal wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI >>>>>> servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 >>>>>> US.
1,000,000 x $500,000 = $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
My guess would be a server stuffed with GPU:s.
1. A huge rackmount frame the size of a refrigerator
2. Two Intel Zeons with 64 or 128 cpus each
3. Many terabytes of ram in 20 to 40 memory slots
4. 10 to 40 SSD hard drives of 1 TB to 8 TB
5. four to twelve Nvidea H100 or H200 boards
6. multiple power supplies
7. etc, etc, etc, etc, etc, etc
In other words, it looks like a 1980s / 1990s mini computer in a rack
configuration with many swapable / upgradable modules.
Here is just one variant:
https://www.dell.com/en-us/dt/servers/specialty-servers/poweredge-xe-
servers.htm#scroll=off&tab0=0&accordion0
Lynn
Just what I would have expected! It will be fun if there's another AI-
winter. Then we'll be eating GPU:s for breakfast given all the oversupply >> that will exist! When do you think disillusionment will set in (if at all)?
The AI disillusionment will start when the companies try to layoff their customer support departments and find out that they cannot do so.
Lynn
On 11/15/2024 1:48 PM, William Hyde wrote:
Lynn McGuire wrote:
Where is all of this electric power going to come from, especially
since we are shutting down 100+ coal power plants per year in the USA ?
As I understand it about 170 are due to be shut down between now and
2030. And given that delays in this field are not uncommon, I suspect
the actual number will be less.
Since we shut down our coal plants fifteen years ago, the number of
respiratory alerts in Toronto has gone from thirty per summer to one
or less.
That alone made it worth the cost.
William Hyde
Yeah, I should have put SWAG on that 100 per year number. The USA has
204 coal power plants left. About a dozen or so here in Texas. Nine of
the coal power plants I worked at in the 1980s have been shut down in
the last 15 years.
https://www.statista.com/statistics/859266/number-of-coal-power-plants-by-country/
Good luck on getting China to shut down their 1,161 coal power plants.
They are adding a new coal power plant every week still in China (SWAG).
Bag houses on the exhaust gas would have stopped the flyash in the air problem. Your old coal power plants
Lynn McGuire wrote:
On 11/15/2024 1:48 PM, William Hyde wrote:
Lynn McGuire wrote:
As I understand it about 170 are due to be shut down between now and
Where is all of this electric power going to come from, especially
since we are shutting down 100+ coal power plants per year in the USA ? >>>
2030. And given that delays in this field are not uncommon, I
suspect the actual number will be less.
Since we shut down our coal plants fifteen years ago, the number of
respiratory alerts in Toronto has gone from thirty per summer to one
or less.
That alone made it worth the cost.
William Hyde
Yeah, I should have put SWAG on that 100 per year number. The USA has
204 coal power plants left. About a dozen or so here in Texas. Nine
of the coal power plants I worked at in the 1980s have been shut down
in the last 15 years.
https://www.statista.com/statistics/859266/number-of-coal-power-
plants-by-country/
Good luck on getting China to shut down their 1,161 coal power plants.
They are adding a new coal power plant every week still in China (SWAG).
Bag houses on the exhaust gas would have stopped the flyash in the air
problem. Your old coal power plants
When I was in public school our class was taken on a trip to see this
same spanking new technological marvel of a power plant.
Scary.
William Hyde
On Fri, 15 Nov 2024, Lynn McGuire wrote:AI
On 11/15/2024 3:23 AM, D wrote:
=20
=20
On Thu, 14 Nov 2024, Lynn McGuire wrote:
=20
On 11/14/2024 2:18 PM, D wrote:
=20
=20
On Thu, 14 Nov 2024, Scott Lurndal wrote:
=20
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with =
rack=20=20=20servers.=A0 I have a few facts that might just blow you away. >>>>>>>=20=20
The expected number of AI servers in the USA alone is presently a >>>>>>> million (SWAG).=A0 The current cost for a single AI server is = $500,000=20
US.
=A0=A0 1,000,000 x $500,000 =3D $500 billion US of capital.
First, What is your source for this data?=A0 Be specific.
=20
Second, define precisely what an "AI server" is.
My guess would be a server stuffed with GPU:s.
