Sundar Pichai, CEO of Google, has defined the present artificial intelligence (AI) as a jagged AI (AJI), which isn’t perfect but explodes. Although the performance is continuously developing to catch up with to artificial intelligence (AGI), it also has problems similar to hallucinations, information errors, and basic mistakes.
Pichai CEO recently appeared in Rex Friedman Podcast and described the present stage of AI as ‘AJI (Artificial Jagged Intelligence)’. It is a symbol of the current AI, where excellent abilities and ridiculous mistakes coexist.
“This term was probably the primary to be written by Andrey Carpash,” he said. Carpash is well generally known as an open AI co -founder and computer vision expert.
In truth, Carpashi said on X (Twitter) last yr, “The state -of -the -art model has made a word called ‘jagged intelligence’ to elucidate the strange characteristics of quite simple problems, similar to solving complex mathematics problems but not understanding that 9.9 is larger than 9.11.”
For instance, large language models (LLMs) are sometimes difficult to make ridiculous decisions or count numbers in easy games. Because of this, “human beings have a linear improvement of information and problem solving, but AI isn’t.”
Pichai CEO also sympathized with this, saying, “The model does an excellent job, and there’s just a few R in ‘Strawberry’ or an easy calculation.”
“It’s now within the age of AJI, and even in dramatic progress, some are still a transitional phase that doesn’t work.”
I used to be cautious in regards to the time of AGI. It was difficult to achieve this completely by 2030. Nonetheless, he emphasized that “even in the event you don’t reach AGI, you’ll have already got a shocking level of progress.”
As well as, 4 core areas of humanity development that AI will bring, ▲ improving the native language -based knowledge access, accelerating science discovery, responding to climate change, and promoting economic growth. This can be the major vision he announced in September 2024 by the UN ‘Summit of the Future’. At the identical time, he stressed that a scientific labeling system is required to obviously distinguish AI production content in the longer term.
As well as, he recently reported just a few key messages to the event of AI.
To begin with, he discussed the phenomenon of ‘token explosion’. Which means that Google’s ‘Geminai’ has increased by about 50 times from 9.7 trillion to 480 trillion in monthly tokens in a yr.
He said, “Each token has experienced the moment of ‘Aha’ through AI,” he said.
He was also convinced that AI would transcend easy technology progress and act as a ‘leap package of civilization’. “AI will provide recent abilities to all 8 billion human beings,” he said. Which means that the opportunity of AI is thought to be a fundamental expansion of humanity, not an extension of existing technology.
Google has already developed an excellent -powered AI model, but it surely also attracted attention by saying that it isn’t commercialized since it is simply too slow and expensive. By way of performance, we now have developed a high -performance AI that exceeds the present models, but it surely is difficult to supply services in point of fact.
It isn’t just emphasis on the positive side of AI. The interior discussion of AI endangered risks is definitely progressing and admitted that “the elemental risk is sort of high.”
By Park Chan, reporter cpark@aitimes.com
