GlobalLogic Germany’s Post

⚡️ Nvidias neue Open-Source-AI-Modelle Nvidia hat Nemotron-4 340B vorgestellt, eine Familie von Open-Source-Sprachmodellen. Diese Modelle sind darauf ausgelegt, hochwertige synthetische Trainingsdaten zu erzeugen und leistungsstarke AI-Anwendungen in verschiedenen Branchen zu entwickeln. Die drei Modelle Base, Instruct und Reward arbeiten zusammen, um neue, leistungsstarke Sprachmodelle zu trainieren. Instruct erstellt synthetische Trainingsdaten, während Reward die besten Beispiele filtert. Die Nemotron-4-Modelle sind genauso gut oder besser als andere führende Open-Source-Modelle und übertreffen sie in verschiedenen Tests. Nvidia hat außerdem Mamba-2 Hybrid veröffentlicht, das in der Genauigkeit andere transformerbasierte Modelle übertrifft. Mit dieser neuen Modellfamilie bietet Nvidia eine kostenlose, Open-Source-Lösung, die in der Lage ist, qualitativ hochwertige AI-Anwendungen zu entwickeln und weiter zu verbessern. #Nvidia #OpenSource #KünstlicheIntelligenz

In the current WirtschaftsWoche cover story "Chip, Chip hurray", my colleagues Michael Kroker, Jörn Petring, Konrad Fischer and I look at the unprecedented boom in the chip industry due to generative artificial intelligence. "Estimates that the industry will grow to a trillion in annual revenue were dismissed as aggressive not long ago, but now they seem too conservative," believes Intel Corporation CEO Pat Gelsinger. Last year, the industry generated $523 billion in revenue. But will demand really continue at such a dynamic pace, or is this another typical tech bubble? Who is benefiting and what role is Europe playing? It is clear that GPU pioneer NVIDIA is currently the big winner due to the demand for AI accelerators - with fantastic margins. Its G100 cards are said to cost $3200 to produce. Volume buyers are expected to pay around $25,000 per unit.  On March 18, Nvidia CEO Jensen Huang will double down. In his eagerly awaited keynote speech at GTC, he will reveal details about the next product generation - "Blackwell".  It is clear that Nvidia is "in the crosshairs", according to Daniel Newman. No industry likes to be dependent on a dominant supplier," says the CEO of The Futurum Group consultancy. Tech companies rely on their own developments and challengers like AMD. Intel, on the other hand, is also betting on becoming a major contract manufacturer for AI chips, alongside TSMC and Samsung Semiconductor. But Intel still has to convince investors that its manufacturing technology can compete. Is the hype around AI just a bubble typical of the tech industry? Futurum's Newman doesn't think so. "AI is going to change the economy massively," he says. "I think the potential is actually underestimated at the moment."

KI-Innovationen: Nvidia enthüllt nächste Generation an KI-Beschleunigern 🏃 🚀 Nvidia präsentiert mit Blackwell eine bahnbrechende Entwicklung im Bereich der KI-Technologie. Die nach dem Mathematiker David Blackwell benannte Blackwell-Architektur übertrifft ihre Vorgängermodelle in Leistung und Effizienz bei Weitem. Besonders erwähnenswert ist die signifikante Leistungssteigerung: Blackwell trainiert KI-Modelle viermal schneller als die bisherige Grace-Hopper-Generation und benötigt hierfür nur ein Viertel der Energie. 💡 Dieser Fortschritt eröffnet Unternehmen enorme Potenziale. Die Einsatzmöglichkeiten erstrecken sich von der effizienteren Entwicklung umfangreicher KI-Modelle bis hin zu fortschrittlichen Simulationen. Die Technologie ermöglicht es, Produkte und Prozesse zunächst digital zu simulieren, bevor sie in der Realität umgesetzt werden. Dadurch können Innovationszyklen beschleunigt und Kosten signifikant reduziert werden. 📢 Die Blackwell-Plattform ist nicht nur für den Einsatz in Rechenzentren konzipiert, sondern unterstützt auch innovative Anwendungen wie Nvidias 3D-Umgebung Omniverse und Apples Computerbrille Vision Pro. Diese Integrationen erweitern die Einsatzmöglichkeiten von KI in Unternehmen und fördern eine umfassendere digitale Transformation. 📈 Ein weiterer beachtenswerter Aspekt ist Nvidias dominierende Stellung auf dem Markt. Mit Blackwell festigt Nvidia seine Position gegenüber Wettbewerbern wie AMD und Intel und verspricht, den Vorsprung weiter auszubauen. Dieser technologische Vorteil könnte sich langfristig als entscheidender Erfolgsfaktor in der KI-Industrie erweisen. 📥 Teilen Sie Ihre Gedanken zum Thema in den Kommentaren. 👉 Besuchen Sie gerne auch unsere Website: https://www.vaudience.ai/   ❗️ Wir bieten Ihnen eine partnerschaftliche und lösungsorientierte Beratung und zeigen Ihnen auf, wie Ihr Unternehmen von Künstlicher Intelligenz profitieren kann.   🗓️ Vereinbaren Sie noch heute Ihre kostenlose Erstberatung mit uns.   Hier finden Sie weiterführende Artikel zum Thema:  🔗 https://lnkd.in/dtg2BGg9 🔗 https://lnkd.in/deDP3M32 🔗 https://lnkd.in/dzAUEcuX #Nvidia #Blackwell #KuenstlicheIntelligenz #KI #Digitalisierung #Mittelstand #Mainfranken

