beyond addressing obvious PR problem ChatGPTs blandness smooths the transition to a world with lifelike AIs after playing around with DAN for a few days, the (semi) lack of censorship provides a striking realism factor that makes conversations slightly more unsettling
GENERATIVE AI
-
ChatGPT transforms simple English into a programming language
By
–
The fact that ChatGPT basically turned simple English into a coding language needs to be discussed more.
-
Browser-native AI webpage summarization feature launches
By
–
Browser-native “shorten webpage” feature. AI features starting to appear in interfaces near you: https://t.co/St1e8LsizK
— Greg Brockman (@gdb) 11 février 2023Browser-native “shorten webpage” feature. AI features starting to appear in interfaces near you:
-
AI Impact on Software Engineering Future with Joseph Sirosh
By
–
Is AI making software engineering the only job of the future? Former Microsoft and Compass CTO, Joseph Sirosh, weighs in on the @storiesinAI podcast.
— Ganesh💰Padmanabhan ⚕️🦾 (@_ganeshp) 11 février 2023
Check out the full podcast here:https://t.co/OXFJpV1133 pic.twitter.com/0mNPVldDViIs AI making software engineering the only job of the future? Former Microsoft and Compass CTO, Joseph Sirosh, weighs in on the @storiesinAI podcast. Check out the full podcast here: https://
youtu.be/njAZXuAUpY4 -
AI conversations integrating multimodal models like DALL·E 2
By
–
Impresionante poder tener conversaciones de este nivel sobre lo que quieras y cuando quieras. Poder soltar conceptos como DALL·E 2 y que los busque y lo integre en una conversación. La respuesta de la IA, muy alineada con la de muchos compis de arte que he hablado estos meses.
-
Transparent AI business model: free content versus paid personalized services
By
–
Y coño, que mi modelo de negocio es muy transparente. ¿Quieres acceder a mis conocimientos de IA?
Sienta el culo y ponte a ver las horas y horas de vídeos gratuitos que hay en Youtube. ¿Que quieres algo rápido, directo y personalizado? Pues paga… o charla con ChatGPT yoquesé -
Data Selection and Distribution Functions in Large Language Models
By
–
The expectation is replaced by an average over tokens (a few trillion for the largest LMs) so F can be very general. If a human is selecting the y’s among other y’s to create a dataset, then the human is F, hopefully being sensible
-
AI devalues creative work: video editor perspective
By
–
Hola La Sexta, acabo de terminar de grabar un metraje de dos horas para mi próximo vídeo. ¿Os lo mando para que me lo editéis? No cuento con presupuesto para hacer este tipo de trabajos pero jajaj lol. Luego harán el reportaje y dirán que la IA devalúa el trabajo y tal.
-
Self-Training: Filtering Functions and Model Ranking Systems
By
–
Self-training: F(x,y) is a filtering/ranking function, eg., what we call a reward/return. The input x may be chosen by humans, but the model generates the y’s and F ranks and selects for further rounds of self-training. F can be explicit or implicit (human in the loop as in RLHF)
-
Learning Methods Unified Through Gradient Optimization Framework
By
–
Funny @sirbayes Learning methods — supervised, RLHF, policy gradients, Dagger, self-training — can be seen as optimisation with the following gradient: grad = Expectation_x,y [ F(x,y) grad log p(y|x) ] Choices of F and how x and y are produced determine the learning type 1/n