1. A huge rackmount frame the size of a refrigerator
2. Two Intel Zeons with 64 or 128 cpus each
3. Many terabytes of ram in 20 to 40 memory slots
4. 10 to 40 SSD hard drives of 1 TB to 8 TB
5. four to twelve Nvidea H100 or H200 boards
6. multiple power supplies
7. etc, etc, etc, etc, etc, etc
=20
In other words, it looks like a 1980s / 1990s mini computer in a =
https://www.dell.com/en-us/dt/servers/specialty-servers/poweredge-xe-=20configuration with many swapable / upgradable modules.
=20
Here is just one variant:
=
AI-=20servers.htm#scroll=3Doff&tab0=3D0&accordion0=20
=20
Lynn
Just what I would have expected! It will be fun if there's another =
all)?winter. Then we'll be eating GPU:s for breakfast given all the = oversupply=20
that will exist! When do you think disillusionment will set in (if at=
=20
The AI disillusionment will start when the companies try to layoff = their=20
customer support departments and find out that they cannot do so.
Lynn
Let us study the example of Klarna with great interest! I think they did=
exactly that. Instead of having people answering questions about loans = and=20
credit, they will let the LLM:s do it and I then assume have a vastly=20 >smaller teams of humans catching the hallucinations and edge cases.
On 11/15/2024 1:48 PM, William Hyde wrote:?
Lynn McGuire wrote:
=20
Where is all of this electric power going to come from, especially=20
since we are shutting down 100+ coal power plants per year in the USA=
or=20=20
As I understand it about 170 are due to be shut down between now and=20
2030.=A0 And given that delays in this field are not uncommon, I = suspect=20
the actual number will be less.
=20
Since we shut down our coal plants fifteen years ago, the number of=20
respiratory alerts in Toronto has gone from thirty per summer to one =
=20less.
=20
That alone made it worth the cost.
=20
William Hyde
Yeah, I should have put SWAG on that 100 per year number. The USA has=20
204 coal power plants left. About a dozen or so here in Texas. Nine of=
the coal power plants I worked at in the 1980s have been shut down in=20
the last 15 years.
=20
https://www.statista.com/statistics/859266/number-of-coal-power-plants-b= y-country/
Good luck on getting China to shut down their 1,161 coal power plants.=20 >They are adding a new coal power plant every week still in China (SWAG).
Bag houses on the exhaust gas would have stopped the flyash in the air=20 >problem. Your old coal power plants probably just had electrostatic=20 >precipitators, if that, which only get 80% of the flyash on a good day.=20 >Bag houses are 98% effective but require more exhaust gas fans due to=20 >their pressure drop (1/4 inch wall thickness woven metal bags).--=20
Most of the companies I have had to try to deal with over the past fewyears are heavily protected by automated telephone/online chat trees
which /never/ get you to an actual person. The closest one came was to
offer me the option of using my phone to chat with one. Sadly, my
phone doesn't do that; it's very primitive.
IOW, they only want to hear about the problems they can imagine you
might have and so have solutions for. Anything else is of no interest
to them.
On 17/11/2024 04:02, Paul S Person wrote:few
[SNIP]
Most of the companies I have had to try to deal with over the past =
years are heavily protected by automated telephone/online chat trees
which /never/ get you to an actual person. The closest one came was to
offer me the option of using my phone to chat with one. Sadly, my
phone doesn't do that; it's very primitive.
=20
IOW, they only want to hear about the problems they can imagine you
might have and so have solutions for. Anything else is of no interest
to them.
Many - I can't say most - of the voice-(non-)response systems can be=20 >forced to pass you to a human if you keep saying, "I want to speak to a=20 >human", or, "I need to speak to a person", repeatedly.
That's my general goto when I get stuck in a vast maze of twisty=20
passages, all alike.
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 US.
1,000,000 x $500,000 = $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
On Sun, 17 Nov 2024 14:09:54 +1100, "Gary R. Schmidt"
<grschmidt@acm.org> wrote:
On 17/11/2024 04:02, Paul S Person wrote:
[SNIP]
Most of the companies I have had to try to deal with over the past fewyears are heavily protected by automated telephone/online chat trees
which /never/ get you to an actual person. The closest one came was to
offer me the option of using my phone to chat with one. Sadly, my
phone doesn't do that; it's very primitive.
