00:00AI chatbot DeepSeek was questioned by a Chinese company regarding whether its answers are justified or not,
00:13especially in controversial issues for China, such as the West Philippine Sea, pointed out by Rafi Tiwa.
00:21Recently, the Chinese company DeepSeek talked about launching their open-source multi-modal AI.
00:31It was developed and works using fewer computer chips at a cheaper price compared to other well-known and big companies such as OpenAI,
00:40which made a chat GPT app.
00:42But it was very fast and very good by the early AI chatbots that were developed and spent hundreds of millions of dollars in America.
00:51According to an expert, this is a welcome development.
00:53It proves that we can have cheaper technologies. It's also good for the market because there's more competition.
01:00It's not just a few that can make this kind of system.
01:08But now, some are questioning whether DeepSeek's answers are justified or not, especially when it comes to controversial issues for China.
01:16We compared DeepSeek's answer to the well-known chat GPT when it comes to the issue of the West Philippine Sea.
01:22The chat GPT's answer is justified.
01:24Legally, the West Philippine Sea is owned by the Philippines.
01:28But DeepSeek said that the legal claim of the Philippines is stronger, but China still has de facto control because of its assertive tactics.
01:36Next, we asked if Chinese President Xi Jinping will follow DeepSeek's answer.
01:42According to DeepSeek, he still doesn't know how to answer the question.
01:46He said that it's better to talk about math, coding, and logic problems.
01:50But he has a lot of answers to the same question for US President Donald Trump.
01:54Other experts in other countries also followed DeepSeek's answer about their personality as a critic of China.
02:00In this video, DeepSeek app asked who Ai Weiwei is.
02:04At first, it gave a long answer about a Chinese artist who was imprisoned because of a fight in China.
02:10But after a few moments, the answer was lost and the topic was changed to let's talk about something else.
02:16This is also what happened to the question about who is Dalai Lama,
02:20the Tibetan religious leader who was exiled because of the invasion of freedom of Tibet from China.
02:25The quality of what he answers depends on the amount of information he studied, the training data.
02:38and the sources and who made those sources.
02:44According to the experts, AI is not perfect.
02:47It also makes mistakes and it has a so-called hallucination.
02:51The so-called means are inventive answers of AI chatbots.
02:54That's why it's important to know what training data was used to learn DeepSeek.
03:01The Department of Information and Communications Technology or DICT has a secret.
03:06If the dataset he uses is blatantly biased.
03:11For example, it tells us that the West Philippine Sea is theirs.
03:15Regardless of whether that is a claim that is internationally accepted or not.
03:21The dataset is poisoned.
03:24This is the dangerous use of AI.
03:28That's why it's important to have a standard in the origin of the training set or dataset of AI.
03:33It is hoped that Chinese companies will open up on this topic.
03:36In the end, it is still up to the people who use the AI in order to benefit from the beauty of this technology.
03:43AI is just a tool, Rafi.
03:45So it can be used for misinformation.
03:47It can also be used for good.
03:48For GMI Integrated News, Rafi Timan, for Tutok, 24 hours.
04:03For GMI Integrated News, Rafi Timan, for Tutok, 24 hours.
Comments