I mostly agree but to be short - AI checking AI is always was, is and will be BIASED let's not forget that we have bigger bias in AI than in any other plaace. Bias is key to realizing why AI is not that good.
Finally I have reado numerous papers, scientific ones which not simply claim but prove AGI is impossible due to infinite loop of dilemma within own machine links
Nir, I think youāre spot on. The real edge now comes from smart engineering, not just bigger models.
Live data connections, multi-modal integration, and self-correcting systems might not make flashy headlines, but theyāre what turn AI into something truly reliable.
This was a great read! I really like how you framed the āscaling slowdownā not as a doom-and-gloom moment, but as a turning point for AI.
I think a lot of people outside the AI space still assume progress is mostly about making models bigger, so your breakdown of why thatās hitting limits (data quality, compute costs, and diminishing returns) really helps reframe the conversation. Also agree with you on self-correction. That āinner voiceā idea is fascinating because it feels like one of the most human-like qualities AI could have. And yes, the AGI hype vs reality section was spot on. I think people underestimate how much entirely new thinking and architectures will be needed before we get anywhere close.
Amazing! many important topics were covered, thx. For the road to AGI I came across this tweet from a developer who claims he found a new architecture based on researching artificial imagination, seems interesting, hope you check it out and share your thoughts
āResearchers have also found that having one AI check another AI's work can catch mistakes and improve results. The goal is to give AI an "inner voice" that can notice when something might be wrong.ā ā¦. we have thisā¦ and everything else you suggested, already in our system ā¦ imagine that, a non phd uneducated idiot is ahead of the elitesā¦ you wonāt find the AI you seek in math or science or engineering ā¦ you are looking in the wrong places ā¦ hereās a hint ā¦ itās in art ā¦.
Luckily, the LLM hype is coming to an end. Thank you, GPT-5. That was not expected. š
I don't think it is coming to an end. However, the scaling is less feasible and helpful than it was before.
I mostly agree but to be short - AI checking AI is always was, is and will be BIASED let's not forget that we have bigger bias in AI than in any other plaace. Bias is key to realizing why AI is not that good.
Finally I have reado numerous papers, scientific ones which not simply claim but prove AGI is impossible due to infinite loop of dilemma within own machine links
I would be interested to read those proofs, as I'm a bit skeptical that such a thing can be proved.
Sure.
Here is the very first from Max.
https://philarchive.org/rec/SCHAII-17
Another one from CS
https://cse.buffalo.edu/~rapaport/Papers/lands.pdf
Nir, I think youāre spot on. The real edge now comes from smart engineering, not just bigger models.
Live data connections, multi-modal integration, and self-correcting systems might not make flashy headlines, but theyāre what turn AI into something truly reliable.
ā Xās Tech Brief
weāve officially hit the point where ājust make it biggerā doesnāt cut it.
welcome to the awkward adolescence of aiātoo big for its old tricks, not yet grown into real intelligence.
This was a great read! I really like how you framed the āscaling slowdownā not as a doom-and-gloom moment, but as a turning point for AI.
I think a lot of people outside the AI space still assume progress is mostly about making models bigger, so your breakdown of why thatās hitting limits (data quality, compute costs, and diminishing returns) really helps reframe the conversation. Also agree with you on self-correction. That āinner voiceā idea is fascinating because it feels like one of the most human-like qualities AI could have. And yes, the AGI hype vs reality section was spot on. I think people underestimate how much entirely new thinking and architectures will be needed before we get anywhere close.
great comment, and thanks for writing it. happy you loved the read :)
Amazing! many important topics were covered, thx. For the road to AGI I came across this tweet from a developer who claims he found a new architecture based on researching artificial imagination, seems interesting, hope you check it out and share your thoughts
https://x.com/ryunuck/status/1883032334426873858?s=46
thanks for sharing!
No he hasnt. It's nonsense by a crypto pumper with some development background. Nothing will come from this
āResearchers have also found that having one AI check another AI's work can catch mistakes and improve results. The goal is to give AI an "inner voice" that can notice when something might be wrong.ā ā¦. we have thisā¦ and everything else you suggested, already in our system ā¦ imagine that, a non phd uneducated idiot is ahead of the elitesā¦ you wonāt find the AI you seek in math or science or engineering ā¦ you are looking in the wrong places ā¦ hereās a hint ā¦ itās in art ā¦.
thanks for the comment!