NVIDIA создала ИИ-систему для поиска внеземных сигналов, работающую в 600 раз быстрее

GUEST OPINION: Since 2015, NVIDIA's graphics processing units have been at the centre of Breakthrough Listen's computing infrastructure. The scale of data that radio astronomy generates requires the kind of parallel processing that GPUs were designed for, and Breakthrough Listen was scanning at a volume that had no precedent in the history of the search for extraterrestrial intelligence. In November 2025, that partnership produced something new: an AI detection system capable of processing Fast Radio Burst signals at 600 times the speed of previous methods. Six hundred times faster means the system can evaluate in seconds what previously took minutes, which matters when the signals being searched for can appear and disappear within milliseconds. To understand why the speed improvement is significant, it helps to start with what Breakthrough Listen is actually searching for — and why detecting it has always been a data processing problem as much as an astronomy problem. What Breakthrough Listen Is Actually Hunting The sky produces radio emissions constantly — from pulsars, from magnetars, from the cosmic microwave background, from satellites and terrestrial interference. The challenge is not detecting radio waves. It is identifying narrow-band transmissions that natural astrophysical processes cannot produce. Natural sources of radio emission are broadband. They spread energy across a wide range of frequencies. A narrow-band signal, concentrated at a specific frequency with no known natural explanation, is what SETI researchers call a technosignature: evidence of technology, and by extension, of something that built the technology. The program scans more than 800 million radio channels simultaneously, searching for signals narrow enough, structured enough, and directionally consistent enough to qualify as plausibly artificial. Fast Radio Bursts occupy a different but related category. These are intense millisecond-duration pulses of radio energy — almost certainly astrophysical in origin for the most part, produced by neutron stars or magnetars under extreme conditions. They arrive from cosmological distances, appear without warning, and are bright enough to detect across billions of light-years. Monitoring them is scientifically valuable on its own terms, and it also stress-tests the detection infrastructure: the same AI system trained to identify anomalous narrow-band signals needs to reliably distinguish between natural FRBs, human interference, and anything that doesn't fit either category. The NVIDIA collaboration addressed that specific bottleneck — not the question of what's out there, but the question of whether the machinery can evaluate candidates fast enough to catch them. The Westerbork Shift In May 2025, Breakthrough Listen announced a separate and architecturally significant expansion: a partnership with ASTRON, the Netherlands Institute for Radio Astronomy, and the University of Manchester to deploy an all-sky monitor at the Westerbork Observatory in the Netherlands. Previous radio SETI programs, including Breakthrough Listen's early work at the Parkes Radio Telescope in Australia and the Green Bank Telescope in West Virginia, operated by pointing a dish at a specific target — a star, a region of the galaxy, a candidate source — and listening. Pointed observation is precise and deep, but it is also blind to everything outside its field of view. A signal appearing from an unexpected direction, or a transient event at a location the telescope wasn't currently aimed at, would go undetected. An all-sky monitor changes the geometry of the search entirely. Rather than pointing, it watches the full sky continuously. Transient events — signals that appear briefly and without advance notice — can be caught wherever they appear, not only where a telescope happened to be aimed. The Westerbork deployment is the first all-sky component in the Breakthrough Listen network, and it represents a meaningful expansion of the program's detection coverage. The partnership also carries a European dimension. Yuri Milner designed Breakthrough Listen from the outset as an international, open-data program — one where observational data is made publicly available to researchers worldwide rather than held within a single institution. The University of Manchester's involvement in the Westerbork deployment extends the network into the UK research community, and ASTRON's participation brings one of Europe's most capable radio astronomy institutions into the collaboration. Open Data as Architecture Breakthrough Listen's commitment to publishing its data publicly was built into the program from the beginning, and the reasoning is practical as well as philosophical. Ruling out false positives — signals that appear artificial but turn out to be interference, equipment artifacts, or known astrophysical phenomena — requires broad scrutiny. More researchers with access to raw data means more opportunities to identify mundane explanations before a candidate signal gets overstated. The policy also reflects a specific argument that Milner has made in the Eureka Manifesto about the nature of the questions Breakthrough Listen is pursuing. The search for extraterrestrial intelligence is not a proprietary investigation. The questions it is asking — whether technological civilisations exist elsewhere in the universe, whether any of them are detectable at interstellar distances — belong to humanity as a whole. Making the data accessible is a structural expression of that position. A genuine discovery, if one comes, would not be the property of any single laboratory or nation. The program's network now spans multiple facilities. Parkes in Australia has monitored frequencies from 700 MHz to 4 GHz, covering more than 800 million radio channels simultaneously. The Sardinia Radio Telescope in Italy joined the network in 2024. Westerbork follows in 2025. The NVIDIA AI detection system runs across the data pipeline, processing candidates at speeds that were not achievable through conventional methods. The Fermi Paradox and What Infrastructure Answers Milner and Stephen Hawking launched Breakthrough Listen in 2015 with a direct and deliberately simple question: Are we alone? A decade of the most systematic search ever conducted has produced no confirmed technosignatures. That result is not a failure — it is data, and it has value. The Fermi Paradox, which Milner addresses directly in the Eureka Manifesto, describes the tension between the apparent scale of the universe — hundreds of billions of galaxies, each containing hundreds of billions of stars, many with planets — and the absence of any confirmed contact with other technological civilisations. The silence could mean many things. Detection methods may not yet be sensitive enough. Search coverage may not be broad enough. The window in which a civilisation produces detectable signals may be shorter than the window in which another civilisation could receive them. The universe may be genuinely empty of technological life at the scale Breakthrough Listen can currently probe. What the NVIDIA partnership, the Westerbork all-sky monitor, and the open data network represent is the ongoing effort to make the question answerable rather than merely debatable. Every improvement in detection speed, every extension of sky coverage, every new facility added to the network narrows the range of possibilities. The search has not found anything. It has eliminated a growing volume of the sky, across a growing range of frequencies, at a growing level of sensitivity. That elimination is how science moves — not from silence to signal in a single moment, but from less information to more, at the pace that the available infrastructure allows. The Breakthrough Listen infrastructure, taken together, is the most serious attempt ever made to hear something, if something is there to hear. Whether or not it succeeds on those terms, it has already changed what the search looks like and what it is capable of asking. The Scale of the Commitment Breakthrough Listen launched with a ten-year, $100 million commitment from Yuri Milner — the largest private investment in SETI research in history. That funding bought telescope time at major radio observatories, computing infrastructure capable of processing the resulting data volumes, a scientific staff to analyse the output, and an open-source data pipeline that has allowed researchers outside the program to contribute to the analysis. The November 2025 NVIDIA collaboration and the May 2025 Westerbork deployment are extensions of that original commitment, reflecting a program that has matured rather than stalled. The first years of Breakthrough Listen were spent establishing observational baselines and refining what a genuine anomaly would look like in the data. The current period is characterised by expanded coverage, improved detection algorithms, and a network architecture that is genuinely international rather than centred on a small number of flagship facilities. The broader Breakthrough Initiatives portfolio — which includes Breakthrough Listen alongside programs in life detection and physics — reflects a consistent view that the most important scientific questions are the ones that private capital has historically been least willing to fund. Governments fund science when the applications are visible. Markets fund science when the returns are near. The questions Breakthrough Listen is asking operate on neither of those timescales. The answer, if it comes, will arrive on the universe's schedule — and the infrastructure to receive it has to be built and maintained in advance of knowing when that might be.