Moscow State University has announced the creation of “the world’s second or third most powerful” supercomputer, having purchased components for it through a Chinese firm trading on AliExpress. “MSU-270” cost the federal budget at least 2.8 billion rubles: that’s how much, according to the state procurement website, MSU paid for components under the brand name SOLAR PEAK (this brand does not exist on the world market). Only MSU employees are allowed to conquer the “Solar Peak”, whereas earlier applications for computing on the top supercomputers “Lomonosov” and “Lomonosov-2” could be submitted by any academic research institutes and universities throughout Russia. The main academic supercomputer will specialise in the subject of artificial intelligence, and the core AI institute at Moscow State University is headed by Katerina Tikhonova. The high-ranking academic boss realises that Russia is now “following trends rather than creating them” in this “exascale race”. But, according to Tikhonova, the country has “a more pragmatic approach to the development of artificial intelligence”, “conditioned by the tasks and challenges Russia faces” – this is “application in UAVs and in the oil and gas industry”. Already, all four laboratories of the AI Institute are engaged in “dual-use” science.
T-invariant tells us why Russia’s newest and most modern university supercomputer, which by all accounts should be at the top of the world’s TOP 500, is not listed there at all. This is very atypical for civilian supercomputers of large universities. Before that, there was one famous case in China, when in 2021 it was announced very late that the country’s TOP-500 level computers had started working.
This is the second T-invariant investigation into supercomputers. The first story told how law enforcers destroyed Russia’s brightest supercomputing company and have been pursuing its founder for five years.
From left to right: Vsevolod Opanasenko, Viktor Sadovnichy, and unanimous leader Boris Gryzlov at the MSU supercomputer centre, 2010. Photo: MSU website.
How the university learnt about the launch of MSU-270
The staff of MSU and even researchers who perform calculations on the supercomputer “Lomonosov-2” and are in constant contact with the Research Computing Centre (RCC), learned about the launch of “MSU-270” from the media. At least 10 interlocutors at the university told T-invariant.
The launch was announced byMSU rector Viktor Sadovnichy on 30 August 2023. “We are officially opening a new supercomputer created by us. It is 400 petaflops. It may be the second or third in the world in terms of power,” TASS quoted him as saying. The MSU press service released more detailed information on 1 September: the machine is planned to be used to create Russian language models similar to ChatGPT.
“We learnt about it from the news. It [the new supercomputer] took many people by surprise in general. We were Googling in our group to see what came out. Various odious personalities, let’s not name them [Sadovnichy at a press conference at TASS], started saying that this is the best computer in the world, an ‘analogue computer’. Our laboratory was constantly working on Lomonosov-2, and we were not aware of it at all, so we started contacting colleagues, gathering information bit by bit. It became clear that everything was done in secrecy,” a former university employee with experience of calculations on MSU supercomputers tells T-invariant.
The main news about the life of scientists during the war, videos and infographics – in the T-invariant Telegram channel. Subscribe so you don’t miss out.
Industry experts interviewed by T-invariant refused to seriously comment on MSU rector Viktor Sadovnichy’s words that the resulting machine is “the second or third most powerful in the world.” “No one will discuss this profanity. They made a decision not to submit to the rating [TOP-500] of supercomputers. Well, then there is nothing to comment on,” the head of one of the supercomputer companies told T-invariant.
MSU-270 supercomputer.
What do they officially say about “MSU-270”?
Thepress service of MSU reports the following information. The computer will help in conducting scientific research in the field of artificial intelligence (AI), solving tasks for the development of domestic software. The computer with a new architecture based on the active use of graphics processors will form a single computing cluster with the supercomputer “Lomonosov-2”. The total performance of the new supercomputer will be 400 AI Petaflops (the accuracy of calculations is not specified). The architecture of the computer system was “inspired” by the advanced samples of implemented supercomputer projects in the best universities of the world, and the technologies used are based on the practices and existing developments of leading manufacturers. A network with a bandwidth of 200 Gbit/s is used as a computing network. It has high reliability and is characterised by minimal latency. The storage network has similar indicators. For management and integration with external systems, a network with a bandwidth of 100 Gbit/s and the necessary redundancy of components was chosen. The supercomputer complex also includes new power supply, cooling and communication systems. Domestic nodes and components were actively used in the development of the complex.