IOW, they only want to hear about the problems they can imagine you
might have and so have solutions for. Anything else is of no interest
to them.
Many - I can't say most - of the voice-(non-)response systems can be
forced to pass you to a human if you keep saying, "I want to speak to a
human", or, "I need to speak to a person", repeatedly.
That's my general goto when I get stuck in a vast maze of twisty
passages, all alike.
That's been my experience in the past.
But the last time I tried it, all I got was a number I was graciously
allowed to send text messages to. Of course, if you happen to have a
phone that can send text messages, that's just as good in that you may
find yourself working with a human being. Unless it just leads you to
the chat tree used online, of course.
OTOH, when my DSL died my former ISP had people available. I may have
managed to mess things up a bit, but they helped me un-mess them so
things worked out fine (kept my emails, dropped them as an ISP). But
they may be smaller than the company I am describing.
But the last time I tried it, all I got was a number I was graciously
allowed to send text messages to. Of course, if you happen to have a
phone that can send text messages, that's just as good in that you may
find yourself working with a human being. Unless it just leads you to
the chat tree used online, of course.
On 11/14/2024 9:00 AM, Scott Lurndal wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000 US. >>> 1,000,000 x $500,000 = $500 billion US of capital.
First, What is your source for this data? Be specific.
Second, define precisely what an "AI server" is.
BTW, I should have mentioned that the AI Servers are not uniform in any
of their aspects. I just gave the specs for one of the high end
machines that the manufacturer has a three month waiting list for.
On 11/17/2024 10:31 AM, Paul S Person wrote:=20
But the last time I tried it, all I got was a number I was graciously
allowed to send text messages to. Of course, if you happen to have a
phone that can send text messages, that's just as good in that you may
find yourself working with a human being. Unless it just leads you to
the chat tree used online, of course.
You can use email to text but you'd have to send to each of the carriers=
interface addresses.
Lynn McGuire <lynnmcguire5@gmail.com> writes:US.
On 11/14/2024 9:00 AM, Scott Lurndal wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000=
=201,000,000 x $500,000 =3D $500 billion US of capital.=20
First, What is your source for this data? Be specific.
=20
Second, define precisely what an "AI server" is.
BTW, I should have mentioned that the AI Servers are not uniform in any=
parallelof their aspects. I just gave the specs for one of the high end=20 >>machines that the manufacturer has a three month waiting list for.
Some of us actually produce ML hardware. The term AI is an marketing
gimic, not reality.
ML hardware ranges from custom logic in a desktop CPU to massively =
specialized hardware (e.g. plug-in GPUs). An ML-enabled server can
range from a simple ML accelerator block in the CPU itself (Apple,
Google, Amazon, Marvell) to a large, expensive, power-hungry GPU from = Nvidia.
Yes, there are large racks of ML-enabled servers, particularly for
training. No, they're not really anything special other than using
high-end, high performance CPUs, GPUs and high-speed interconnects >(Infiniband, 100 and 400Gb ethernet). Basically no different
than any supercomputer in scale and power requirements.
On Mon, 18 Nov 2024 14:12:34 GMT, scott@slp53.sl.home (Scott Lurndal)
wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:US.
On 11/14/2024 9:00 AM, Scott Lurndal wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI
servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000=
1,000,000 x $500,000 =3D $500 billion US of capital.=20
First, What is your source for this data? Be specific.
=20
Second, define precisely what an "AI server" is.
BTW, I should have mentioned that the AI Servers are not uniform in any= >=20
of their aspects. I just gave the specs for one of the high end=20 >>>machines that the manufacturer has a three month waiting list for.
Some of us actually produce ML hardware. The term AI is an marketing >>gimic, not reality.
ML hardware ranges from custom logic in a desktop CPU to massively = >parallel
specialized hardware (e.g. plug-in GPUs). An ML-enabled server can
range from a simple ML accelerator block in the CPU itself (Apple,
Google, Amazon, Marvell) to a large, expensive, power-hungry GPU from = >Nvidia.
