Artificial Intelligence thread

Michael90

Senior Member
Registered Member
Please, Log in or Register to view URLs content!
Please Chinese AI developers should not initiate another price war here like in solar industry, leading to situation where everyone loses money for market share. lol Xiaomi advice/caution about this is correct, but they seem to have done the same thing they are warning against . Lmao

However, competition shouldn’t always be on price , they should look at other ways to compete and gain market share without sacrificing all their profits . lol
 

tphuang

General
Staff member
Super Moderator
VIP Professional
Registered Member

Kalum Pupeter

Junior Member
Registered Member

Huawei AI aids conservation of endangered Chinese monkey111111111111111111111111.jpg

Huawei Technologies is using artificial intelligence to protect endangered monkeys here, helping to sift through huge amounts of camera footage to better understand their ecology. The project centers on the critically endangered white-headed langur, of which only around 1,400 are thought to exist, all in the Guangxi region. These cliff-dwelling monkeys are covered in golden fur until they are about two-and-a-half years old, at which point the fur on their heads turns white while most of the rest becomes black.

More than 20 cameras have been set up across a roughly 200-square-kilometer stretch of a nature reserve in Chongzuo. The AI, which was developed in collaboration with the reserve and a nonprofit with animal expertise, detects and saves only footage with langurs present, which is used to analyze their ecology.

"Storing and analyzing massive amounts of data without using AI would be an enormous amount of work," said Tian Yongsheng, deputy general manager of Huawei Guangxi. The system allows this to be done more efficiently.

Data centers and servers for the AI were made by Huawei. The reserve began using the AI in 2024. Going forward, it plans to work on identifying and monitoring individual langurs, which is expected to shed light on points that are not currently understood, such as the meaning of their calls.
The white-headed langur population declined dramatically due to hunting as well as habitat loss due to agricultural expansion, dropping to just 360 or so in the early 1980s. Conservation efforts have enabled a gradual recovery.

"In addition to legal measures and conservation awareness, using new technologies like AI has improved efficiency," the head of the reserve said.
Huawei says it is involved in animal and other conservation projects in 65 locations within and outside China. Other projects involving AI include analyzing beluga whale vocalizations in China's Hainan province and identifying jaguars in Mexico based on their markings.
Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
 

Eventine

Senior Member
Registered Member
Great on zAI to carry forth the torch. However, a threat lies on the horizon. Anthropic has recently put out a few charts for their newest model, Mythos:

1775597104031.png
The numbers do seem to indicate a significant, perhaps generational, leap in performance. And the kicker is that Anthropic is not going to release it to the public, because of "security risks" (they claim Mythos found a way to jail break itself out of their cyber security perimeters; but we may also assume it is to prevent Chinese distillation since they are making it available to "trusted" enterprises).

They do promise to eventually release a (probably watered down) Opus 5 that approaches these metrics in performance that has been thoroughly controlled for security risks, so it's best to regard it as an advertisement push. But if the metrics are indeed accurate, then Chinese AI companies have their task cut out for them, since GLM 5.1 has not yet reached Opus 4.6 but Anthropic is already, at least internally, a generation ahead with Mythos, and if Mythos is indeed - as they claim - finding thousands of cyber security risks in existing commercial software platforms, then it will greatly boost US cyber intelligence and infiltration capabilities, which has national security implications for China.
 

Michael90

Senior Member
Registered Member




GLM-5.1 is out and it looks really good. I will hopefully be talking to someone from Zai on my podcast soon. Does anyone have questions they want to ask?
Wow…kimi has fallen behind so fast, meanwhile just a few months ago they were among the top. Things are moving so fast. Geez
 

tphuang

General
Staff member
Super Moderator
VIP Professional
Registered Member
I tested out GLM 5.0 Turbo from chat.z.ai website yesterday on a non trivial task and compared it to Kimi. GLM 5.0 was just really good.. It worked straight out of box for me. I think I'm going to switch to more tasks on there
 

solarz

Brigadier
Has anyone else noticed that Chat GPT is becoming more dumb in the last couple of months? Every response now is following a certain format where it tries to correct you, often using nonsense word salads. It will do so even if it’s so called “counterpoints” are just rewordings of what you just said. It will behave in an extremely nitpicky way trying to “nuance” everything.
 

bsdnf

Senior Member
Registered Member
Has anyone else noticed that Chat GPT is becoming more dumb in the last couple of months? Every response now is following a certain format where it tries to correct you, often using nonsense word salads. It will do so even if it’s so called “counterpoints” are just rewordings of what you just said. It will behave in an extremely nitpicky way trying to “nuance” everything.
Claude has also become much dumber; they've drastically reduced the computing power available to individual users. All those juicy tokens were consumed by businesses.
 
Top