What calculations are planned to be performed according to official information
The order “On the effective functioning of the research hardware and software complex “270 years of MSU” appeared on 7 September 2023 (available at the disposal of T-invariant). A steering committee headed by Sadovnichy was formed, which includes 13 more people.
The same order lists the main computational tasks:
- predicting the quality of heterogeneous channels in data transmission networks based on probabilistic models and machine learning methods.
- development of combined molecular modelling and machine learning methods
- development of AI algorithms for working with graph-type data
- development of machine learning matmetods for processing large textual information
- intelligent image analysis for a wide range of tasks, including high-throughput plant phenotyping and precision farming tasks
- development of neural network approaches to mega-sense plant tasks
- intelligent analysis of big data from multi-wavelength telescopes and space missions, dynamic modelling of galaxy evolution from 3D-stereoscopic data.
All MSU supercomputers are usually supervised by the MSU Research Computing Centre (RCC). This time, the press service emphasised that “MSU-270” will become a working tool of the Faculty of Mechanics and Mathematics and the Faculty of Computational Mathematics and Cybernetics (CMC), as well as the MSU scientific and educational schools “Mathematical Methods of Analysis of Complex Systems” and “Brain, Cognitive Systems, Artificial Intelligence”.
“MSU-270” can accelerate “the emergence of solutions that can help teachers control the attention of younger students in lessons,” Igor Sokolov, dean of MSU’s VMK faculty, told TASS at the time. “We are now engaged in helping teachers: younger schoolchildren lose attention very quickly. The teacher has to quickly track this and react, and this is literally a few minutes. Therefore <…> [we need] systems that will tell the teacher who to pay attention to,” the agency quoted him as saying.
Where will the money come from?
It is known about two orders of the Russian government to allocate funding for “MSU-270”. on 16 December 2021 , a documentappeared , which set the task to allocate in time in the same 2021 MSU 1 billion 182 million rubles for hardware and software complex for research in the field of AI. The second document is dated 31 March 2022 – it talks about 2 billion 424 million roubles for the same purposes.
Judging by the state procurement website (thefirst and second contracts), MSU held two tenders with a single bidder. One contract was for 1 billion 935 million roubles, the second for 899 million roubles. That is, a total of about 2.8 billion roubles.
Is it possible to estimate the budget of “MSU-270”? There is no open information about how much it cost for one or another country to assemble supercomputers included in the world TOP-50. There are only experts’ assumptions. According to Sergey Abramov, the largest Russian supercomputer specialist , at the moment the number one supercomputer in the TOP-500 rating may cost up to $250 million.
If we count formally, “MSU-270” cost the Russian budget about 25-28 million dollars. The experts interviewed by T-invariant do not dare to estimate how much of this sum settled in import chains. However, the purchase scheme was quite typical for the recent “sanctions“ years, and we will talk about it below.
Who built the supercomputer for Moscow State University?
The contractor of both contracts for the creation of the MSU supercomputer was the company Argo Technology EastLLC . One of its founders is Andrei Kuchinsky. Previously he worked as a sales director at Hewlett-Packard, EMC, Dell EMC, Lenovo. In EMC Russia and CIS he held the position of Partner Relations Director.
Now his other company, Main Software LLC, is better known on the market. The most notable are its contracts to create cloud services as an import substitute for Western companies that have left: for Megafon or Bilayn, as well as for the state cloud system “Basis”. Argo Technology East also patented the Argo data storage system (DSS), which was actively promoted on the Russian market by the largest vendor – Lanit Group of Companies. The company has long been engaged in import substitution of key technologies, and information about a Swiss company with a similar name (Argo Technologie S.A.) has appeared at least since 2015.