Yes, there are large racks of ML-enabled servers, particularly for >>training. No, they're not really anything special other than using >>high-end, high performance CPUs, GPUs and high-speed interconnects >>(Infiniband, 100 and 400Gb ethernet). Basically no different
than any supercomputer in scale and power requirements.
1 million of them will rather increase the power requirements,
possibly causing shortages.
On Sun, 17 Nov 2024 21:16:12 -0600, "Jay E. Morris"I was talking about an email to text message interface. For example if
<morrisj@epsilon3.comcon> wrote:
On 11/17/2024 10:31 AM, Paul S Person wrote:
But the last time I tried it, all I got was a number I was graciously
allowed to send text messages to. Of course, if you happen to have a
phone that can send text messages, that's just as good in that you may
find yourself working with a human being. Unless it just leads you to
the chat tree used online, of course.
You can use email to text but you'd have to send to each of the carriers
interface addresses.
Email disappeared as an option decades ago. Which is too bad, it
worked quite well once I figured out that:
Paul S Person <psperson@old.netcom.invalid> writes:
On Mon, 18 Nov 2024 14:12:34 GMT, scott@slp53.sl.home (Scott Lurndal)
wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:US.
On 11/14/2024 9:00 AM, Scott Lurndal wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI >>>>>> servers. I have a few facts that might just blow you away.
The expected number of AI servers in the USA alone is presently a
million (SWAG). The current cost for a single AI server is $500,000=
parallel1,000,000 x $500,000 =3D $500 billion US of capital.=20
First, What is your source for this data? Be specific.
=20
Second, define precisely what an "AI server" is.
BTW, I should have mentioned that the AI Servers are not uniform in any= >> =20
of their aspects. I just gave the specs for one of the high end=20
machines that the manufacturer has a three month waiting list for.
Some of us actually produce ML hardware. The term AI is an marketing
gimic, not reality.
ML hardware ranges from custom logic in a desktop CPU to massively =
specialized hardware (e.g. plug-in GPUs). An ML-enabled server canNvidia.
range from a simple ML accelerator block in the CPU itself (Apple,
Google, Amazon, Marvell) to a large, expensive, power-hungry GPU from =
Yes, there are large racks of ML-enabled servers, particularly for
training. No, they're not really anything special other than using
high-end, high performance CPUs, GPUs and high-speed interconnects
(Infiniband, 100 and 400Gb ethernet). Basically no different
than any supercomputer in scale and power requirements.
1 million of them will rather increase the power requirements,
possibly causing shortages.
It is unlikely that there will be anywhere near one million
ML training setups, which are the power-hungry setups.
ML Inference requires significantly less horsepower and will
be distributed over millions of laptop/desktop/table systems.
It's just as likely that the hype bubble will burst in the
next 18 months.
On 11/18/2024 10:31 AM, Paul S Person wrote:
On Sun, 17 Nov 2024 21:16:12 -0600, "Jay E. Morris"I was talking about an email to text message interface. For example if
<morrisj@epsilon3.comcon> wrote:
On 11/17/2024 10:31 AM, Paul S Person wrote:
But the last time I tried it, all I got was a number I was graciously
allowed to send text messages to. Of course, if you happen to have a
phone that can send text messages, that's just as good in that you may >>>> find yourself working with a human being. Unless it just leads you to
the chat tree used online, of course.
You can use email to text but you'd have to send to each of the carriers >>> interface addresses.
Email disappeared as an option decades ago. Which is too bad, it
worked quite well once I figured out that:
you send email to 5551234567@txt.att.net (substituting my number of
course) it will arrive on my phone as a text. My reply will go back to
your email.
"Jay E. Morris" <morrisj@epsilon3.comcon> writes:
On 11/18/2024 10:31 AM, Paul S Person wrote:
On Sun, 17 Nov 2024 21:16:12 -0600, "Jay E. Morris"I was talking about an email to text message interface. For example if
<morrisj@epsilon3.comcon> wrote:
On 11/17/2024 10:31 AM, Paul S Person wrote:
But the last time I tried it, all I got was a number I was graciously >>>>> allowed to send text messages to. Of course, if you happen to have a >>>>> phone that can send text messages, that's just as good in that you may >>>>> find yourself working with a human being. Unless it just leads you to >>>>> the chat tree used online, of course.