Handelsblatt Exclusive: 🚨 The CEO of the US chip company AMD, Lisa Su, wants to make her company the leader in #AI. The acquisition of Silo AI, one of the largest AI companies in Europe, should contribute to this. "The acquisition of Silo AI significantly strengthens our AI software capacities," Su told Handelsblatt. On Wednesday afternoon, AMD announced that it had acquired the #Helsinki-based company for 665 Mio$. 👾 AMD is gaining around 300 additional employees, who bring with them important expertise. Unlike #ChatGPT developer OpenAI and most other competitors, Silo AI does not train its language models on chips from NVIDIA, but on those from AMD. The company has developed special software for this purpose. ⚡ AMD is considered Nvidia's most important rival for #AIchips. In the long term, AMD boss Su believes her company is in a better position than Nvidia. "AMD offers many more products that go far beyond pure AI chips." According to her, #Nvidia is a "great" company whose graphics and AI chips are in high demand. However, unlike AMD, Nvidia focuses on just one area: "I firmly believe that our customers don't just need specialty chips for a single solution, but a #broad offering," said Su. 🕹 Nvidia focuses on #GPUs, which are used on a large scale in data centers to train AI models. This has made Nvidia the most valuable company in the world. AMD also offers #CPUs, the brains of PCs, notebooks and servers. CEO Su has great ambitions: "We are still at the very beginning of the AI revolution. My goal is for us to lead this revolution." ==> Link to our story on AMD and Silo AI below. Link to the Lisa Su interview of my colleague Stephan Scheuer in the comments.

🎙️Terra Quantum AG CEO and Founder, Markus Pflitsch, just spoke at the Süddeutsche Zeitung #SZGipfel Economic Summit in Berlin in an engaging conversation with Süddeutsche Zeitung’s deputy editor-in-chief, Ulrich Schäfer. 👉 Focus of discussion was the future of quantum computing and its potential to spearhead a technological revolution within the commercial sector, a development that is undeniably imperative. 🤝 One point that is supporting to trigger the quantum inflection point is Terra Quantum AG’s newly announced technical collaboration with NVIDIA. Markus Pflitsch said: “We are using our hybrid quantum algorithms delivered on NVIDIA GPUs to deliver significant business impact today. This collaboration will not only accelerate this impact but also help unlock a whole new range of opportunities in the areas of machine learning and simulation — while also setting a pathway for quantum computing.” Check out the top topics of #SZGipfel Economic Summit here (German): https://lnkd.in/eRjUScnW and read more about Terra Quantum AG’s collaboration with NVIDIA here (English): https://lnkd.in/eqVSThr7, in NVIDIA’s blog: https://lnkd.in/dFidD4rC and on our website: https://lnkd.in/e8QQyDwR #QuantumIsNow #QuantumComputing #PioneeringTech #TechInnovation #SZGipfel #Nvidia

Was ich heute zum Thema #Technologie und #Innovation lese. Ein Beitrag von #: Künstliche Intelligenz: Drei Technologie-Schwergewichte fordern Nvidia heraus - klamm.de Bestimmt relevant für einige Kollegen bei #Schattdecor und #MyMineralMix !

