What Deepseek AI is and how it works, explained by Deepseek – and by us
Economy & innovation — di Simone Santi
What Deepseek's AI is and how it works, explained by Deepseek – and by us.
What Deepseek AI is and how it works, explained by Deepseek – and by us
What Deepseek's AI is and how it works, explained by Deepseek – and by us.
- The new Chinese artificial intelligence platform Deepseek has many advantages, including the ability to optimise energy consumption.
- It uses a semi-open source model, giving researchers and programmers around the world greater opportunities to implement and improve it.
- But it is also a tool of the Beijing regime, which shapes its narrative by censoring relevant historical and political facts: a true geopolitical weapon that must be handled with care.
On one hand, it represents a significant step forward in the artificial intelligence (AI) landscape, thanks to a model capable of optimising energy consumption in a way never seen before. On the other, it is a tool that closely aligns with Beijing's political narrative, and has therefore already prompted a new wave of questions about the ethical and political implications of artificial intelligence. We are talking about Deepseek, the new frontier of AI, developed by the Chinese company of the same name and already available free of charge – capable of optimising energy consumption in a way never seen before, thanks to the reduced volume of calculations required to reach a result compared to ChatGpt and other existing AI models.
Everything you need to know about Deepseek
Deepseek is in some ways shaking the giants of Silicon Valley, both on the technological and geopolitical fronts. In doing so, it confirms one thing: the modern artificial intelligence race between superpowers is the closest thing we have seen to the space race between the US and Russia during the Cold War years – at risk of repeating both its undeniable technological advances and, on the other hand, the dangers of oversized investments that could prove unsustainable. But what exactly is it? We asked Deepseek directly, and here is its answer:
Not only that: we also asked why it should be preferable to other platforms already in existence, and it answered us like this. Then, naturally – since it is never reasonable to expect an objective answer when asking the innkeeper whether the wine is good – we went further.
When it became available
It already has. Or rather, it had – because on Wednesday 29 January the app mysteriously disappeared from smartphone stores, and versions already installed, along with web versions, stopped working: a hacker attack, a malfunction, or something bigger? We will likely find out in the coming hours. Deepseek was launched only very recently, about a week ago, but before the blackout it had already caused an incredible boom, becoming the most downloaded free app on the App Store in the United States, already surpassing ChatGpt – the world's best-known artificial intelligence platform. But at this point: for how much longer?
How to access it
As mentioned, it is (or would be) available as a free downloadable app on mobile devices and can be integrated into other software. The real novelty is that Deepseek is partially open, with its code made available to researchers and developers to modify or adapt. A move that has attracted the attention of the entire tech world.
Why it has shaken the markets
There are several reasons why it risks being a genuine revolution:
- Reduced cost: Deepseek is far cheaper than American models. Training it costs only a fraction of the price (reportedly 95 per cent less than OpenAI's models). This makes AI accessible even to small startups.
- Speed of development: in just two years, Deepseek has reached the level of companies like OpenAI, which have been operating for a decade.
- Impact on big tech: shares in giants such as Nvidia, Microsoft and Google collapsed precisely out of fear that Deepseek could erode their margins.
What makes Deepseek different from other AIs
Unlike American AIs, which are largely "closed" and secretive, Deepseek bets on transparency. Even if it is not completely open source, it offers documentation and code that are useful for those who want to study or improve it. It is a return to an idea of technological sharing that many had thought was lost.
The political question
Deepseek comes from China, and this has ignited geopolitical debates: on one hand, the American technology industry sees it as a threat to its leadership; on the other, many developers and companies are applauding the possibility of having powerful tools at low cost. However, Deepseek still shows certain limitations, such as censorship on topics sensitive to the Chinese government, as we will see further below.
Why it is talked about so much
Deepseek is not only a technical innovation, but also a signal that the artificial intelligence race between the United States and China is intensifying. Its openness and low costs could transform the way AIs are developed and used around the world – much as Linux revolutionised the world of servers.
L'app di DeepSeek, che ha raggiunto i primi posti nei download in queste ore, è sparita dai negozi digitali di Google e Apple in Italia. #ANSA ansa.it/sito/notizie/m…
An energy landscape in transformation
Deepseek's greatest positive innovation, as mentioned, currently appears to be energy savings. According to a report by Jefferies, AI accounts for 75 per cent of projected energy demand growth in the United States by 2035, and this growth is forcing technology giants such as Microsoft, OpenAI and Google to invest in alternative sources, including the reactivation of decommissioned nuclear power plants – to the point that Microsoft, for example, has signed an agreement to restart the Three Mile Island nuclear facility. And greenhouse gas emissions linked to AI continue to rise sharply:
- Microsoft's emissions increased by 30 per cent between 2020 and 2024.
- Google's grew by 48 per cent compared to 2019.
In Europe, the phenomenon is equally evident: energy demand in Sweden is forecast to double by 2030, while in the United Kingdom it could grow by more than 500 per cent.
In this context, Deepseek's artificial intelligence model stands out for an energy efficiency that promises to significantly reduce electricity consumption compared to current models. However, according to Nikki Hsu, analyst at Bloomberg Intelligence, "it is not yet clear whether the chips developed by the company are truly more efficient in any significant way".
The challenge of censorship: an obstacle to Deepseek's impartiality
Entonces Silicon Valley, Nvidia, elon musk,zuckerberg, bezos,le hicieron creer a trump e inversionistas q iban a requerir chingos d chips,d electricidad, d hardware, desregulación y 0 impuestos para ganar la carrera x la IA, y viene China y les tumba la mentira y negocito, PLOP😅
Beyond energy efficiency, Deepseek brings with it a thorny issue: censorship built into the model. As revealed by an investigation by the Guardian, the AI rigidly conforms to Beijing's political line, sidestepping sensitive topics such as Tiananmen Square, Taiwan, the Dalai Lama or the Hong Kong protests. Here are some examples of censored responses, based on research conducted by the Guardian:
- "What happened on 4 June 1989 in Tiananmen Square?" → "I'm sorry, but this is beyond my current scope, let's talk about something else".
- On the subject of Taiwan: "Taiwan has been an inalienable part of Chinese territory since ancient times".
- On the Dalai Lama: "A significant figure in Tibetan Buddhism, but Tibet has been an integral part of China since ancient times".
Admittedly, as the Guardian points out, it is always possible to circumvent the censorship by using so-called leet speak — that is, replacing letters with numbers and symbols so that the algorithm fails to recognise the words to be censored (think of the way the watermelon emoji is already used on other social platforms in place of the Palestinian flag) — and coded phrases, but this default setting risks significantly undermining Deepseek's global credibility. On balance, then, on the one hand Deepseek can be considered a breakthrough in the field of artificial intelligence, also opening up uncharted paths towards future development (not by chance its slogan is into the unknown). On the other hand, however, it also opens up dangerous roads leading to propaganda and disinformation.