You can use email to text but you'd have to send to each of the carriers >>>> interface addresses.
Email disappeared as an option decades ago. Which is too bad, it
worked quite well once I figured out that:
you send email to 5551234567@txt.att.net (substituting my number of
course) it will arrive on my phone as a text. My reply will go back to
your email.
Have you tried it recently? Didn't work for me.
On 11/18/2024 10:45 AM, Scott Lurndal wrote:
Paul S Person <psperson@old.netcom.invalid> writes:
On Mon, 18 Nov 2024 14:12:34 GMT, scott@slp53.sl.home (Scott Lurndal)
wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:parallel
On 11/14/2024 9:00 AM, Scott Lurndal wrote:
Lynn McGuire <lynnmcguire5@gmail.com> writes:
I am on the periphery of the ongoing blanketing of the USA with AI >>>>>>> servers. I have a few facts that might just blow you away.=20
The expected number of AI servers in the USA alone is presently a >>>>>>> million (SWAG). The current cost for a single AI server is $500,000= >>> US.
1,000,000 x $500,000 =3D $500 billion US of capital.
First, What is your source for this data? Be specific.
=20
Second, define precisely what an "AI server" is.
BTW, I should have mentioned that the AI Servers are not uniform in any= >>> =20
of their aspects. I just gave the specs for one of the high end=20
machines that the manufacturer has a three month waiting list for.
Some of us actually produce ML hardware. The term AI is an marketing
gimic, not reality.
ML hardware ranges from custom logic in a desktop CPU to massively =
specialized hardware (e.g. plug-in GPUs). An ML-enabled server can
range from a simple ML accelerator block in the CPU itself (Apple,
Google, Amazon, Marvell) to a large, expensive, power-hungry GPU from = >>> Nvidia.
Yes, there are large racks of ML-enabled servers, particularly for
training. No, they're not really anything special other than using
high-end, high performance CPUs, GPUs and high-speed interconnects
(Infiniband, 100 and 400Gb ethernet). Basically no different
than any supercomputer in scale and power requirements.
1 million of them will rather increase the power requirements,
possibly causing shortages.
It is unlikely that there will be anywhere near one million
ML training setups, which are the power-hungry setups.
ML Inference requires significantly less horsepower and will
be distributed over millions of laptop/desktop/table systems.
It's just as likely that the hype bubble will burst in the
next 18 months.
""Prediction is very difficult, especially about the future." — Niels Bohr" https://www.reddit.com/r/quotes/comments/10at8mq/prediction_is_very_difficult_especially_about_the/
BTW, my buddy who programs AI Servers at the big dumb company, agrees with you. But they are selling hundreds of the huge AI servers each month and are
struggling to meet demand.
Lynn
On 11/18/2024 2:26 PM, Scott Lurndal wrote:graciously
"Jay E. Morris" <morrisj@epsilon3.comcon> writes:
On 11/18/2024 10:31 AM, Paul S Person wrote:
On Sun, 17 Nov 2024 21:16:12 -0600, "Jay E. Morris"
<morrisj@epsilon3.comcon> wrote:
On 11/17/2024 10:31 AM, Paul S Person wrote:
But the last time I tried it, all I got was a number I was =
aallowed to send text messages to. Of course, if you happen to have=
mayphone that can send text messages, that's just as good in that you=
tofind yourself working with a human being. Unless it just leads you=
ifI was talking about an email to text message interface. For example =the chat tree used online, of course.
You can use email to text but you'd have to send to each of the = carriers
interface addresses.
Email disappeared as an option decades ago. Which is too bad, it
worked quite well once I figured out that:
toyou send email to 5551234567@txt.att.net (substituting my number of
course) it will arrive on my phone as a text. My reply will go back =
worked.your email.=20
Have you tried it recently? Didn't work for me.
