New trends

What algorithms are used to develop neural networks?

Why are the contradictions in the field of artificial intelligence intensifying in the so-called collective West? What trends does this escalating conflict very clearly outline? Let us figure out what is happening. 

French pie

The AI Action Summit was recently held in Paris, with about 1,000 people from 80 countries taking part in the forum. They discussed the prospects and potential threats associated with the development of neural networks. Belarus and Russia were not invited to the event; however, this was entirely expected. Yet, many unexpected things happened there as well. As usual, they started out strong, discussing the provision of access to independent, reliable and trustworthy neural networks for as many people as possible. Well, in general, for everything good, against everything bad — a lot of beautiful phrases for the media, nothing more. The fact is that any neural network only learns what it was trained on. A lot was said about the ethical aspect of neural networks, but discussing this is like talking about the ethics of a felt-tip pen. After all, you can draw a splendid picture with it, or you can write a four-letter word. It is just a tool, and the issues of ethics lie with the user, not with the tool. The main disputes at the summit were covered up by the tinsel of general statements, yet if you look at what actually happened, a very interesting picture emerges.
Jamal Atif, Scientific Director of the French nation-wide AI programme (PEPR IA), started the fun by saying, “We are witnessing the coming to power of a new administration in the United States and deregulation. At the same time, we see the emergence of a new AI model from China — DeepSeek. We should talk about it too. Now we have open-source models that work just as well as closed models. And this is very interesting. Both China and the developers from DeepSeek have shown us that this can be done, that people are capable of this. New developments should be not only in the United States and China, but all over the world. We need brains, we need talent.” 
In other words, France has directly stated that it is not particularly happy to be trailing behind the United States and does not mind at all trying its hand in the field. Looking at what America will invent and hoping that it will be shared with Europe is not very promising. Proprietary developments would be much more useful; moreover, there is no need to pay for them. So France has outlined its ambitions to become a European leader in the field of neural network implementation. Yes, Europe so far is hopelessly behind the US and China in this industry, although France still has some grounds for such ambitions.

Demarche

Now we have come to the most interesting moment. As the summit came to an end, an international agreement was developed, according to which ‘AI must be open, inclusive, transparent, ethical, safe, secure and trustworthy taking into account international frameworks for all’. The wording is more than vague and not particularly binding, but all participating countries signed the agreement, with the exception of the United States and the United Kingdom. This really became a sensation of the summit. Why did they refuse to sign? They said that all this does not correspond to their national interests. Both the UK and the US expressed concerns that international regulation could slow down the development of neural network technologies, and criticised the accepted standards for collecting data needed to train neural networks. In the end, they proudly muttered that they were not ready to co-operate with authoritarian regimes that could use these developments for military purposes. That sounds especially amusing against the background of the rejection of restrictions on data collection. Naturally, neural networks in the United States will be used exclusively to create peaceful weapons that bring the light of democracy to people. Just for your information: if you use a certain American service, you can assume with 97 percent probability that you are participating in the training of a neural network.
Obviously, a decision was made to limit the distribution of their own technologies and developments, as well as to prevent any international interference in their research. If Washington and London had taken such a position a couple of months ago, it would have been very painful for the other participants. But the fact is that China has already presented its own chips for training neural networks, which are very effective in their task. 
The fact that the US is trying with all its might to put the already collapsed monopoly back together is quite natural. Yet in fact, we get a technological split and a neural network race. 
Everyone understands perfectly well that this is no longer an economic issue, but a strategic one. The main question here is whether the US and the UK will be able to outplay the 58 countries, including China, that signed the agreement. Or rather — only the US, since the UK cannot boast of any achievements in this domain, it just decided to play on the US side out of habit. In this situation, I would bet on China.  

Our prospects

Events around the neural network industry are developing extremely rapidly, and, naturally, the question arises: how will all this affect Belarus? Well, we have this conversation and the signing of a similar agreement yet to come. At the end of last year, Russian President Vladimir Putin invited his colleagues to join the co-operation in the international BRICS+AI Alliance, “The international alliance of national associations and development institutions in the field of artificial intelligence from the BRICS countries and other interested states will give a significant boost to this co-operation. We are launching this alliance today, at the Artificial Intelligence Journey conference, and expect that new foreign participants, including my colleagues, will join this event.”   
The work of the alliance has already begun, and it is here that Belarus should take the most active part. 

After all, Belarus has exactly the same advantage as France — we have our own nuclear energy, which makes our country very attractive for investment in the field of neural network development. Belarus can become a fairly serious player in the neural network market at least in our region if we do not miss the moment. 

CHIP: THE HEART OF AI 

It is when the USA comes and says with a sweet smile that they are ready to co-operate and share their technologies with the world that we will understand that the ‘hegemon’ did not succeed in securing a new leadership in neural network. Most likely, this may happen in the foreseeable future. After all, the US no longer has the opportunity to slow down its competitors with sanctions on the supply of equipment for training neural networks — China produces its own chips, and America is no longer a monopolist. Indeed, in conditions of fair competition, fighting with the whole world does not look very promising.

TO THE POINT

The attractiveness of France for investors in the field of neural networks is that it has a certain number of nuclear power plants that provide excess generation, sufficient for such ambitions. We have already entered an era when it is not enough to have only the technology; you also need to have the energy to use this technology. All this was reinforced by the recent statement made by Emmanuel Macron about the allocation of €109bn for investments in French neural network projects. The statement has been made, and France is laying claim to its piece of the pie that is just beginning to be prepared.

By Yury Terekh