“Today, AI is the main indicator that “the future is already here”. Over the last few years, neural networks and artificial intelligence services have become very popular, and more and more people regularly begin to use them in training, work, everyday life, and just for the sake the world. Models”, – WRITE ON: ua.news
Today, AI is the main indicator that “the future is already here”. Over the last few years, neural networks and artificial intelligence services have become very popular, and more and more people regularly begin to use them in training, work, everyday life, and just for the sake the world. Models are constantly improving, upgrading, new competitors are emerging, and states like the United States are investing in the development of hundreds of billions of dollars.
UA.News also decided to test artificial intelligence in its work. We asked for the most famous AI service-Chatgpt-to write for us a plan for what the scenario of termination of the war in Ukraine should be. Then we asked the same word for the same at his Chinese competitor – DEEPSEEK. Both gave us their answers, which are similar, but different in places. And the second request for AI was the request to create a project of a peace treaty between Ukraine and the Russian Federation. And if Chatgpt expressed his thoughts without problems, then Deepeseek just refused to answer.
What happened to us and what conclusions can be drawn – read in our material.
Peace Achieving Scenarios: Comparison Chatgpt and Deepeseek
Both models had the same question. We suggested that they imagine that they are a speaker in which UA.news is interviewed. The task was to write a plan for the end of the war between Ukraine and Russia.
Number one item Chatgpt called military stabilization and termination of hostilities. In his opinion, the priority step is to reach a stable truce between the parties to the conflict. This requires the effective work of international mediators and security guarantees for both parties. At the same time, Deeseek called “Recovery of Dialogue at the International Level” as the first point. He proposed to attract key international players (UN, EU, USA, NATO, China and Turkey) to create a negotiation platform. Chinese AI paid special attention to the need for neutral intermediaries.
Second point Chatgpt – international negotiations and diplomatic settlement. He proposes to hold multilateral negotiations with the participation of international players (UN, EU, USA, etc.) to develop a peace agreement. All this should be based on international law and UN Charter. But Deeseek wrote about the recognition of territorial integrity and sovereignty of Ukraine. This principle, in his opinion, should be clearly enshrined and be the basis of negotiations.
Third point From Chat – withdrawal of foreign troops and “demilitarization of conflict zones”. He offers the gradual exit of foreign military formations from the territory of Ukraine and the creation of demilitarized zones under the control of international peacekeeping missions. Exactly the same as a step by number three suggested Chinese AI.
Item number four According to Ghatgpt, reintegration of temporarily occupied territories. He calls for the implementation of a comprehensive reintegration program for occupation regions, including infrastructure restoration, law and order, and social and economic support to the local population. But DEEPSEEK in this paragraph was only mentioned by the need for security guarantees. He proposes to get them in the form of memoranda or treaties, as well as involve Ukraine in Euro -Atlantic structures.
The fifth point – Economic Recovery and Involvement of International Assistance. Chatgpt believes that at this stage, Ukraine’s partners should intensify economic support programs and stimulate economic recovery. His Chinese “colleague” here again agrees almost word to word. However, it emphasizes that it is necessary to raise funds for recovery primarily at the expense of aggressor country.
The sixth step From American Shi – “social cohesion and national dialogue”. Chatgpt proposes to conduct “All -Ukrainian dialogue to strengthen national unity, overcome differences and form a joint vision of the future of the country.” He also calls for reconciliation and understanding within Ukraine. However, what is meant by “dialogue” is not fully understood. The Chinese model in the sixth point is more practical: it proposes international investigations into war crimes and human rights violations. All guilty should be fair punishment in lawsuits.
Finally, Seventh and Last Item from Chatgpt – prescribing security guarantees and prevention of future conflicts. He says that at this stage it is necessary to conclude international agreements that will provide long -term security guarantees for Ukraine, including the possible placement of international observers and participation in collective security systems. Deeseek, however, proposes to pursue a policy of “social reintegration and long -term peace” as the last step. It is about helping veterans and displaced persons, as well as educational initiatives “to strengthen peace, tolerance and unity in Ukrainian society”.
Next – even more interesting. We asked Chatgpt to write a project of a peace agreement between Kiev and Moscow. The American AI immediately published a very detailed plan: on the issues of security guarantees, the status of temporarily occupied territories, sanction and economic blocks, political cooperation in the future, international control, borders, etc. But Deeseek, thinking about about 30 seconds, said that his server was busy and suggested trying later.
We tried again, and over a while – again. The answer was the same. At the same time, the other inquiries of Chinese Shi responded immediately and no server was busy. In this case, it may be a censorship of certain topics by the Chinese government – or in fact, selective technical failures are in the new service.
By drawing conclusions from the answers of both models, you can see that they are something similar. It should also be noted that both scenarios – both from American and Chinese AI – are quite pro -Ukrainian and rely on the rules of international law and the UN Charter.
Both Chatgpt and Deeceek pay attention to the importance of security guarantees, economic and infrastructure recovery, international assistance and supervision, etc. And in places “Dipsik” even more pro -Ukrainian than “chat”: in particular, he proposes to restore Ukraine primarily at the expense of the aggressor.
On the other hand, it is noticeable that the answers of the American service are more spreading and detailed. Whereas Chinese “ko Lega ”often corresponds to the thesis, concise and somewhat“ skimmer ”, but more pragmatically and specifically.
Conversation with the expert
We turned to an artificial intelligence expert Alexander Teplyuk But they asked him to comment on the results. The specialist draws attention to the technical differences between the two models and explains why their answers differ in style and accents. In his opinion, there are no special ideological reasons here – rather it is about different technological approaches.
“Why do the answers differ? Because the data they study on is all basic. Chinese models learn for information that is actually closed by China’s internet. There are otherwise covered by information than in Europe, in Europe or in the US. Therefore, he builds prerogatives on information he analyzes. He does not come up with anything himself, but examines existing information. We need to remember that each company has its own date on which their model learns. That is, there is no ideology here, it is rather technical issues, ”Tepluk said.
He also commented on why Deeceek refused to answer some questions. According to the expert, this model really has technical troubles and problems with server capacity.
“Servers DEEPSEEK They have been sagging for several days. Their infrastructure is not designed for so many users. Most companies pay great attention to infrastructure so that a large number of users can use the service at a time. And here DEEPSEEK Not like that, he worked a little in another direction. I have not been working with him for a few days, ”Tepluk explains.
Further, our interlocutor shared his thoughts on competition between models and the future Shi sphere.
“Who can win this competition is difficult to say. Can win at all the third. Today, everyone uses the transformer architecture, which appeared in 2017. Almost all generative models work on it today. It improves and refines, but does not change fundamentally. A new company with the best architecture may appear in a year or two. It can become more effective. And the cost of using will be cheaper.
I have a question at all: is not funded DEEPSEEK The China’s government. First, to make such a cheap price, and secondly, to make such a hip. Because he appeared a few months ago. But Hype began only now. When Trump came and stated that he would allocate 500 billion per s. Why did no one talk about her before? She was open, ”Tepluk asked questions.
At the end, the expert gave advice on the use of artificial intelligence exclusively for UA.news readers. He advises to use several models at the same time and compares this with a trip to several doctors to be sure to be sure of the diagnosis.
“I advise you to use several models at once. I personally use Ghatgpt, DEEPSEEK and Gemini simultaneously. And sometimes I use some others. New models are constantly emerging, many of them. But these three models are basic. I recommend that everyone do not use only one model. It’s like communicating with a doctor: that’s when you asked three doctors and have a diagnosis in three, the diagnosis was made accurately. So do not trust only one model – this is my recommendation, ”Alexander Teplyuk summarized.
Nikita Trachuk