Perceptron 1958: Galti Karke Seekhne Wali Pehli Machine

Shodh-patra ek nazar me

1958 me Frank Rosenblatt ne "The Perceptron" me ek aisi machine ka varnan kiya jo udaharan dekhkar seekhti hai — galat jawab par apne andar ke vajan (weights) badal leti hai.

Feynman samjhao: gulel ka nishaana

Socho tum gulel se nishaana laga rahe ho:

• Pehla nishaana daayein chhoota → agli baar thoda baayein
• Phir upar chhoota → thoda neeche
• Har galti ek sandesh hai: kitna, kidhar sudharna hai

Perceptron yehi karta hai — bas tezi se, ganit se:

• Galat jawab = chhoota nishaana
• Sudhaar = galti ke aakaar ke anupaat me vajan badlo
• Niyam pakka: agar hal seedhi rekha se sambhav hai, to machine use zaroor dhoondh legi (convergence theorem)

Akhbaaron ne hungama macha diya tha: "ye machine chalegi, bolegi, sochegi!" Vaada zyada tha, par beej sahi tha.

Kyun mahaan hai?

Ye seekhne wali machine ka pehla gambhir prayaas tha — aur iska "galti se sudhaar" wala niyam aaj ke optimization (gradient descent) ka dada hai. Har AI jo "training" se guzarti hai, 1958 ki us gulel ki vanshaj hai.

Ek line me

Hoshiyaar wo nahi jo galti na kare — jo har galti se nishaana sudhaare.