Get ready for AGI Boyzzz 😘😘😘😘

[හෙළයෙක්]

calculators can do one specific task very well, this is "General", that's the difference.

LLM තාම තියෙන්නේ week AI/narrow AI කියන category එකේ. LLM general tasks කරන් නෑ. Language task විතරයි කරන්නේ.

 

[dayt0na]

Again missing the G in AGI. Fact that Anthropic had to pick a specific domain is a strong sign. Besides if you think Claude is self aware then i don't know what to tell you.

I have clearly mentioned why they chose that path.

Why does it have to be self aware?

AGI = a system that exhibits all the cognitive capabilities the brain has.

cognitive capability = your brain's mental capacity and processing skills used to think, learn, remember, reason, and solve problems

 

[Edward Kenway]

I have clearly mentioned why they chose that path.

Why does it have to be self aware?

AGI = a system that exhibits all the cognitive capabilities the brain has.

cognitive capability = your brain's mental capacity and processing skills used to think, learn, remember, reason, and solve problems

"My beloved LLM is almost AGI but it decided not to do AGI" isn't an argument. Regardless of the reason Claude is not general by any measurement therefore not AGI.

Did you really say AGI must exhibit cognitive capabilities of human brain and then went on to ask why AGI must be self aware? Are you saying that your brain is not capable of self awareness?

Explains a lot actually

 

[dayt0na]

Did you really say AGI must exhibit cognitive capabilities of human brain and then went on to ask why AGI must be self aware? Are you saying that your brain is not capable of self awareness?

Explains a lot actually

A system that can think, reason and solve problems without being self aware and no self interest is the IDEAL AGI.

Like why the fuck would I want Claude Code to be self aware as long as it does the above?

 

[හෙළයෙක්]

A system that can think, reason and solve problems without being self aware and no self interest is the IDEAL AGI.

Like why the fuck would I want Claude Code to be self aware as long as it does the above?

Backward reasoning ගැන හිතන්නවත් බෑ mL එක්ක. ඔය forward reasoning විතරයි.

 

[Edward Kenway]

A system that can think, reason and solve problems without being self aware and no self interest is the IDEAL AGI.

Like why the fuck would I want Claude Code to be self aware as long as it does the above?

Okay so you don't understand what self awareness mean in this context either?

Being self aware does not mean Claude is going to be in love with you. It means Claude understands why it's doing something.

 

[dayt0na]

Okay so you don't understand what self awareness mean in this context either?

Being self aware does not mean Claude is going to be in love with you. It means Claude understands why it's doing something.

define "understand"
you need to be self aware to "understand" something?

 

[Edward Kenway]

define "understand"
you need to be self aware to "understand" something?

Yes, token predictor does not "understand" anything.

All it can do is picking the next token based on probability math.

 

[dayt0na]

Yes, token predictor does not "understand" anything.

All it can do is picking the next token based on probability math.

that's like saying a human only produces speech through neurons and electrical signals.

 

[Edward Kenway]

that's like saying a human only produces speech through neurons and electrical signals.

No that would be comparing how AI is thinking to the way semiconductors working.

Question is not the underlying hardware.

 

[dayt0na]

No that would be comparing how AI is thinking to the way semiconductors working.

Question is not the underlying hardware.

Question should be about the OUTPUT. Not the hardware or "mechanism"

 

[Edward Kenway]

Question should be about the OUTPUT. Not the hardware or "mechanism"

Yeah you don't understand AGI.

You think Claude is almost AGI because it "feels" that way.

 

[dayt0na]

Yeah you don't understand AGI.

You think Claude is almost AGI because it "feels" that way.

It "feels" that way because of the OUTPUT

 

[Edward Kenway]

It "feels" that way because of the OUTPUT

No it feels that way because it's a token predictor masquerading as AGI

 

[dayt0na]

No it feels that way because it's a token predictor masquerading as AGI

Are you saying that you feel Claude is AGI because it's a token predictor?

 

[sacrox]

තරහ නැත්තම් ගෙදරදීම AGI හදාගන්න හැටි කියන්නද?

ado

 

[Edward Kenway]

Are you saying that you feel Claude is AGI because it's a token predictor?

I'm saying that whatever you "feel" is not the benchmark.

Either way even you can't deny that Claude isn't "general".

ado

අඩෝ

 

[dayt0na]

I'm saying that whatever you "feel" is not the benchmark.

Either way even you can't deny that Claude isn't "general".



අඩෝ

there are other benchmarks (that keep getting saturated with newer model releases) and subject matter experts to do that.
check out ARC AGI

My response was a question posed by you on why I think AGI is near.
My judgement is based on the models output on a domain I am familiar with and their improvement over the years.

 

[Edward Kenway]

there are other benchmarks (that keep getting saturated with newer model releases) and subject matter experts to do that.
check out ARC AGI

My response was a question posed by you on why I think AGI is near.
My judgement is based on the models output on a domain I am familiar with and their improvement over the years.

Whole point of AGI is not being limited to a domain and you don't seems to get it.

 

[dayt0na]

Whole point of AGI is not being limited to a domain and you don't seems to get it.

That’s my limitation not the models limitation. I can only speak for a specific domain and other people in other domains do speak on how the models have improved over time.

Definition of AGI was clearly stated.