Interestingly, back in 2021 , it was reported that Argo storage systems were used in the Platon system (truck toll collection service), as well as in the infrastructure of Modul (established in the early 90s, founders – the largest organisations of the Russian military-industrial complex: NPO “Vympel” and Radio Instrument Engineering Research Institute). At the same time, the high-performance Argo storage system was used in the work of Modul STC with the aim of AI tasks (in particular, for DGX NVIDIA clusters ); CNEWS reported that “the second most productive DGX NVIDIA cluster in Russia” was installed there.Thus, we can conclude that Argo Technology East was chosen to supply equipment for the MSU-270 supercomputer for a reason.
Interestingly, Kuchinsky is also one of the founders of LLC “Centre for Technological Support of MIREA”. This company was established by the Centre for Technological Support of Education of the largest technical university in Russia, which has been working for the Russian military-industrial complex for over 70 years – MIREA – Russian Technological University (formerly the Moscow Institute of Radio Engineering, Electronics and Automation).
It is important to note that the Centre is based at the Institute of Artificial Intelligence of this university.
T-invariant wrote to Kuchinsky with a request for comment. At the time of publication, we had not received a reply.
Andrey Kuchinsky
Synopsis
What else is MIREA known for lately?
MIREA has a lot to do with the topic of UAVs, including the use of AI capabilities. The university reports on its resources that it is producing UAVs that are being tested by the 45th Airborne Brigade of the Russian Armed Forces. The university has also recently developed an anti-drone radio countermeasures system in two versions: Vivaldi K and Vivaldi R. At the same time, MIREA bypassed Canadian sanctions andbought a Canadian radar detector for drone detection for 4.5 million roubles and three drone 1200 anti-drone guns for 400,000 roubles.
How and from what was the supercomputer for Moscow State University assembled?
Before Russia’s full-scale invasion of Ukraine, supercomputers in Russia were assembled in direct communication with leading technology giants: NVIDIA, Intel, IBM, HPE and others. The purchase of components was accompanied by service and engineering support. Here is how Sergey Abramovpreviously described the process in an interview with T-invariant:
– To bring a serious chip through parallel import is half a job. Every TOP-500 (especially TOP-100) level project is not just a simple purchase of hardware. When everything was normal, we did not just buy, we communicated directly with companies. As an example: I had to communicate with Intel vice-presidents: we discussed the project with them, so that a certain batch of processors would be available to us on time, so that they would fit it into their plans, so that the prices would be right, and so that we would promote the project in the media together. We were allocated hot engineering support in case of problems. Engineering support is critical: these systems are built on the newest chips, so they’re not exactly proven. And if there were problems, our developers and Intel (or AMD, or Mellanox, or NVIDIA, or Dolphin…) engineers would sometimes spend hours and days in close communication (email, phone – the problem must be solved)”.
How was the supercomputer “MSU-270” assembled now? In the documentation on the government procurement website, all purchased components are listed under the brand SOLAR PEAK (solar peak). This trade mark is listed in the declarations of the EAEU, but this brand does not exist on the world market.
T-invariant was able to find mention of the SOLAR PEAK trade mark in the declarations of conformity with the requirements of the technical regulations of the Eurasian Economic Union (EAEU) of Argo Technology East itself. A Chinese company supplying products under this brand is also listed there. It is called GREATZC TECHNOLOGY (HK) LIMITED, and it, for example, sells equipment on the AliExpress website.
The company is also on the lists of organisations supplying goods to Kazakhstan. There is a mention of the trade mark SOLAR PEAK from GREATZC TECHNOLOGY (HK) LIMITED in the declarations of the company “Imexprofi”. It is interesting that the company specialises in supplies from China and their website contains such an advertising text (stylistics, punctuation and spelling are preserved) :
“In the new logistical realities with a team of 25 employees, since the challenging year 2022, we have brought more than 900 deliveries from abroad, mostly with equipment.”
