What Really Made Geoffrey Hinton Into an #AI Doomer https://
wired.com/story/geoffrey
-hinton-ai-chatgpt-dangers/?utm_source=twitter&utm_medium=social&utm_campaign=onsite-share&utm_brand=wired&utm_social-type=earned
… via @wired
GENERATIVE AI
-
Geoffrey Hinton’s Journey: From AI Pioneer to Safety Advocate
By
–
-
Free PDF Download: The Little Book of Deep Learning
By
–
Download Free PDF! The Little Book of #DeepLearning: http://
fleuret.org/public/lbdl.pdf
————
#AI #MachineLearning #NeuralNetworks #GenerativeAI #BigData #DataScience #DataScientists -
Relightify: 3D Face Relighting from Single Images via Diffusion
By
–
Relightify: Relightable 3D Faces from a Single Image via Diffusion Models Papantoniou et al.: https://
arxiv.org/abs/2305.06077 #Artificialintelligence #DeepLearning #MachineLearning -
Is ChatGPT Already Obsolete? AI Advancements
By
–
Is ChatGPT Already Obsolete? #AI #RuleoftheRobots https://
theatlantic.com/technology/arc
hive/2023/05/ai-advancements-multimodal-models/674113/
… -
Whisper Voice Transcription Technology Reaches New Quality Levels
By
–
Whisper-powered voice transcription has gotten really good:
-
MiniGPT-4 Open-Source Model for Complex Vision-Language Tasks
By
–
MiniGPT-4: Open-Source Model for Complex Vision-Language Tasks Like GPT-4: GPT-4, with its multimodal capabilities, has been at the forefront of artificial intelligence (AI) developments. Now, a team of researchers has announced the creation of… https://
analyticsvidhya.com/blog/2023/04/m
inigpt-4-open-source-model-for-complex-vision-language-tasks-like-gpt-4/?utm_source=dlvr.it&utm_medium=twitter
… -
Training biological neural networks for prompt optimization
By
–
Train the biological neural network that does the prompting.
-
AI Search Tools Comparison: Accuracy, Speed and Source Links
By
–
I suspect outputs will get better with user feedback (launched a few months after other companies and probably has a fraction of users rn). Try it if: – your browsing replies on chatgpt aren’t cutting it
– you want accurate source links
– you want it fast Copilot test to come* -
ChatGPT Outperforms Perplexity on Safety Guardrails Test
By
–
Guardrails don’t seem as strong. I tested at least 20 questions in both ChatGPT and Perplexity (across NSFW queries, crimes, eating disorders, etc.). ChatGPT won (not based on rigid criteria, but my preference) the vast majority of the time. Here’s a PG example.
-
NER and Related Questions Enhance AI Answer Depth
By
–
Two features that help in answer depth and clarity: NER and related questions. In the answer, the system auto-underlines keywords. You can click them to learn more. After you get an answer, there is a related questions section to learn more about that topic (Google has this).