SAN FRANCISCO, United States— No doubt, the artificial intelligence (AI)-related announcements by Nvidia during the CES 2024 in Las Vegas stood out as the most consequential for the industry. Nvidia is reported to have unveiled a generative AI suite during the keynote address, described AI models that Nvidia has unveiled previously, and claimed these to be AI systems capable of producing text, images, or any other media content. This growth saw Nvidia becoming the world’s largest company, and the most important factor had been its advancements in technology. As expected, the event NVIDIA had been waiting for did assist with returning optimism, though I yearned for fear and expectation meeting.
Returning to Anticipated Scale
Attendees and deck analysts expected the return of “deal frenzy” with halls crowded at CES, which started on Jan 5. Avi Greengart of Techspotential analyzed the trade show’s difference from past years, stating, “Even with AXIOM battling for area-wide meetings, we expect large crowds. In 2022, it was a stripped-down version of rush hour, empty rooms, and no plugins to fancy hotel rooms.” He explained that later he expects disabled access and many behind-closed-doors meetings, explaining, “which is what trade shows are all about.” From earlier Tuesday, competing corporations took their akafosgrs. What was actually a vast square of land in excess of 18 acres, pavilions, halls, as well ass, not only in London but also on-site for participants, proved reality.
AI Models Applicable for Business and Robotics Domains
Jensen Huang, the co-founder and the chief expo of Nvidia, launched the AI models in his keynote speech that had opened CES a week ago“”g, the co-founder and the chief expo of“ Nvidi”a, launched the AI models in his keynote speech that had opened CES a week ago. These innovations are expected to automatically regulate the business processes of companies through building their own AI agents. In addition, they also aim to assist manufacturers of robots and self-driving cars for their system training by offering basic tools necessary in building AI-powered technologies.
Cosmos: World Foundation Model
One of the released models is called “Cosmos,” which Huang deemed as the first blow to each country’s foundational AI model. Huang noted its intended task as “It’s teaching, not creating content. It’s teaching the AI to consider the physical world,” and echoed that he plans to make the model available for download. Huang made an analogous statement by explaining the nature of his hope for Cosmos by saying, “We really hope Cosmos will do for the world of industrial AI what Llama 3 has done for enterprise AI,” which means the expectation is based on the reality that Meta released Llama 3 for free in April last year and it triggered the adoption of generative AI in businesses.
Llama Nemotron Suite
The models were also made as Llama models for enterprise purposes, converting them into the Llama Nemotron suite. Huang noted that Llama 3 had approximately 650,000 downloads from Meta and was respun into other models 60,000 times. He described Llama 3.1 as a “complete phenomenon” and “singularly the reason why just about every enterprise and every industry has been activated to start working on AI.” Llama Nemotron would come in three sizes: nano, super, and ultra, with differing power and speed demands. They were designed specifically for “agentic” AI applications where multiple intelligence models could work independently to proffer solutions to intricate problems and automate monotonous functions. Like Cosmos, Huang stated the Llama Nemotron models would also be distributed freely.
Pushing the Need for AI Hardware
Nvidia’s core business is as a hardware manufacturer; they began as a producer of gaming and graphics GPUs. The company found itself at the core of the AI revolution because model training and inference were accelerated to an astonishing degree by the same ‘massively parallel’ processing done in GPUs. Constructing AI models and freely distributing them makes sense for Nvidia because it increases the demand for the company’s physical goods. Nvidia’s GPUs have purportedly been in hot demand because of the crypto mining frenzy as well as the AI boom, which has led some gamers to voice complaints of being shut out of the market.
The Intersection of Gaming and AI
Nvidia’s transformation from a gaming to an AI business this week occurred almost full circle at CES. Huang announced gaming-oriented Nvidia GPUs (RTX 50 series) that use AI inference, instead of traditional graphics computation, to render pixels. These new GPUs purportedly possess the same performance as some of Nvidia’s top-of-the-line GPUs but retail for under half their pricing. The reason is that the most expensive computational work is done on Nvidia’s graphic models AI servers and the cards used during the computation are inexpensive. Huang said out of 35 million pixels in a video, 33 million would be inferred by models (94% AI, 6% traditional); 6% is required as input. This approximate 94% ratio is a claim that Nvidia’s market cap is roughly this figure. Shares skyrocketed tenfold this decade (US$14 to US$140) when the GPU demand for training models at Nvidia soared, like with ChatGPT.
Other Themes from CES
Creative Strategies analyst Carolina Milanesi mentioned, “…the way we are going to be connected will be a topic to create great buzz at CES” together with metaverse propulsion (VR goggles, mixed reality), Web 3 (NFTs, blockchain, the decentralized internet), and potentially crypto (although Milanesi noted FTX/SBF could be a dampener). Pandemic effects/health tech (tech for remote health care) and environment (gadgets for rubbish, energy apps, sustainable materials) also fall under other themes. Greengart focused on the “environment” theme, saying, “If you’re the type of person who is off the grid and growing vegetables, then CES probably doesn’t cater to you,” adding, “But, I do commend companies that figure out how to make their products and supply chain more sustainable.”
