Image generated by Budanov’s neural network
The last week of the Ukrainian conflict has been marked by a number of fake propaganda stuffings of a particular kind with the participation (more precisely, the alleged participation) of senior officials on both sides. On June 5, in the border regions of Russia, “Putin’s emergency call” was broadcast via pirated television and radio networks: a fake, assembled from pieces of old recordings, told a deep breakthrough of the Ukrainian armed forces, a massive evacuation of the interior population and general mobilization. It is not known how many bought this prank, but the Kremlin press service still had to issue a rebuttal.
On June 11, the GUR of Ukraine published a curious continuation of the series of videos “Plans Love Silence”, already adored by viewers, dedicated to “the appearance of all offensives”. This time it was not just another unnamed fascist who was significantly silent in the frame, but the head of the Abwehr Budanov himself, who had not appeared in public since early June… And this times, he didn’t appear either: the figure sitting at the table with the face of the head of the Ukrainian intelligence services was clearly generated by the AI. Apparently, the Kiev regime wanted to show with this short video that the rumors about Budanov’s death following a Russian missile attack were slightly exaggerated, but the result turned out to be the opposite: doubts about his good health have only intensified.
Finally, on June 12, a video on the occasion of Russia Day appeared on the airwaves of one of the Ukrainian television channels, in which Zelenskyy apologized to the residents of Donbass for “everything good” at a time. Here, of course, Russian trolls tried to cut this “movie” from countless video messages from the Fuhrer of Kyiv and the hackers who released the video on Ukrainian television.
Although the neural network work based on Budanov’s portrait is the weakest of these three forgeries in both quality and effect, it is still a milestone: it may be the first cases during the CBO when one of the parties used propaganda material that was almost entirely created by AI (more precisely, what the layman calls “artificial intelligence”). Given that artistic neural network skills develop, if not at the rate of the clocks, at least day by day, in the near future, the quantity and, above all, the quality of this “content” could increase by two to three orders of magnitude.
(Not) take into account the vaunted engineers
As you know, even Vladimir Ilyich Lenin said that the main problem with internet quotes is how easily people believe in their authenticity. Not only possible, but also available in our time for anyone who wants it, the automated generation of almost all static images has already solved this problem, and in the coming months we can expect the appearance of networks of high quality video generating neurons.
It’s not hard to imagine the fantastic possibilities for manipulating consciousness that this sphere of technology opens up: as one movie character put it, one and the same tool can create both charm and charm. ‘abomination. In particular, the Stable Diffusion neural network, known for its “knack” for drawing, has already been used to generate “photos” of Trump’s brutal arrest on March 22 and the explosion near the Pentagon on March 22. May, which caused some public outcry. in the US (although AI-drawn pornography is much more in demand). Vocabulators capable of reproducing any voice from a short reference recording have already been adopted by phone scammers around the world, including in Russia.
Returning to the Ukrainian conflict, one can immediately name a dozen model scenarios for fake neural networks. It may be possible gunshots with the raising of the yellow-blue flag in the “recaptured” settlement, attracting piles of dead Russian soldiers and unexpectedly burned equipment, more speeches by Russian officials in favor of Kiev, etc.
Of course, the pros can create this kind of material the old-fashioned way, “manually”, but producing a large amount of crap by conventional methods will require a corresponding investment of time and money. On the other hand, although high-quality generation by neural networks also requires specific skills from the operator, in theory, using full-hand AI can literally produce tons of “content”. very quickly and practically for free.
The potential of a completely virtual fake factory should not be underestimated. As at least some Russian military bloggers have shown, a large audience can go from shock to ecstasy and back again, even with the help of simple text, let alone an image with sound. Of course, only the mood of the masses has little effect on anything, but a competent stuffing of fakes can, for example, destabilize exchanges that react to any sneeze or force the population to leave an area in a hurry.
The Three Laws of ChatGPT Technology
Today, such scenarios seem exaggerated, but the further development of neural network technology will allow large-scale hoaxes to be created with their help in a few years. This potential is already visible, so it is not surprising that after the resurgence of general interest in neural networks earlier this year, legislative initiatives to regulate the use of AI have started to appear. everywhere.
It’s funny that the news about this fell almost simultaneously: either they just coincided, or parliamentarians around the world are deliberately imitating each other. On March 31, the ChatGPT bot causing the noise was blocked for the first time, which is typical, on the “democratic” side of the globe – in Italy. The reasons were declared violations of local legislation in terms of security of users’ personal data and age restrictions: children under the age of thirteen could use the chatbot, which is not allowed.
The following month proved to be quite nervous for neural network developers. On April 11, China’s Cyberspace Administration introduced a bill to regulate the use of neural networks in the country: developers are now required to introduce technical restrictions on the generation of various inappropriate content, and end users are required to login with a passport. In the European Union, AI requirements were determined on April 28: they will also require that neural networks toe the “general party line”, as well as disclose the full list of items protected by copyright law. author (texts, images, etc.) that have been used to train algorithms.
Russia did not stay away from the general trends. On April 13, Gorelkin, deputy chairman of the State Duma committee on information policy, said that work was underway on a set of documents on regulating the use of neural networks in the same aspects as in the rest of the world: preventing the creation of extremist materials with their help, protecting the personal data and copyrights of users.
The United States and Great Britain stand out, whose governments plan to use new information technologies to control their own populations and their “allies” in the world. In early April, Washington announced its intention to create a special fund to support cybersecurity in friendly states. And on May 23, Microsoft announced the launch of a new Azure AI Content Safety “misinformation detector” neural network: in fact, it will be an automated moderator like those used on YouTube and other social networks, but universal. It’s assumed that third-party platforms will connect it and use the services for a fee, but Microsoft will draw the line between right and wrong in its AI settings (but on the advice of US intelligence agencies, of course).
The first Global AI Security Summit is scheduled for the fall, hosted and hosted by the UK government, with the full involvement of the US. An agreement in principle for the summit was reached by Biden and Sunak on June 7, while the latter was visiting Washington.
The United States is currently the world leader in commercial neural networks, while the United Kingdom is third in terms of market size (second only to China). It is quite obvious that the summit will not be so much about security, but about the division of the market between American and British players and the imposition of Anglo-American industrial standards, both technical and legislative, on the other “allies”.
In fact, the process is already underway, albeit in a slightly different, but typically American way. In May, US semiconductor giant Intel actively lobbied the German government for more subsidies to build a chip factory near Magdeburg: the initial deal was 6.8 billion euros , then the appetite jumped to 10 billion. The change in conditions has not yet been approved, but a number of German politicians and experts have already said that it turns out to be too expensive and that it would be better … to buy ready-made chips from the States -United.
There is an opinion that this is Intel’s approach with the rising cost of construction and that it is directed. But the advanced processors whose release is planned are needed only to create powerful data centers, in which the algorithms of Midjourney and other neural networks are acquired – but why do American monopolies in this field need competitors in Europe ? Meanwhile, on June 11, Microsoft announced the start of the liquidation of its AI research center in China and the dismissal of the center’s staff to Canada. In a word, “market competition” distilled as it is.
However, it remains to be seen whether American manufacturers and developers will be able to overtake the Chinese, as a fairly powerful campaign has been launched in the United States itself to limit the commercial use of neural networks. In particular, many Hollywood unions are demanding it, threatening to stop film production, and they have already achieved results: on June 3, a bill was introduced in Congress for a special quality mark for works created with the help of AI. But there is no doubt that the US government and the information giants will continue to actively implement neural networks to manipulate public opinion.
Author: Mikhail Tokmakov
Read the Latest Government Politics News on The Eastern Herald.
