Send Link mailing list submissions to
[email protected]
To subscribe or unsubscribe via the World Wide Web, visit
https://mailman.anu.edu.au/mailman/listinfo/link
or, via email, send a message with subject or body 'help' to
[email protected]
You can reach the person managing the list at
[email protected]
When replying, please edit your Subject line so it is more specific
than "Re: Contents of Link digest..."
Today's Topics:
1. Re: Does current AI represent a dead end? (Stephen Loosley)
2. Asian clouds target small and medium enterprises in APAC in
search of growth (Stephen Loosley)
3. US Military, "We want to share data, and fight the way we
want in actual warfare scenario.? (Stephen Loosley)
4. Re: Does current AI represent a dead end? (Tom Worthington)
----------------------------------------------------------------------
Message: 1
Date: Fri, 06 Dec 2024 14:29:15 +1030
From: Stephen Loosley <[email protected]>
To: "link" <[email protected]>
Subject: Re: [LINK] Does current AI represent a dead end?
Message-ID: <[email protected]>
Content-Type: text/plain; charset="UTF-8"
Kim writes,
> So, is there hope? I believe, though I would be happy to be proved wrong on
> this
> ? that current generative AI systems represent a dead end, where exponential
> increases of training data and effort will give us modest increases in
> impressive
> plausibility but no foundational increase in reliability. I would love to see
> compositional approaches to neural networks, hard as it appears.
Kim, thanks for this interesting post, was wondering if you might expand on your
above-mentioned "compositional approaches to neural networks," if you've time.
Thanks also to Roger & Kate for equally interesting responses. Both are most
apt!
Cheers
------------------------------
Message: 2
Date: Fri, 06 Dec 2024 14:52:23 +1030
From: Stephen Loosley <[email protected]>
To: "link" <[email protected]>
Subject: [LINK] Asian clouds target small and medium enterprises in
APAC in search of growth
Message-ID: <[email protected]>
Content-Type: text/plain; charset="UTF-8"
Chinese clouds target small and medium enterprises in APAC in search of growth
Smaller buyers see deep discounts and suddenly worry less about regulatory
issues
By Laura Dobberstein Thu 5 Dec 2024
https://www.theregister.com/2024/12/05/chinese_cloud_vendor_apac_prices/
Canalys APAC Forum:
Chinese cloud vendors are undercutting their US counterparts on price and
making inroads into the Asian small and medium business market, according to
analyst firm Canalys.
Speaking at the Canalys APAC Forum event yesterday in Indonesia, analyst Yi
Zhang observed that AWS, Google, and Microsoft currently operate 32 datacenters
across Asia-Pacific, compared to 20 run by Alibaba, Huawei, and Tencent.
But the Chinese clouds' facilities represent almost half of their offshore
presences ? a sign they are targeting the region.
Zhang said China's clouds sometimes appeal because they each have strengths
drawn from their parents main activity: Alibaba Cloud serves etailers well,
Tencent Cloud has products for the entertainment industry, and Huawei Cloud is
strong in core IT.
But the analyst feels the *big advantage* China's clouds offer is price ?
typically 30 to 50 percent less than rival hyperscalers.
This cost stage makes cloud service more affordable and accessible, and to
those price sensitive customers, this can mean the difference between being
able to access cloud or being entirely priced out, explained Zhang.
Small businesses, in particular, are prioritizing affordability over other
considerations, including data compliance ? a long-standing issue for Chinese
cloud vendors. Concerns often arise regarding adherence to local data
sovereignty laws and the potential for government oversight or access.
It's understandable that some partners may be cautious about working with
Chinese vendors due to the customer concerns about data compliance, the analyst
acknowledged.
However, attitudes are shifting. Market acceptance has increased in the past
year, with two thirds of channel partners now seeing greater opportunities with
Chinese vendors compared to a year ago.
Zhang described the expansion of Chinese cloud vendors into APAC as a
*strategic necessity,* driven by changing economic conditions and declining
domestic enterprise demand. The Chinese cloud market has seen growth plunge
from 66 percent in 2020 to just 14 percent in the first half of 2024, forcing
vendors to seek revenue beyond their borders.
The APAC region was a natural choice for expansion because of its geographical
proximity, cultural ties, and a booming digital economy.
According to Canalys, the region accounts for 53 percent of overseas revenue
for China's top three cloud vendors.
--
------------------------------
Message: 3
Date: Sat, 7 Dec 2024 01:28:46 +1100
From: Stephen Loosley <[email protected]>
To: [email protected]
Subject: [LINK] US Military, "We want to share data, and fight the way
we want in actual warfare scenario.?
Message-ID:
<sy5p282mb4409497e19e2827f1d96152ac2...@sy5p282mb4409.ausp282.prod.outlook.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
New joint data standards could come in early 2025
The anticipated document would help ensure compatibility between
training and simulation systems across services.