=20
Yes, just before I posted. Both to the text and reply back to email =
On Mon, 18 Nov 2024 14:54:28 -0600, "Jay E. Morris"
<morrisj@epsilon3.comcon> wrote:
On 11/18/2024 2:26 PM, Scott Lurndal wrote:
"Jay E. Morris" <morrisj@epsilon3.comcon> writes:
On 11/18/2024 10:31 AM, Paul S Person wrote:
On Sun, 17 Nov 2024 21:16:12 -0600, "Jay E. Morris"I was talking about an email to text message interface. For example if >>>> you send email to 5551234567@txt.att.net (substituting my number of
<morrisj@epsilon3.comcon> wrote:
On 11/17/2024 10:31 AM, Paul S Person wrote:
But the last time I tried it, all I got was a number I was graciously >>>>>>> allowed to send text messages to. Of course, if you happen to have a >>>>>>> phone that can send text messages, that's just as good in that you may >>>>>>> find yourself working with a human being. Unless it just leads you to >>>>>>> the chat tree used online, of course.
You can use email to text but you'd have to send to each of the carriers >>>>>> interface addresses.
Email disappeared as an option decades ago. Which is too bad, it
worked quite well once I figured out that:
course) it will arrive on my phone as a text. My reply will go back to >>>> your email.
Have you tried it recently? Didn't work for me.
Yes, just before I posted. Both to the text and reply back to email worked.
I don't suppose there is an easy way to relate a given cell phone
number to the carrier and so to the @ address?
On Mon, 18 Nov 2024 14:54:28 -0600, "Jay E. Morris"
<morrisj@epsilon3.comcon> wrote:
On 11/18/2024 2:26 PM, Scott Lurndal wrote:
"Jay E. Morris" <morrisj@epsilon3.comcon> writes:
On 11/18/2024 10:31 AM, Paul S Person wrote:
On Sun, 17 Nov 2024 21:16:12 -0600, "Jay E. Morris"I was talking about an email to text message interface. For example if >>>> you send email to 5551234567@txt.att.net (substituting my number of
<morrisj@epsilon3.comcon> wrote:
On 11/17/2024 10:31 AM, Paul S Person wrote:
But the last time I tried it, all I got was a number I was graciously >>>>>>> allowed to send text messages to. Of course, if you happen to have a >>>>>>> phone that can send text messages, that's just as good in that you may >>>>>>> find yourself working with a human being. Unless it just leads you to >>>>>>> the chat tree used online, of course.
You can use email to text but you'd have to send to each of the carriers >>>>>> interface addresses.
Email disappeared as an option decades ago. Which is too bad, it
worked quite well once I figured out that:
course) it will arrive on my phone as a text. My reply will go back to >>>> your email.
Have you tried it recently? Didn't work for me.
Yes, just before I posted. Both to the text and reply back to email worked.
I don't suppose there is an easy way to relate a given cell phone
number to the carrier and so to the @ address?
On 11/19/2024 11:25 AM, Paul S Person wrote:
I don't suppose there is an easy way to relate a given cell phone
number to the carrier and so to the @ address?
No, there isn't. US phone numbers have been movable between carriers
for decades. I did it myself.
In article <vhih8r$1uic7$1@dont-email.me>,
Cryptoengineer <petertrei@gmail.com> wrote:
On 11/19/2024 11:25 AM, Paul S Person wrote:
I don't suppose there is an easy way to relate a given cell phone
number to the carrier and so to the @ address?
No, there isn't. US phone numbers have been movable between carriers
for decades. I did it myself.
There is a national database -- obviously, because the incoming calls
have to get routed to the right carrier. I seem to have misplaced the
name of this database; it's operated under an FCC contract by an
organization called the Number Portability Administration Center.
Every phone call and SMS message requires a lookup in this database.
Originally, calls would be routed on the basis of the local exchange
carrier to which the NPA-NXX was assigned, and then numbers which had
been "ported out" would be kicked back to the originating carrier with
a "not our number" message to force a database lookup. These days,
it makes sense to just do the database dip unconditionally. (There's
caching in the carrier networks to minimize their database access
charges.)
Whether there's any way for normal people to get access to this
information I don't know.
-GAWollman
On 11/19/2024 12:44 PM, Garrett Wollman wrote:
In article <vhih8r$1uic7$1@dont-email.me>,
Cryptoengineer <petertrei@gmail.com> wrote:
On 11/19/2024 11:25 AM, Paul S Person wrote:
I don't suppose there is an easy way to relate a given cell phone
number to the carrier and so to the @ address?