🚀 NVIDIA pusht AGI: Eine neue KI-Ära beginnt 🚀 Im Herzen des Silicon Valleys führt NVIDIA, unter CEO Jensen Huang, die KI in eine aufregende Zukunft: Die Entwicklung Künstlicher Allgemeiner Intelligenz (AGI). Mit dem Ziel, Maschinen zu erschaffen, die menschliche Aufgaben ebenbürtig ausführen können, steht NVIDIA an der Spitze einer Revolution, die das Potenzial hat, alles zu verändern. Dieser mutige Schritt in Richtung AGI könnte nicht nur die Art und Weise, wie wir arbeiten und leben, neu definieren, sondern auch die Grenzen des Möglichen erweitern. NVIDIA's Vorreiterrolle zeigt, dass wir am Beginn einer Ära stehen, in der Technologie unsere kühnsten Träume übertrifft. #NVIDIA #AGI #KIRevolution #SiliconValley #ZukunftDerTechnologie https://lnkd.in/dvp5z4YA

Stories of world-famous companies almost being swallowed up by a larger competitor and thus almost never becoming famous are part of Silicon Valley folklore. The most famous example is Google, which was almost sold to its competitor Excite for 750,000 dollars in 1999. Or Yahoo, which wanted to take over Facebook for almost one billion dollars. And of course there is also an anecdote about NVIDIA, currently one of the most highly valued companies in the world thanks to its AI chips. In 2006, chip manufacturer AMD wanted to enter the graphics card industry and considered acquiring Nvidia. The deal fell through because Nvidia boss Jensen Huang and the then AMD CEO Hector Ruiz were unable to reach an agreement. AMD bought ATI instead. Google, Meta and Nvidia are now among the most influential companies in the tech industry. Excite, on the other hand, no longer exists. Yahoo - now a mere shadow of its former self - is owned by the investment company Apollo Management. But AMD, which almost went bankrupt due to the high debt burden from the ATI purchase, has really blossomed under its current boss Lisa Su. What's more, AMD, which, like Nvidia, has its headquarters in Santa Clara in Silicon Valley, has become Nvidia's most important competitor. AMD is the number 2 in the market for AI accelerators, ahead of Intel Corporation with its Gaudi series and Google with Tensor, that is mainly used internally. The gap to Nvidia is, of course, huge. Nvidia's sales of AI accelerators are estimated at 105 billion dollars this year, while AMD will probably reach five billion dollars. Now AMD boss Lisa Su is making a big purchase. The expertise to expand the market position is to come from ZT Systems, or more precisely from its engineers. The strategy behind these and other acquisitions - and why the ATI deal should be a warning.

1/ Nvidia setzt auf eine Mischung aus menschlichen und synthetischen Daten, um das Problem fehlender Trainingsdaten für humanoide Roboter zu lösen. Mit einem Apple Vision Pro steuern Menschen die Roboter und zeichnen Beispielaktionen auf. 2/ Das Simulationsframework RoboCasa multipliziert die aufgezeichneten Daten um den Faktor 1000 oder mehr. MimicGen generiert aus den Originaldaten zusätzliche Handlungsabläufe und filtert Fehler heraus. So können teure menschliche Daten durch GPU-beschleunigte Simulationen ersetzt werden. 3/ Nvidia-CEO Jensen Huang erläutert auch das "Drei-Computer-Problem" in der Robotikentwicklung: Ein Computer für die KI-Erstellung, einer für die Simulation und einer für die Ausführung. Open-Source-Tools wie RoboCasa und MimicGen sollen das Robotik-Ökosystem im Einklang mit Nvidias Soft- und Hardware wachsen lassen.

On Monday I attended the opening keynote of NVIDIA's GTC developer conference in San Jose, which Bank of America analysts have dubbed the Woodstock of the AI industry. In fact, the lines around the SAP Center and the enthusiasm of the attendees reminded me of past Apple product launches when founder Steve Jobs would present and wow his audience. Although Nvidia's products are tough fare. At the keynote, according to analyst Patrick Moorhead, Nvidia CEO Jensen Huang had to prove that he was still ahead of the competition. Indeed, Huang delivered. His latest-generation AI accelerator, Blackwell, is said to be up to 30 times faster at inference - applying AI models - than its predecessor, Hopper, which is considered the gold standard. And Huang left no doubt that he no longer sees Nvidia as a hardware supplier, but as a platform provider. In other words, on a par with Apple and Microsoft. In fact, Nvidia is closing the gap on Apple, with only $450 billion separating them in market value. And Apple seems to be so far behind in gen AI that it even has to use technology from its competitor Google. As much as I don't begrudge Huang his success-he's been at it for over thirty years-the growing power of Nvidia is somewhat frightening. Today, Nvidia already plays a kind of referee in the market, and thanks to the bottlenecks in demand for its products, it can participate in interesting start-ups. With the Nvidia Inference Microservices (NIM) presented at GTC, the company could extend its power even further. Nvidia may be able to be the gatekeeper which AI models are part of these services and therefore successful in the market - and which are not.