Artificial Intelligence thread

[tphuang]

Any reason why Bytedance does not have a benchmark coding and general llm. Video generation is where they dominate but they are nowhere to be seen in coding benchmarks. Or it it that they just don't focus on it.

https://twitter.com/i/web/status/2001907871223812563

they are certainly not as good as others in coding, but recent update apparently has been an improvement. I mean they seem to focus on stuff that customers would use on their apps.

As a daily user of all the major Chinese chat LLMs, Doubao probably has the best user experience for general non-thinking knowledge queries. It usually responds the fastest and includes pictures and relevant Douyin videos. The voice and audio interface is the smoothest, and it's a very satisfying experience asking questions using voice and listening to the response.

For drawbacks, unlike other chat apps which all use an abstract logo, Doubao uses an AI avatar mascot, which is annoying to me but maybe appeals to more mainstream people? There's also no dark mode support which all other apps support.

For more in-depth thinking type questions or things where I want lots of data, I like DeepSeek thinking mode or Kimi K2 since they tend to respond with more tables making things more readable.

Recently The Information wrote an article on OpenAI's recent struggles and one highlight is that mainstream users didn't care much about the benchmark of their thinking models and prefer speed of response, so their thinking models don't really drive subscription conversion. Perhaps it's a trap for product focused LLM companies to try to spend all their efforts in beating benchmarks and Bytedance has the right priority on response speed for the masses.

right, it seems like a lot of what they do is to improve app experience rather than productivity gain and deep research that others are focused on.

https://twitter.com/i/web/status/2002061341540884505

this is very interesting. ByteDance continues to work with more OEMs. They are ambitious, 150 to 200m AI phones is their current goal.

 

[bsdnf]

ByteDance's Doubao consumes an average of 50 trillion tokens per day, matching or even surpassing Google's figures (they haven't updated their latest data yet).

 

[tphuang]

https://twitter.com/i/web/status/2002202322164981789

I think the token usage for Xiaomi models will go up quickly once it releases AI phones using its AI assistant based on these great models.

 

[bsdnf]

I believe it will be extremely challenging yet inevitable for AI agent phones to become mainstream. TikTok will support this trend, while Alibaba, Tencent, and Pinduoduo will strongly oppose it.

AI agents will ruthlessly disregard traditional methods of capturing user attention, such as recommendations, promotions, and advertisements. Users may not even need to open apps, which would be another blow to shopping apps(already heavily impacted by live streaming and short-video sales).

Traditional picture-based and personalized push ads will become obsolete, but it will not affect sales models like TikTok’s, which rely on video consumption as a traffic gateway. Just like television commercials replaced newspaper commercials. User attention will become an extremely scarce resource, making the literal task of "catching the eye" a significant challenge.

 

[tphuang]

https://twitter.com/i/web/status/2002744608246112301

This is the kind of stuff that the west just can't compete in.

You can bring ideas to market so fast in China. The entire supply chain is all there.

https://twitter.com/i/web/status/2002744608246112301

this is the first time I have heard of Tuya/涂鸦 AI platform

Please, Log in or Register to view URLs content!

, but it looks like they can get you everything you need to design and bring to market on a new AI toy.

 

[bsdnf]

https://twitter.com/i/web/status/2002744608246112301

This is the kind of stuff that the west just can't compete in.

You can bring ideas to market so fast in China. The entire supply chain is all there.

https://twitter.com/i/web/status/2002744608246112301

this is the first time I have heard of Tuya/涂鸦 AI platform

Please, Log in or Register to view URLs content!

, but it looks like they can get you everything you need to design and bring to market on a new AI toy.

Is this running locally on the toy or in the cloud? Previously, Huawei's AI dolls used cloud computing power, with a model of daily free attempts plus a membership subscription.

 

[bsdnf]

I saw ads for "Ant Afu" at bus stops, touting it as an AI health assistant and consultant. It was originally called Ant AQ, it seems they've chosen to specialize in a different market segment because its download numbers have significantly lagged behind Qwen.

Alibaba's internal competition has gotten a bit crazy. Not only are the Ant and Qwen teams each developing their own models, but they're also launching their own AI apps.

 

[tphuang]

Is this running locally on the toy or in the cloud? Previously, Huawei's AI dolls used cloud computing power, with a model of daily free attempts plus a membership subscription.

i would imagine on cloud. Otherwise, I'm not sure why Alibaba Cloud account would be talking about them.

 

[9dashline]

Please, Log in or Register to view URLs content!

 

[tphuang]

https://twitter.com/i/web/status/2002885054070919521

China Telecom update on its data centers. The Lingang one is interesting, since it's a 290MW & 300k card cluster for 25B RMB. Also, if it's just 25B RMB for such a data center, then a 1GW data center probably only cost like 80B RMB.