Imho, many LLMs are facing with hallucinations and it is indeed an unsolved problem that hinders LLM capabilities towards AGI

Just curious about you guys thought on
- Will we solve hallucination by 5 years span?
- Do you think hallucinations is caused by AR-LLM design?

Cooking AI at Eidy (eidy.cloud) /running, /poetry /mathrock

/receipts minter
5k - 20:14 
10k - 41:24 
21k - 1:38:02 
42k - 3:28:34

My take is that 
- this problem will persist for more than 5 years

On the second take, I have two opinions
- AR-LLM can mitigate hallucination my sophisticated decoding method (meta decoding, etc.). I assume that hallucination is caused by non-optimal decoded sequence
- AR-LLM is flawed as Yann said