Artificial Intelligence thread
- Thread starter 9dashline
- Start date
The article explicitly says it is not preventable in such a way, because the creator used private pictures.
And the victim indeed had no social media presence.
Lovart ai launched
The fact that the demo video Lovart ai uses to exemplify their use case looks eerily similar to trending faux commercials that I have seen on X/Twitter for other gen-AI video production tutorials makes me do a double take.
For example, here is a tweet showing how to use VEO3 to do the exploding box demo commercial:
I don't know if that was the video that started the trend (likely not), but now every social media platform is clogged with copy-cats claiming this or that softwear app can do something that KLING or VEO already does, it's just rediculous.
I am really sceptical about whether a lot of the "innovative" AI startups we regularly read about are actually innovating anything at all, or are they just coming up with new ways of rebranding exisiting off the shelf and open source tech to appeal to increasingly specific market segments. After all, they know that in the rush to go "AI-first" they can count on a certain amount of FOMO to drive consumption.
And legitimate copycat companies are one thing, but then I also suspect the same companies of rebranding themselves just to preempt potential competitors. After all, creating the illusion of product diversity is a big part of all aspring monopolies, so how many of the startups we are going to see are really owned by the same parent companies just slapping on a slightly modified logo and user interface that puts the same buttons in a different order?
We see this in almost every other industry, so why would AI be any different? Realme, Oppo, Vivo, OnePlus, and iQOO are really all just remarketed from BBK Electronics. Audi, Lamborghini, Porsche, Bentley, Bugatti, Skoda, and Volkswagen are all owned by the Volkswagen Group. I predict the real AI giants will use the EXACT same cognitive tactics you see with other products around price or quantity anchoring, decoy pricing, illusion of scarcity, etc. to make users think they HAVE to have subscriptions to multiple AI tools that all do roughly the same thing if you know how to use them.
Last edited:
Alibaba now out with Qwen3-MT, a translation model that supports 92+ languages out there.
so question for fellow members, I'm going to have an upcoming episode on AI regarding to just the infrastructure part of things. But after that, it gets a little harder. I have used a lot of AI so plan to do a Q&A on that. But I need people to actually give me some questions to answer, so if you can reply to me and pm me with some questions/topics to explore that would be great.
Also, I'm seriously weak on the part with regard to training/building models and transformer architecture, so I'm looking for someone who can talk about that. If you know this stuff, please PM me.
Also, I'm seriously weak on the part with regard to training/building models and transformer architecture, so I'm looking for someone who can talk about that. If you know this stuff, please PM me.
I’m curious on the use cases of AI in assembly lines. Are image recognitions being use for QA and flexible manufacturing, etc. Do they need constant connectivity to large data centres or local edge computing.
Thanks
Image recognition has been deployed in industrial inspection for decades (dating back 40-50 years), with China's private sector adopting it for at least 20 years. Its integration into flexible manufacturing (agile/flexible production) gained momentum recently alongside Industry 4.0 initiatives.
Huawei's Pangu model leads China's industrial-grade AI models (with several competitors emerging). While Western industrial AI primarily leverages LLMs, multimodal, and vision models—where LLMs handle basic scheduling while enhanced LLMs and vision models support inspection—these remain auxiliary skills. Truly specialized industrial models like Pangu are architected around time-series data processing at their core.This enables real-time analysis of hundreds of thousands of streaming parametersacross production systems, predicting inter-parameter relationships to preempt failures while automatically syncing with ERP systems like inventory management. The outcome: high-efficiency, high-quality scalable production aligned with agile manufacturing goals.
Deployment flexibly spans cloud or on-premise—cost dictates the choice. Budget-constrained setups opt for cloud (requiring robust connectivity), while capital-intensive operations choose on-premise (remotely maintainable by technicians). For AI engineers, operational deployment is alwaysremote. Crucially, industrial AI rejects Western subscription models—a monetization desperation tactic. Hardware retrofits for on-premise industrial AI (costing hundreds of thousands to single-digit millions RMB) are marginal compared to infrastructure investments.
Recent fringe attacks in Chinese AI forums accusing Pangu of plagiarism—citing its QWEN training—merit eye-rolls. Pangu's industrial DNA centers on time-series processing.Its QWEN integration for human-machine interfaces (expanding to smart driving, IVI systems, mobiles) leverages QWEN's MIT/Apache2.0 license—now the global standard for commercial LLM foundations post QWEN2/2.5. Reasons: unparalleled open-source licensing and training stability (unlike Llama's notorious instability). Meanwhile, U.S. "open-source" models impose commercial restrictions amid a pivot toward proprietary models.
Key takeaway: LLMs’ Q&A paradigm fails in real industrial AI. True industrial AI processes massive streaming data(vision feeds, sensor networks, robotics telemetry)—a domain where China currently dominates. Huawei's Pangu already serves 200+ industries. Europe/U.S. lack comparable deployment scale and data access (China's SASAC enables central SOE data sharing; Western private firms neversurrender core data to IT vendors).