T-invariant has sent a request for comment to the GREATZC TECHNOLOGY mailing list and is awaiting a response.
All components in the tender are signed as SOLAR PEAK, but if you search for information on other parameters of purchased cables, nodes, modules and other things, all this equipment is identified as Nvidia products (and switches as HUAWEI), which stopped all official relations with the Russian market after 24 February 2022. T-invariant sent an enquiry to Nvidia’s European, Chinese and American offices and has not received a response so far.
GREATZC TECHNOLOGY, judging by its website, mainly trades in embedded computer hardware. “The segment is very specific. Most often it is on-board equipment for aviation, transport, including military equipment, of course. Stationary equipment can also be used, but less often – for working in harsh conditions: dust, dirt, quarries, mines, chemical production, something like that,” a market expert explains to T-invariant.
“Nvidia has a strict policy to counter illegal imports and circumvent sanctions. There are regular training sessions among employees. But if a Chinese company bought it legally, then further it can resell components, and what can be done about it!!!? As for service and engineering support, there can be two options. In the first case, a “black” end buyer, having no support from our side, loses 20-30% in performance, but nevertheless gets a working machine. In the second case, a conditional Chinese company can build a test cluster, test the new hardware, processors, see if everything works, receiving all the necessary support from Nvidia,” says T-invariant engineer familiar with the rules of the company.
Up-to-date videos about science during the war, interviews, podcasts and streamings with famous scientists on T-invariant’s YouTube channel. Become our subscriber!
Who has the right to perform calculations on “MSU-270”?
On the previous MSU supercomputers (“Chebyshev”, “Lomonosov” and “Lomonosov-2”), any team from any university or research institute in Russia could submit a request for calculations. “I don’t remember that anyone with an adequate request was refused. Moreover – the most interesting works of other organisations were reported to officials, the most advanced research was proud and collected in the annual report. The moment of prestige was very important: MSU showed that good science was done at Lomonosov. And scientific groups themselves had to inform in their reports (for example, grant reports) that the results were obtained with the help of “Lomonosov” and send half-page reports on the results to SIVC. Especially regional scientists were helped with access by RNF infrastructure grants. With “MSU-270″ it doesn’t work like that: as I understand it, only MSU scientists count on it,” one of the heads of the Natural Science Faculty of MSU (who recently resigned) tells T-invariant.
An executive of one of the largest companies in the industry also agrees with him. “The machine is controlled by special people who now play a much bigger role in any university – people in epaulettes. As far as I know, access is restricted even within MSU, only certain groups and people who are politically mature and vetted are allowed to work,” the interlocutor tells T-invariant.
“The procedure for gaining access has always been very transparent,” says an MSU employee, ” There are documents and a list of organisations that worked on the Lomonosov supercomputersin the public domain . “Previously, in order to get access, a letter was written to the rector of MSU with a request to grant access to specific employees (examples are available at T-invariant’s disposal)”.
Now the rules have changed, and the wording is as follows: “The head and responsible executor of the planned work should be full-time employees of Moscow University. As part of the subsequent execution of the work, it is possible that MSU teams may cooperate with other organisations”. Theoriginal message has been preserved in the archive, and there is also an example from the MSU Chemistry Department website. Those wishing to do so are given a questionnaire, which reads as follows:
Questionnaire to the application for the use of the supercomputer “MSU 270”
The “MSU 270” supercomputer will be used to perform computational experiments to solve problems related to artificial intelligence (AI) technologies. The architecture of the “270 Years of MSU” supercomputer is based on the use of high-performance graphics accelerators. The supercomputer is planned to be used for solving tasks that require large computing resources (related to the analysis of large amounts of data). To get access to the supercomputer, it is necessary to pass the initial selection of projects.