No, there isn't. US phone numbers have been movable between carriers
for decades. I did it myself.
There is a national database -- obviously, because the incoming calls
have to get routed to the right carrier. I seem to have misplaced the
name of this database; it's operated under an FCC contract by an
organization called the Number Portability Administration Center.
Every phone call and SMS message requires a lookup in this database.
Originally, calls would be routed on the basis of the local exchange
carrier to which the NPA-NXX was assigned, and then numbers which had
been "ported out" would be kicked back to the originating carrier with
a "not our number" message to force a database lookup. These days,
it makes sense to just do the database dip unconditionally. (There's
caching in the carrier networks to minimize their database access
charges.)
Whether there's any way for normal people to get access to this
information I don't know.
-GAWollman
As I said in the other message, https://www.whitepages.com/ and the
reverse number lookup did give my carrier. I have not moved carriers in
over 20 years though so someone who has would have to verify that it
still gives the right carrier.
On 11/19/2024 12:44 PM, Garrett Wollman wrote:
In article <vhih8r$1uic7$1@dont-email.me>,
Cryptoengineer <petertrei@gmail.com> wrote:
On 11/19/2024 11:25 AM, Paul S Person wrote:=20
I don't suppose there is an easy way to relate a given cell phone
number to the carrier and so to the @ address?
No, there isn't. US phone numbers have been movable between carriers
for decades. I did it myself.
There is a national database -- obviously, because the incoming calls
have to get routed to the right carrier. I seem to have misplaced the
name of this database; it's operated under an FCC contract by an
organization called the Number Portability Administration Center.
Every phone call and SMS message requires a lookup in this database.
=20
Originally, calls would be routed on the basis of the local exchange
carrier to which the NPA-NXX was assigned, and then numbers which had
been "ported out" would be kicked back to the originating carrier with
a "not our number" message to force a database lookup. These days,
it makes sense to just do the database dip unconditionally. (There's
caching in the carrier networks to minimize their database access
charges.)
=20
Whether there's any way for normal people to get access to this
information I don't know.
=20
-GAWollman
=20
As I said in the other message, https://www.whitepages.com/ and the=20 >reverse number lookup did give my carrier. I have not moved carriers in=20 >over 20 years though so someone who has would have to verify that it=20
still gives the right carrier.
In article <vhji7h$24o52$1@dont-email.me>,
Cryptoengineer <petertrei@gmail.com> wrote:
On 11/19/2024 6:01 PM, Jay E. Morris wrote:
On 11/19/2024 12:44 PM, Garrett Wollman wrote:
In article <vhih8r$1uic7$1@dont-email.me>,
Cryptoengineer <petertrei@gmail.com> wrote:
On 11/19/2024 11:25 AM, Paul S Person wrote:
I don't suppose there is an easy way to relate a given cell phone
number to the carrier and so to the @ address?
No, there isn't. US phone numbers have been movable between carriers >>>>> for decades. I did it myself.
There is a national database -- obviously, because the incoming calls
have to get routed to the right carrier. I seem to have misplaced the >>>> name of this database; it's operated under an FCC contract by an
organization called the Number Portability Administration Center.
Every phone call and SMS message requires a lookup in this database.
Originally, calls would be routed on the basis of the local exchange
carrier to which the NPA-NXX was assigned, and then numbers which had
been "ported out" would be kicked back to the originating carrier with >>>> a "not our number" message to force a database lookup. These days,
it makes sense to just do the database dip unconditionally. (There's
caching in the carrier networks to minimize their database access
charges.)
Whether there's any way for normal people to get access to this
information I don't know.
-GAWollman
As I said in the other message, https://www.whitepages.com/ and the
reverse number lookup did give my carrier. I have not moved carriers in
over 20 years though so someone who has would have to verify that it
still gives the right carrier.
Interesting, thanks!
pt
https://xkcd.com/1129/
| Sysop: | Tetrazocine |
|---|---|
| Location: | Melbourne, VIC, Australia |
| Users: | 4 |
| Nodes: | 8 (0 / 8) |
| Uptime: | 191:55:39 |
| Calls: | 62 |
| Files: | 21,500 |
| Messages: | 70,991 |