{"id":1855,"date":"2026-08-03T03:19:08","date_gmt":"2026-08-03T03:19:08","guid":{"rendered":"https:\/\/futureknowledge.in\/?p=1855"},"modified":"2026-08-03T03:19:08","modified_gmt":"2026-08-03T03:19:08","slug":"qa-nvidia-genai-chief-explains-why-open-models-matter-in-ai","status":"publish","type":"post","link":"https:\/\/futureknowledge.in\/?p=1855","title":{"rendered":"Q&A: Nvidia genAI chief explains why open models matter in AI"},"content":{"rendered":"
\n
\n
\n
\n
\n
<\/div>\n<\/section>\n

When Nvidia CEO Jensen Huang speaks, the tech industry listens. He used his first-ever post on X<\/a> last week to argue that open AI models<\/a> \u201cstrengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.\u201d<\/p>\n

Nvidia is a chip company (with a\u00a0focus on GPUs). But it also has its own AI models that include Nemotron, an open-weight model that\u2019s free to download and modify. The company is also pushing for open AI technologies and security through the Nemotron Coalition<\/a>\u00a0 and the newly launched Open Secure AI Alliance<\/a>.\u00a0<\/p>\n

With Huang\u2019s recent comments in mind, Computerworld<\/em> sat down with Kari Briski, Nvidia\u2019s vice president of generative AI (genAI) software, to find out more about open models and why they matter for enterprise and sovereign applications.<\/p>\n

\n
\"Kari
\n

Kari Briski, vice president of generative AI software at Nvidia.<\/p>\n<\/figcaption><\/figure>\n

Nvidia<\/p>\n<\/div>\n

What do open models really give enterprises and countries? <\/strong>\u201cOpen models allow enterprises and countries to own, inspect, and adapt models with their own data. The learning rate of adoption differs by region and enterprise, so there are different pockets of acceleration, but all the models have made tremendous progress closing the gap. What\u2019s interesting about Nemotron is that we also publish our data, and that opened up a new world of engagement.\u00a0<\/p>\n

\u201cEnterprises reached out, saying, \u2018Thank you, but I want to understand why you put this set of data out.\u2019 That got them in the mindset that they could curate, create, capture, and control their own data.\u201d<\/p>\n

When you develop these open models, do you have CIOs in mind, or sovereign uses? Where does the rubber hit the road? \u201c<\/strong>Both. At the highest level, it\u2019s very much the same: use cases, data sets, outcomes you want to achieve. It used to be question and answer, now it\u2019s agentic workloads, getting stuff done, calling tools. To get a task done, which tool do you call? That\u2019s different for every enterprise and every local region, which is why we match local models to local ecosystems, with post-training on those local models.\u00a0<\/p>\n

\u201cOne enterprise can have 2,000 tools; a local region, 2,000 regional tools. [They\u2019re] fundamentally the same, but with different go-to-market motions: reinforcement learning environments, synthetic data generation, or anonymizing and differential privacy for healthcare and banking data.\u201d<\/p>\n

Why does Nvidia put its models out in the open? What do you get out of it? <\/strong>\u201cWe\u2019re building Nemotron first and foremost for ourselves, to understand our systems running at scale \u2014 not just for training but also for inference, which allows us to iterate on things like model architecture for token efficiency. We believe in a thriving ecosystem. When you put a model out into the open, you get more startups, more builders, lower entry barriers.\u201d<\/p>\n

What\u2019s the connection between open models and sovereign AI? \u201c<\/strong>To be very honest, it\u2019s about bootstrapping a region that otherwise didn\u2019t have the compute to get to a base-level model. Centers of excellence historically lived in higher education or research, and in certain pockets of the world, getting onto a compute cluster was grant-based: get on, get off, then find another grant. That start-stop, not being able to constantly iterate, is difficult. It\u2019s the scaling laws of AI: the more compute, the more intelligence; the more access, the faster you can drive them. Open models and open data are that bootstrap. You don\u2019t have to recreate capturing the knowledge of the internet as a pre-training model.\u201d<\/p>\n

Regionally, things are different. Voice and word of mouth are big in South Asia, chatbots less so. What\u2019s the approach going forward? Is it an SLM approach? <\/strong>\u201cYou\u2019re going to hate my answer: it depends. Language is ever evolving. Go back to first principles: AI is infrastructure. Not just the model, but the harness, the skills, the runtime. All of this is a new computing platform, and when we deploy it, it\u2019s how can it understand and adapt. We\u2019re at the tip of the iceberg integrating AI into everyday applications. The more it can understand and update even dialects, the better.\u00a0<\/p>\n

\u201cUnderstanding those niche areas is what drives the data flywheel of deploying AI. It won\u2019t be overnight. That\u2019s why this is a new industrial revolution. We have to lay the infrastructure everywhere for these models to update. And to your point, it could be a large teacher model that at the edge is an SLM. It depends.\u201d<\/p>\n