“There have always been long queues at MSU for computing, and there were a lot of people within the university who wanted it. Perhaps now MSU-270 is used for AI tasks, and they [tasks] are terribly voracious and require a lot of resources,” says a former employee of the Moscow State University Research and Development Centre.
The new supercomputer will apparently also be used by the MSU administration as a generous gesture to other selected colleagues and institutions. For example, in late November 2023, a cooperation agreement was signed by MSU and JINR (Dubna) and in a press release Rector Sadovnichy, among other things, says that he is “ready to use the potential of the Research Computing Centre of MSU, including the newest supercomputer “MSU-270″ with unique computational characteristics, which was put into operation on 1 September this year”. This is the only public statement by MSU found by T-invariant about granting access to the new supercomputer to anyone outside the university.
JINR has its own supercomputer called Govorurun, but it is, according to scientists interviewed in Dubna, “outdated and constantly breaking down,” and the new NICA collider will need a lot of computing power to operate fully. T-invariant plans to talk about the supercomputer in Dubna, created by one of the largest companies in the industry – RSK Group – in one of the next articles in this series.
What became of Lomonosov-2?
If you pay attention to the photos of “MSU-270”, you can notice a striking resemblance of the new machine with the old “Lomonosovs”, the interlocutors point out. This is partly explained in the university’s press release: “The computer with a new architecture based on the active use of graphics processors will form a single computing cluster with the Lomonosov-2 supercomputer.” One of the participants of computing at MSU comments on this: “it is important to declare that a new computer was built, not an old one modernised: both for purely marketing reasons, and it is much more pleasant to report and give interviews afterwards”.
“The black boxes in the photo are not Nvidia, but ordinary Supermicro (but still Nvidia cards and AMD processors inside. They are a kind of “clone”, a generic of the original DGX from Nvidia. It looks like they just removed all the panels of T-Platforms (the company that assembled the Lomonosovs – T-invariant), and the cabinets were left over from the long-disassembled Lomonosov-1, they were used again,” said a former MSU employee who previously worked on Lomonosov-2.
Why MSU-270 is a dark horse and what happened to the Russian TOP-50 supercomputer rating, which was maintained by MSU?
The experts interviewed by T-invariant are reluctant to assess the level of the new MSU supercomputer, because the university has not provided data to the world TOP-500, and the Russian rating, which for many years was maintained by the Research and Development Centre of MSU, stopped being updated after Russia’s invasion of Ukraine. This reason, of course, is not directly stated, but the following announcementis listed on the website :
“Under the current conditions, the practice of TOP-50 lists to objectively reflect the changes in both the application of high-performance computing systems and advanced supercomputing technologies is incapable and may contribute to incorrect interpretation of statistics, which contradicts the goals of the project. Therefore, we have decided to temporarily suspend the publication of new editions of the rating. We are grateful to all those who for many years have contributed to the development of this project and replenished it, and sincerely hope that in the foreseeable future we will be able to announce the resumption of publication of the lists. Compilers of the TOP-50 rating”.
The rating has stopped being updated because after 24 February 2022 all components (both for MGU-270 and for upgrading other Russian machines – three for Yandex, two for Sber and one for MTS) are supplied via grey or black imports,” several market participants interviewed believe.
Viktor Sadovnichy shows Putin the supercomputer centre. Photo: kremlin.ru
ChatGPT is ours! Who supervises and finances the work on AI at MSU
In his latest address to the Federal Assembly in early 2024, Putin said that Russia “must achieve technological sovereignty in the areas of generative artificial intelligence and large language models.” He noted that a new state strategy for AI development has been approved. “The capacity of domestic supercomputers should be increased 10 times by 2030,” the Russian president set the task. The topic of supercomputers (as well as everything digital) is included in the national project “Data Economy”, on which the Russian authorities have planned to spend 700 billion roubles by 2030 (the end of Putin’s current term).
