News ChatGPT's ancestor GPT-2 jammed into 1.25GB Excel sheet — LLM runs inside a spreadsheet that you can download from GitHub

[Admin]

Software developer and self-confessed spreadsheet addict Ishan Anand has jammed GPT-2 into Microsoft Excel. More astonishingly, it works – providing insight into how large language models (LLMs) work, and how the underlying Transformer architecture goes about its smart next-token prediction.

ChatGPT's ancestor GPT-2 jammed into 1.25GB Excel sheet — LLM runs inside a spreadsheet that you can download from GitHub : Read more

 

[BX4096]

This video should be a mandatory watch for everyone who rants about the "AI" being "smarter" than the humans, "stealing" data from creators, or being "sentient" (as that Google dummy claimed).

 

[stoppostingcrapnews]

This video should be a mandatory watch for everyone who rants about the "AI" being "smarter" than the humans, "stealing" data from creators, or being "sentient" (as that Google dummy claimed).

Yes, I'm sure we should judge human intelligence based on its ancestors as well like Eutherians. I'm sure you can explain all the emergent properties of LLMs like learning languages it wasn't taught and the ability to pass theory of mind tests that wasn't in the training data.

I'm also sure you know exactly how the human mind works which is why you're so certain technology cannot replicate it. /s

 

[jimbo007]

This video should be a mandatory watch for everyone who rants about the "AI" being "smarter" than the humans, "stealing" data from creators, or being "sentient" (as that Google dummy claimed).

No, it only shows that even an Excel spread sheet can do so many things a fraction of which you may hope to understand.

 

[craigss]

What a load of crap, however we will continue like sheep to blindly allow these mega companies to invade and analyse all our lives and data and pigeon hole us until one day you will wake up not having any control or say in how you live think or work.
Non of this is necessary, none of these AI machines can ever hope to be a tenth as clever or as intuitive as a human brain yet we are all blindly jumping into the rabbit hole because big data companies tell us we need it, create the need then supply it Nvidia did very well

 

[cknobman]

Yes, I'm sure we should judge human intelligence based on its ancestors as well like Eutherians. I'm sure you can explain all the emergent properties of LLMs like learning languages it wasn't taught and the ability to pass theory of mind tests that wasn't in the training data.

I'm also sure you know exactly how the human mind works which is why you're so certain technology cannot replicate it. /s

Or he could be saying that the human mind is so complex that technology is currently, and possibly never, able to replicate it.

AI is just an imitation of intelligence. All its "decisions" are based on what we understand intelligence to be and therefore its output is a direct reflection of who creates it.

 

[NinoPino]

No, it only shows that even an Excel spread sheet can do so many things a fraction of which you may hope to understand.

Is like saying that even a pen can do so many things a fraction of which you may hope to understand.