AI companies in Nepal told me, \u201cWe can\u2019t innovate because we don\u2019t have access to a GPU.\u201d They want a model that runs on the hardware they have. Is that the way you look at it? <\/strong>\u201cIf you know Nvidia, it\u2019s all about the ecosystem, and we want to enable developers. This is why we have many different sizes: Nano, Super, and Ultra of the Nemotron family, not just for where you deploy, but for developers to iterate on smaller GPUs, then scale to a more robust model like Ultra.<\/p>\n

\u201cWe run as a model-as-a-service across the cloud providers. Day zero, they all had it ready to go, along with our inference partners, optimized and efficient for their workload. We don\u2019t pitch the checkpoint over the wall; we help them take it that last mile, and partners get early access, so on release day it\u2019s available on whatever platforms they use.\u201d<\/p>\n

The industry works together on open technology like Linux. Can you partner on models, and is the focus on performance or quality? \u201c<\/strong>On performance versus quality, it\u2019s both. You can\u2019t have a high-quality model that is slow or heavy, and you can\u2019t have a fast model that sucks. We\u2019re going after three dimensions: efficient, state-of-the-art, and open. And models do work together today \u2014 a planning agent routes a question to the best model to complete the task.<\/p>\n

\u201cOn partnering, that\u2019s our goal in true open source and in the Nemotron Coalition, bringing the best and brightest minds with the commitment to open source and collaboration. Members contribute in different ways: pre-training new architectures, post-training RL environments, contributing data. The goal is everybody working together on one model. That\u2019s why the coalition matters for model architecture \u2014\u00a0 token efficiency, latent mixture of experts, changing how we route it. These are new architectures we\u2019re thinking about. We need these ideas coming in from others.\u201d<\/p>\n

Is the latest open-source model always the best? In developing countries, some go back to older models they\u2019ve tested enough to predict the responses. \u201c<\/strong>Models aside, that\u2019s true for any software. You do an upgrade and it\u2019s just not working the way it worked before. The results aren\u2019t better. When you build a system around it with certain prompts, a model might not react the way it used to. This is why we built the coalition, why we work with a very close set of partners. We pull their evaluation benchmarks in-house to make sure we\u2019re not regressing, only improving.\u00a0<\/p>\n

\u201cYou have to switch your mind with AI and the way you test it. Is the latest model necessarily the best? I\u2019m going to say yes. Companies and countries need the skills to quickly evaluate, update prompts, and adopt new models.\u201d<\/p>\n

Can I fork an Nvidia model, put it on Hugging Face, and do what I want with it? Do you take lessons from the forks and benchmarks? \u201c<\/strong>We have a very open license. We put out reduced precision NVFP4 checkpoints. Those are the most popular, especially with Ultra, because people want the smallest footprint to run that robust model. Even with mature models, there\u2019s all kinds of quantization happening and getting posted back, and I love that community engagement. I love seeing different forks of our models. We track those, too, which gives us an idea of what matters to people. That\u2019s the purpose of putting it out in the open: to see how they change it, how they need to adapt it, then put it back out for the rest of the world to enjoy.<\/p>\n

\u201cBenchmarks are table stakes, not the ceiling of where we need to go, so we\u2019re always looking for new benchmarks and new workloads. In the last 90 days, the style of workload changed dramatically, going from question-answer pairs to agentic workloads, and those workloads mattered to make sure we were tracing our model, which led to us being able to fully trace it. We want to show that you can be just as intelligent in a short amount of time, compute efficient, token efficient.\u201d<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

Source: Read the original article on www.computerworld.com<\/a><\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

When Nvidia CEO Jensen Huang speaks, the tech industry listens. He used his first-ever post on X last week to argue that open AI models \u201cstrengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.\u201d Nvidia is a chip company (with a\u00a0focus on GPUs). But it also has its own AI models that include […]<\/p>\n","protected":false},"author":1,"featured_media":1856,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[36,3],"tags":[18,29,33],"class_list":["post-1855","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-share-suggestions","category-technology","tag-impact-amzn","tag-signal-avoid","tag-stage-stage-4"],"_links":{"self":[{"href":"https:\/\/futureknowledge.in\/index.php?rest_route=\/wp\/v2\/posts\/1855","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/futureknowledge.in\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/futureknowledge.in\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/futureknowledge.in\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/futureknowledge.in\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1855"}],"version-history":[{"count":0,"href":"https:\/\/futureknowledge.in\/index.php?rest_route=\/wp\/v2\/posts\/1855\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/futureknowledge.in\/index.php?rest_route=\/wp\/v2\/media\/1856"}],"wp:attachment":[{"href":"https:\/\/futureknowledge.in\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1855"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/futureknowledge.in\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1855"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/futureknowledge.in\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1855"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}