Machines will overtake us in five years

Taking the stage in his signature leather jacket on Monday, Nvidia co-founder and CEO Jensen Huang warned the crowd at San Jose’s SAP Center that two hours of math, science and algorithms awaited them. “I hope they realize this is not a concert,” Huang said. “You’ve come to a developer conference.”

Nvidia’s developer conference, called GTC, plays out something like a concert; That’s why the New York Times has named it the “Woodstock of Artificial Intelligence.” An estimated 300,000 people will participate online and in person this week. At its peak, approximately 70,000 people followed the live YouTube broadcast of the conference.

To audience applause, Huang introduced a “very, very big GPU” to work with those systems. According to the company, the “AI superchip” called Blackwell can run generative AI models with trillions of parameters about 25 times faster and consume much less power than its predecessor.

Nvidia’s next generation of artificial intelligence graphics processors, called Blackwell, will cost between $30,000 and $40,000 per unit, CEO Jensen Huang told CNBC’s Jim Cramer on “Squawk on the Street” Tuesday.

“We had to invent new technology to make this possible,” Huang said, showing the Blackwell chip. The CEO estimated that Nvidia spent about $10 billion on research and development costs.

The price suggests that the chip, which will be in high demand for training and deploying AI software like ChatGPIT, will be priced in the same range as its predecessor, the H100, known as Hopper, which The price will be between $25,000 and $40,000 per chip. Chip as per analyst estimates. The Hopper generation, introduced in 2022, marked a significant increase in the price of Nvidia’s AI chips compared to the previous generation.

Continue reading story

“Digital twins” to perform all types of simulations.

Huang also showed off the possibilities of creating “digital twins” – high-quality digital simulations of systems or products – through the company’s Omniverse platform, which can now connect to Apple’s Vision Pro. Automakers use digital twins to simulate driving conditions and test new features. Digital versions of warehouse layouts allow companies to experiment with new configurations or quickly spot potential bottlenecks. For example, the company’s Earth-2 platform is already being used to predict extreme weather patterns.

Developer conferences are an opportunity for tech companies to showcase their new products and capabilities, but they typically don’t generate that much buzz. However, shares of Santa Clara, California-based Nvidia have soared, adding more than $1 trillion in market value since the beginning of the year. It is currently the third most valuable company in the United States after Apple and Microsoft.

Huang founded Nvidia in 1993, and was initially known for creating graphics processing units designed to improve the gaming experience on personal computers. As artificial intelligence evolved over the past decade, GPUs found a new purpose in powering supercomputers that train AI models, and Nvidia gained customers including OpenAI, Microsoft, Google, and Amazon.

Huang told attendees that, in the past, most computing was limited to retrieving and recombining existing content. In the future, Huang said, most content will be generated for the user on the spot by AI that understands context and can provide information “exactly the way you want it.”

Huang also predicted that developers would soon be able to “assemble an AI team” to program software. For now, the company has introduced the Nvidia Inference microservice, which allows companies to access pre-trained models and APIs, accelerating application development.

Huang also showed off the company’s progress in robotics, specifically in creating simulated environments for developing real-world robot software through a platform called Isaac. To cap off the task, Huang ended his keynote speech on stage with two Star Wars droids that Disney had trained using the Nvidia platform.

In five years, humans could be overtaken by AI

But Huang didn’t just talk about products on his program. The company’s CEO then predicted that artificial generative intelligence (AGI) would surpass humans within five years. He made a statement in a question and answer session with members of the press following the annual GTC Developers Conference; However he clarified that everything depends on how the term is defined.

As TechCrunch points out, Huang estimates that AGI is actually difficult to estimate without a specific definition. For his prediction, Huang built on an artificial intelligence capable of performing many of the tasks performed by humans today.

Imagine that the definition of AGI is passing human tests, like “math, reading, reading comprehension, reasoning, medical exams, law exams… you name it, a bunch of tests.” And if AGI is defined as doing well at all of them, “it will be better than any human. Do you think a computer can do that in five years? Then the answer is probably yes.” “

Of course, the Nvidia CEO says that if you only ask questions about “general intelligence” rather than defining the ability to pass a test, the answer is more confusing. In addition, Huang also looks forward to a future generated by AI, in which all the information humans consume will be generated by this technology and every pixel on the screen will be the work of an AGI.

More news that may interest us:

On video | Figure’s humanoid robot is capable of talking and reasoning thanks to AI