2/16, 8:19 AM] Jaspreet Bindra : If humans had not 'hallucinated', would we have some of our most important discoveries and creations?
German chemist Friedrich Kekulé discovered the molecular structure of benzene when he was ‘hallucinating’ about a snake biting its own tail. Mendeleev reportedly had a ‘vision’ of the periodic table, Edison claimed to mine his dreams for material. Writer Stephen King claimed to have dreamt up his novel "Misery" during a somnolent transatlantic flight, and the masterpieces of Van Gogh and Salvodar Dali were often inspired by the hallucinations inside their heads.
So, are we looking at LLMs (Large Language Models) hallucinating the wrong way?
What if we thought of them as a feature, and not a bug, and take its help to make astonishing discoveries like the aforementioned gentlemen?
I explore this very counterintuitive view in my Tech Whispers column in Mint today: https://www.livemint.com/
[2/21, 8:56 AM] Rakesh Biswas: Cited @~Jaspreet Bindra in a social media DJ driven UDLCO happening today in the AI healthcare India group :
[2/21, 6:49 AM] Rakesh Biswas: Yes in this case our hypothesis is that this could be non AI generated content masquareding as AI
[2/21, 8:16 AM] Vijayasimha Ajarananda: Ha, ha. Hallucinating human's beware, your digital twin will be ASH karoing (Artificially synthesised hallucination).
[2/21, 8:23 AM] Rakesh Biswas: Our biological human networks are sufficient for positive hallucinations reinforcing creativity.
There's an interesting piece written and shared by Jaspreet Bindra about how AI LLM hallucinations are a booster to their creative potential likening it to Kekule's discovery of the benzene ring which was a hallucinationary dream (aren't all dreams that?)!
Salvador Dali was known to encash on them for his biological LLM products that sold for zillions!
[2/21, 8:24 AM] Vijayasimha Ajarananda: What most of us refer to as stupidity,?
[2/21, 8:28 AM] Rakesh Biswas: Absolutely!
Learning to ask stupid questions is the first step to creativity!
[2/21, 8:39 AM] Vijayasimha Ajarananda: Loved the moment Alpha Go check-mated the human expert with some "crazy irrational and inhuman" moves confusing the human and making him lose the game.
[2/21, 8:39 AM] Vijayasimha Ajarananda: https://youtu.be/WXuK6gekU1Y?
[2/21, 8:53 AM] Ravi Kiran Bhaskar: I specifically excel at that, still skeptical if I am past the first step 😂
[2/21, 9:20 AM] Prof Asoke Talukdar: Is curiosity based research not about asking stupid (?) questions? 😊
[2/21, 9:24 AM] Rakesh Biswas: Perhaps it's not just the first step but an integral component of many more steps that are arranged like a circle. One has to keep returning to the "stupid" step from time to time!
[2/21, 9:26 AM] Ravi Kiran Bhaskar: Local minima in all its glory 😀
[2/21, 9:28 AM] Khushi : Asking 'stupid (?)’ questions is like using a GPS – it helps you navigate the unknown.
It is essential for the learning.
[2/21, 9:29 AM] Rakesh Biswas: That looks like the Duning Kruger curve. It needs a little bit of a modification! Especially when we reach the plateau we don't know it but actually doubt it.
Also the curve is possibly more complex and intertwined again with the initial maximas and minimas!
[2/21, 9:31 AM] Ravi Kiran Bhaskar: Yup an overly simplistic version for the uninitiated naïveté like me 😃
[2/21, 9:32 AM] Rakesh Biswas: You've really plateaued well 😂
[2/21, 9:45 AM] Prof Asoke Talukdar : Phases of learning:
1) Unconscious Incompetence
2) Conscious Incompetence
3) Conscious Competence
4) Unconscious Competence
[2/21, 9:47 AM] Rakesh Biswas: When we are conscious of our incompetence we could be parallely unconscious about our competence @Ravi Kiran Bhaskar
No comments:
Post a Comment