Authors: Katie Paul, Anna Tong and Krystal Hu
NEW YORK/SAN FRANCISCO: Prominent computer scientist Fei-Fei Li is creating a startup that uses human-like processing of visual data to make artificial intelligence (AI) capable of advanced reasoning, six sources say to Reuters, which would be a step forward for technology.
Li, considered a pioneer in the field of artificial intelligence, raised money for the company in its latest round of seed funding. Investors included Andreessen Horowitz, a Silicon Valley venture capital firm, three sources said, and Radical Ventures, a Canadian firm she joined as a research partner last year, according to two other sources.
Spokespeople for Andreessen Horowitz and Radical Ventures declined to comment. Li did not respond to requests for comment.
Li is widely known as the “Godmother of Artificial Intelligence,” a title derived from the nickname “Godfathers” often used to refer to the trio of researchers who won the world of computing’s top prize, the Turing Award, in 2018 for their groundbreaking achievements in technology artificial intelligence.
continued below
In describing the startup, one source pointed to a talk Li gave at the TED conference in Vancouver last month, in which she said the cutting-edge research includes algorithms that convincingly extrapolate what images and text will look like in three-dimensional environments and act on these predictions using a concept called “spatial intelligence.”
To illustrate the idea, she showed a photo of a cat with its paw outstretched, pushing a glass towards the edge of the table. She said that in a split second, the human brain is able to assess “the geometry of that glass, its place in 3D space, its relationship to the table, the cat and everything else,” and then predict what will happen and take action to prevent it. . .
“Nature has created this virtuous cycle of seeing and acting, driven by spatial intelligence,” she said.
Her own lab at Stanford University has tried to teach computers “how to behave in a 3D world,” she added, for example by using a immense language model to produce a robotic arm that could perform tasks such as opening a door and preparing a sandwich in response. to verbal orders.
Li made a name for herself in the field of artificial intelligence for developing a large-scale image dataset called ImageNet, which helped usher in a generation of computer vision technologies that could reliably identify objects for the first time.
He co-directs the Human-Centered AI Institute at Stanford University, which focuses on developing artificial intelligence technologies in ways that “enhance the human condition.” In addition to her academic work, Li led artificial intelligence at Google Cloud from 2017 to 2018, served on Twitter’s board and advised policymakers, including at the White House.
Li lamented the funding gap for artificial intelligence research between the well-resourced private sector on the one hand and government scientists and labs on the other, calling on the U.S. government to adopt a “moonshot mentality” to invest in scientific applications of the technology and risk research.
Her Stanford University profile states that she is on partial leave from early 2024 to slow 2025. Research interests listed in her profile include “cognitively inspired artificial intelligence,” computer vision, and robotic learning.
On LinkedIn, he describes his current job as “newbie” and “something modern”, starting in January 2024.
By jumping into the world of startups, Li joins the race of the hottest AI companies to teach their algorithms common sense to overcome the limitations of current technologies, such as immense language models that tend to spit out meaningless lies in the middle of otherwise dazzling human reactions.
Many argue that the ability to “reason” must be demonstrated before AI models can achieve artificial general intelligence, or AGI, which refers to the threshold at which a system can perform most tasks as well or more efficiently than a human.
Some researchers believe they can improve reasoning by building larger and more sophisticated versions of current models, while others say the way forward involves using modern “world models” that can extract visual information from the physical environment around them to develop logic, replicate how children learn.
Most read on the Internet
