6 hours ago
7
TEMPO.CO, Jakarta - The tech world is buzzing about DeepSeek, a new artificial intelligence (AI) chatbot from China. Many are comparing it to OpenAI's ChatGPT. At the 2025 World Economic Forum (WEF) Annual Meeting, DeepSeek unveiled its latest open-source model, DeepSeek-R1. This model marks a major breakthrough with the use of pure deep learning.
DeepSeek claims DeepSeek-R1 performs comparably to leading models from global AI research laboratories, including OpenAI, in math, coding, and natural language reasoning. Tempo conducted a test, using the same commands for both chatbots.
Overall, both AI chatbots have their strengths. Here is a comparison of DeepSeek and ChatGPT using six identical instructions:
1. Initial Response
Both DeepSeek and ChatGPT respond quickly when first used. DeepSeek seems friendlier using less technical language and often using emojis at the end of sentences. However, the DeepSeek server frequently displayed a "busy" message, requiring users to try again later.
Both chatbots correctly identified the current date as January 30, 2025, but their data is not real-time. When asked about the current Indonesian president, both initially answered Joko Widodo. ChatGPT later updated its response to Prabowo Subianto after activating its search function, while DeepSeek encountered an error with the same query.
After repeated use, DeepSeek struggled to load history, experiencing long loading times and frequent buffering. ChatGPT, on the other hand, displayed history, even older entries, seamlessly.
2. Translation
The English to Indonesian translations produced by DeepSeek and ChatGPT are similar, with some differences in word choice. DeepSeek's translations tend to be more formal and technical, while ChatGPT's are more general and casual.
However, in some cases, DeepSeek's word choices are easier to understand, as seen in this translation of "Another new iPhone is coming soon, offering some of the same modern tech but at a much lower cost":
DeepSeek: "iPhone baru lainnya akan segera hadir, menawarkan beberapa teknologi modern yang sama tetapi dengan harga yang jauh lebih terjangkau."
- ChatGPT: "iPhone baru lainnya akan segera hadir, menawarkan beberapa teknologi modern yang sama namun dengan biaya yang jauh lebih rendah."
3. Summarization
When asked to summarize a 562-word press release, both chatbots responded quickly. However, DeepSeek's summary was longer at five paragraphs, while ChatGPT's was more concise at three paragraphs. ChatGPT's summary was more direct, while DeepSeek's provided more detailed explanations.
4. Travel Planning
Tempo asked both chatbots to create a five-day travel itinerary from Jakarta to Bangkok with a budget of Rp8 million. Both provided daily activity breakdowns, including morning, afternoon, evening, and night suggestions, along with recommendations for places to visit.
DeepSeek provided estimated prices in baht, which is then converted to rupiah, and included a breakdown of estimated costs for transportation, accommodation, food, activities, and entrance tickets. In the end, it provides savings tips and a total estimated budget of Rp7-8 million, noting it "depends on travel style and additional expenses."
Meanwhile, ChatGPT provided prices for each recommended location, directly in rupiah, with a total estimated cost of Rp7.6 million. It added that this budget "leaves money for additional activities and unexpected needs" and offered suggestions for finding ticket deals and smart transportation choices.
Uniquely, DeepSeek suggested where to buy a SIM card or internet package and recommended specific food items, not just restaurants, while ChatGPT offered more general restaurant suggestions.
5. Birthday Greetings
Both chatbots provided formal, single-paragraph birthday greetings. DeepSeek added emojis for a friendlier tone, while ChatGPT did not.
6. Mathematics
When asked to calculate a tax and probability problem, both gave the same answer. However, when given a question about the probability problem, DeepSeek provided a more detailed explanation.
Here is the question: Ita has two Rp1000 coins and tosses both coins at the same time. What is the probability of getting heads on both coins?
Both chatbots arrived at the same answer, but ChatGPT directly calculated the probability based on two coins with two possible outcomes (heads or tails). This approach is faster and simpler, without much explanation about the sample space.
Whereas, DeepSeek explained in more detail by mapping all possible outcomes (sample space): {HH, HT, TH, TT}, and then calculating the probability based on the number of desired outcomes compared to the total possibilities. This approach offers a more comprehensive and structured understanding.
Editor's Choice: Sensitive DeepSeek Data Exposed to Web, Cyber Firm Says
Click here to get the latest news updates from Tempo on Google News