By Lauren C. Williams Senior Editor December 5, 2024
https://www.defenseone.com/defense-systems/2024/12/new-joint-data-standards-could-come-early-2025/401450/
Re:Training & Simulation, Pentagon I/ITSEC
* Bell said developing cross-domain solutions that work effectively will
require ?continued investment,? but is critical for the U.S. military to
continue to train with allies and partners.
* ?We want to be able to share data with them so that we can train as we
can fight, and we can fight the way we want to in an actual warfare
scenario,? Bell said.
ORLANDO, Fla.?The Pentagon?s top body in charge of developing military
requirements is expected to release new data standards next year that
will make it easier for combatant commands and services to conduct
large, exercises in virtual environments.
The joint requirements oversight council, or JROC, is currently working
on a capstone initial capabilities document, which will outline data
standards for simulation and modeling systems the military uses.
The document is expected in March, said Adm. Christopher Grady, vice
chairman of the Joint Chiefs of Staff, during the Interservice/Industry
Training, Simulation and Education Conference.
Standards for data architecture and data sharing are necessary so the
joint force can take advantage of modeling and simulation technologies.
?There's so much data out there that you all are working with?How do we
manage data so that we can bring it into these very sophisticated
modeling and simulation engines that are going to help us with that
dominant decision-making?? he said.
?But from a joint perspective, we need to have standards right. So that
as we build out, for instance, the Joint Live, Virtual and Constructive
framework, I want the services and the? [combatant commands]? to be able
to plug into that seamlessly.
So standardization will be really, really important.?
Still, there?s work remaining, and the Pentagon plans to ask Congress
for funding to ?build out those robust standards so that everybody can
see where they fit in,? Grady said.
?I think, from a big arc perspective, that will be really helpful,? he said.
Adopting joint standards would make it so various simulation systems
used by each of the services can talk to each other, which would make it
easier to hold large, joint training exercises, said John Bell, chief
technology officer HII Mission Technologies, which supports the Joint
Live, Virtual, Constructive training environment the combatant commands
use.
?If I'm a Navy guy and I'm running a Navy exercise with Navy simulation
systems and we want to do a joint exercise with the Air Force, how do we
make sure all of our training systems can talk to each other in a
meaningful way, so that we train together in a meaningful way?? Bell said.
?So we have to set those standards at a joint level.?
And when it comes to data sharing across classification levels, the
military needs to have? machine-based solutions to keep operators from
having to toggle between computers to share information with mission
partners, Grady said.
?I want everything to be fully informed. And that's very, very challenging.
So the more that we can get away from swivel-chairing data, the more
that we can have cross-domain solutions from the highest classification
level all the way down, the better will be,? he said, describing how
operators will spin in their chair to share information between systems
of different classification levels.
That?s a task better suited for computers, but developing cross-domain
solutions that are effective is very difficult, Bell said.
?We want to have computers be able to do that. And we have these
cross-domain solutions in place, but they're very difficult to develop,?
he said.
Bell said developing cross-domain solutions that work effectively will
require ?continued investment,? but is critical for the U.S. military to
continue to train with allies and partners.
?We want to be able to share data with them so that we can train as we
can fight and we can fight the way we want to in an actual warfare
scenario,? Bell said.
--
------------------------------
Message: 4
Date: Sat, 7 Dec 2024 09:28:47 +1100
From: Tom Worthington <[email protected]>
To: [email protected]
Subject: Re: [LINK] Does current AI represent a dead end?
Message-ID: <[email protected]>
Content-Type: text/plain; charset="utf-8"; Format="flowed"
On 5/12/24 15:33, Kim Holburn wrote:
> https://www.bcs.org/articles-opinion-and-research/does-current-ai-represent-a-dead-end
>
> ... Current AI systems have no internal structure that relates meaningfully
> to their functionality. ...
The clever bit about Generative AI is it produces useful results without
apparently "knowing" anything. There are other ways to do AI.
This form of criticism seems to be saying: "But it shouldn't work!".
This current fad for generative AI will likely pass, but will boost
funding of other approaches. An example is Google's DeepMind which was
awarded a Nobel Prize in chemistry for protein prediction AI. They
originally set out to have the system learn purely on its own. That did
not work, so they added some explicit knowledge of chemistry.
https://en.wikipedia.org/wiki/Google_DeepMind
--
Tom Worthington http://www.tomw.net.au
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 665 bytes
Desc: OpenPGP digital signature
URL:
<https://mailman.anu.edu.au/pipermail/link/attachments/20241207/bc52afb5/attachment.sig>
------------------------------
Subject: Digest Footer
_______________________________________________
Link mailing list
[email protected]
https://mailman.anu.edu.au/mailman/listinfo/link
------------------------------
End of Link Digest, Vol 385, Issue 7
************************************
