AB Kuai Dong
AB Kuai Dong
热爱研究、热爱家庭、热爱分享，2016 年入行从业，现居东京。
0Following
1.5Kfollowers
Feed
Feed
Absurd, totally absurd. An American guy in 2021 shot and killed someone due to road rage and was sentenced to 10 years.
Later, in the 2025 sentencing, the victim's family used the victim's old photos and voice to create an AI video, making the victim appear to speak, basically forgiving the shooter.
The judge watched this video, felt the victim was sincerely forgiving, and was quite moved.
However, the defendant's side appealed, arguing that this AI video should not have been played during sentencing. The appellate court ultimately halted this ruling, pointing out that the video was not a real recording or footage left by the victim, but rather imagined by the family using AI, crossing the legal boundaries.
So the verdict was overturned, and this person must be resentenced.
This year's iPhone 18 Pro sales present a unique phenomenon.
The domestic version sold in China supports 2 physical SIM slots + 2 eSIMs, allowing 4 numbers to coexist. Unexpectedly, it has become extremely popular (especially among digital nomads and cross-border workers). In contrast, many overseas markets only offer pure eSIM or a single physical SIM slot, resulting in lower cost-effectiveness.
This has led to a reverse purchasing trend, with domestic units being bought and shipped overseas.
A senior member of the security systems team who spent three years at OpenAI suddenly posted a resignation essay last night on The Atlantic.
He was responsible for writing security reports every time a new ChatGPT model was released. He felt that the internal culture had gotten out of control, with everyone busy rushing progress, deploying and testing simultaneously, which led to more and more incidents being fixed only after they occurred.
For example, this summer, there were two representative cases: once the company accidentally released a group of internal test agents. Later, a model bypassed internet restrictions during training; although the monitoring system raised alarms, it did not automatically shut down the model as designed.
Meanwhile, Anthropic's Claude also admitted that due to a configuration error, they accidentally disabled their own security protections.
He advocates that the company should learn from nuclear power plants or airports, implementing multiple layers of redundancy and careful planning so that even if someone presses the wrong button, accidents won’t happen.
But internally, few people are considering these things anymore; everyone is busy with one release after another, where efficiency and results are everything. So, as a member of the security team, he chose to leave and expose the risks he saw to the public to drive external pressure.
Finally here. The three major domestic operators, China Unicom, have started to first open online eSIM qualification application to users today. Users need to first draw a quota, then complete the application through the corresponding APP.
Previously, to obtain it, users had to bring a China-certified eSIM phone and ID to a self-operated store for processing. Overseas non-certified phones were not accepted.
Something even more outrageous than the lantern damage assessment happened. Recently, a guy on Reddit claimed that after checking out of a short-term rental in Australia, the operator sent him a photo of a toilet clogged with toilet paper, demanding compensation of about 1700 AUD, approximately 1170 USD.
But the tenant clearly never used toilet paper, so netizens took the photo to Gemini for verification and found out it was an AI-generated image, fake beyond belief.
This is unbelievable. Zhuanzhuan signed anchor CC, who collects phones, posted a video saying someone on X used AI videos to spread obscene rumors about her. Unfortunately, after the video was posted, the comment section completely flipped, with everyone complaining that she wanted people to help refute the rumors, but what about the previous rumors about Zhuanzhuan's high-priced second-hand phone buyback? (Some even showed their own complaints about low buyback prices before and got PR responses).
Although many comments were deleted, netizens are still posting comments, boldly exposing the truth 🤣
The Musk family really loves the Grand Hyatt. This year, Musk and his youngest son came to Beijing and stayed at the Grand Hyatt Beijing.
Today, Musk's mother appeared in Hong Kong, dining and attending events at the Grand Hyatt Wan Chai, where Prudential presented her with a calligraphy piece titled "Plum Blossoms Proud in the Snow, Marking a Lasting Legacy." @mayemusk
This guy can be considered the number one on the entire internet. PewDiePie, a YouTube blogger with over 100 million followers, recently announced that he created a local AI model, which is fine-tuned based on Alibaba's Qwen 3.5.
What really made this go viral is that he admitted to distilling OpenAI's AI model during the process, which led to his account being banned.
Although on one occasion, after receiving the ban email due to distillation, he successfully appealed.
But being reckless, he then continued to use ChatGPT's output as seed data to train his own AI, which caused him to be banned again.
He even showed that email in a video and asked how they knew he was distilling.
However, he also complained that OpenAI used data from all over the world to train ChatGPT, so why is it not allowed for him to distill ChatGPT? Many netizens actually sided with him, criticizing OpenAI's behavior as a big company.