Katerina Tikhonova
A leading artificial intelligence centre should be established at Moscow State University – this was announced by Kirill Dmitriev, CEO of the Russian Direct Investment Fund (RDIF), at a meeting with the president back in May 2019. Putin then supported the idea and called it a crucial element in the development of a national AI strategy. As a result, in 2020 the Institute of Artificial Intelligencewas established at Moscow State University , which, according to a number of media outlets, was headed by Katerina Tikhonova – Putin’s daughter and CEO of the National Intellectual Development Foundation (operating under the Innopraktika brand, founded by Moscow State University). The connection between the two organisations is described in the news on the foundation’s website. In addition, Innopraktika is implementing a large-scale multi-billion dollar investment project to create the MSU Science and Technology Valley in Moscow.
There is little information about Tikhonova’s participation in the life of MSU on the official pages of the university (there are mentions on the pages of Innopraktika and there is a page on the portal of scientific activity of MSU “Istina”). Among other things, Tikhonova also appears as a member of the Academic Council of MSU (archive, 2015).
Katerina Tikhonova at MSU in 2012. Photo: MSU website
The AI Institute has a lot of interesting research and publications (here is the annual report for 2023), including work on language models similar to popular chatbots. The task of creating a domestic answer to ChatGPT is under discussion, and universities are currently working on it, several interlocutors confirmed to T-invariant.
In April 2023, employees of Moscow State University and Skoltech, speaking at the Data Fusion conference, said that the expected date of its appearance is 2025. Konstantin Vorontsov, head of the Laboratory of Semantic Learning of Artificial Intelligence (now part of the Institute of AI of MSU) , suggested at that time that it should be called ” Our GPT”, and Ivan Oseledets, director of the Centre for Artificial Intelligence Technologies, argued that “teaching AI to languages and cultures of the country’s peoples requires significant investment and resources, including ethnographic expeditions, philologists and philosophers”.
“Until February 2022, the Tikhonova Institute had its own computing power, and they counted a lot of things on their base. Tikhonova already has a whole building of her own at MSU, and to what extent MSU-270 was built under them, it’s hard to say. But she does not need money and resources, “Innopraktika” is supported by almost all big business in Russia”, – says one of the employees of the Research and Development Centre of MSU, who no longer works at the university.
Large language models and other competition with companies like OpenAI or Google are beautiful and spectacular, but, according to Tikhonova herself, in Russia “a more pragmatic approach to the development of artificial intelligence”, “conditioned by the tasks and challenges facing Russia” is “application in UAVs and in the oil and gas industry”. She said this while speaking at the Data Fusion-2024 conference. The five-minute video of Tikhonova’s speech is titled “Multipolarity of the Future Technological World: Russia’s Role”.
“Russian artificial intelligence has its own national specifics, it has its own research on large language models and generative AI, but Russia lags behind in the pace of development from the global one and so far we rather follow trends than create them,” Tikhonova believes. In her opinion, international sanctions are not terrible for universities and institutes: “Russia has not been completely excluded from the global context, scientists continue to publish in top world journals and speak at top world conferences. We have everything: people, personnel,” she emphasised.
In addition to multipolarity, Tikhonova is very concerned about sovereignty. For example, the opening of the Institute of Artificial Intelligence at Moscow State University “with the expert support” of Innopraktika, was “coordinated with the security strategy of the Russian Federation”. Putin’s daughter told the country’s largest businessmen and officials about this at the last Innopraktika anniversary forum, the theme of which was “Aggregate Sovereignty”.
Katerina Tikhonova’s speech
What does the Institute of AI of MSU do and what does it have to do with dual-use technologies?
