'I trust AI the way a sailor trusts the sea. It can carry you far, or can drown you'

LIVE SCIENCE

Some commenters highlighted what they believe are LLMs' limitations.

"We don’t trust [large language models] LLMs, nor do we think the fundamental technology can ever become an 'AI' to any degree whatsoever," Varian commented. They added that"the fundamental truth is that LLMs have no concept of time, past, present, or future. A request is processed, the process is handled, and further requests can have previous requests force-fed back to it until it runs out of available ram. That’s the [entirety] of it."

Others suggested that LLMS are unlikely to ever be fully intelligent.

"AI is just about automation, the knowledge humans have acquired so far stored in computers is automatically retrieved on our request. I see no intelligence in AI," Jose p.koshy said. "We can trust it as long as our request to AI is unambiguous and pertains to data only. We cannot leave the decision making process to AI."

Read more | LIVE SCIENCE

---