The full name of this centre is “Institute for Advanced Studies of Artificial Intelligence and Intelligent Systems at Lomonosov Moscow State University”. Judging by the report for 2023, the centre makes quite a lively impression. The works are published in normal journals (including Nature group journals). The topics dealt with by the Institute’s staff are relevant to the current scientific agenda. Among the scientists working with the Institute of AI MSU, it is necessary to mention a well-known specialist – it is neurobiologist Mikhail Lebedev. According to the information on the website, the AI Institute has four laboratories: “Machine Learning and Semantic Analysis”, “Intelligent Video Analysis”,
“Artificial intelligence in bioinformatics and medicine”, “Development of invasive neurointerfaces”. There are four more research groups, which probably have the prospect of becoming full-fledged laboratories.
The first two laboratories are pure digital, the third and fourth are, among other things, “wet” biology and neurobiology. Scientists speak a lot at scientific conferences (mainly in China) and to the general public, publish articles, including popular science articles.
Most of the work in all four laboratories, which is reflected in the report for 2023, is, so to speak, dual-purpose science. Almost any project can be described as working for war, although the results obtained can also be used for civilian purposes.
The most famous project of the Machine Learning and Semantic Analysis Laboratory is the neural network for searching relevant scientific information SciRus-tiny. Academician Alexey Khokhlov characterised this work in his Telegram channel: “…the SciRus-tiny neural network is now applied in a variant when the user can load arbitrary text of any length into the search field. The neural network will then identify those articles from the eLIBRARY database that are closest in subject matter to the entered text”. However, it should be noted that eLIBRARY is not the most complete or authoritative scientific library.
Less is known about another lab project, called the News Collider. The developers write in the report, among other things, the following: “A tool for processing large text arrays “News Collider”. The project develops technologies for semantic markup of texts used in socio-humanitarian research. The main operations include selection of text fragments, their classification, linking and commenting when necessary. Such methods make it possible to analyse large amounts of data (media, social networks, archives) for research on, for example, the impact on public opinion and the identification of potentially dangerous discourses”.
It is well known what discourses are considered “dangerous” by the Russian authorities today. The fact that neural network tools can be used to “scour” social networks and media publications is also well known. The likely use of such a “news collider” is quite imaginable. Most likely, it will be the search for “dangerous discourse”.
The “Intelligent Video Analysis” laboratory is no less important. Judging by the report, an important area of the lab’s work is neural network methods for video compression and high-definition restoration. These methods can be used, for example, when transmitting video over narrow channels, including satellite channels. In times of war, satellite imagery is very important for military intelligence.
The laboratory “Artificial Intelligence in Bioinformatics and Medicine” deals with multiomics, i.e. various areas of genomics, proteomics (protein research), etc. Here, a supercomputer is simply necessary to use such advanced methods as predicting the structure of proteins based on amino acid sequence. This is done, for example, by Google DeepMind’s AlphaFold neural network, but there are other similar solutions. The main areas of the laboratory’s work are pharmacology and diagnostics based on machine learning.
The laboratory “Development of invasive neurointerfaces” deals with the creation of neuroprostheses and various methods of neurorehabilitation. Here they work on the treatment of neuropathic and phantom pains, which often accompany injuries and amputations. War is an epidemic of traumatism.
It should be noted that so far this centre is relatively small. There are hundreds of much more powerful research centres around the world engaged in AI, multi-omics and neurointerfaces research (mostly in the US and China). When Ilon Musk decided to build his supercomputer, he built it in less than a year, and his brainchild, as far as we can judge from open information, is about an order of magnitude more powerful than “MSU-270”. The AI Institute’s computational capabilities – even taking into account the launch of the supercomputer – are rather modest. The coverage of topics is rather limited, even compared to the AI-occupied branches of Sber or Yandex. But it is quite a modern centre.
The scientists at the AI Institute seem to understand how to use the supercomputer. It will not stand idle. Most likely, the majority of applications will be military or dual-purpose, and we will not get open information about many of them in the coming years.
You can support T-invariant’s work by subscribing to our Patreon and choosing a convenient donation amount.
Text: T-invariant Editorial Board
Editors 10.10.2024