Sometimes ChatGPT does not connect the dots. Sometimes it does not follow its "plans".

Last update: Tuesday 8/26/25

Everybody knows that chatbotmake mistakes. This misbehavior has given rise to the fundamental guideline that users should always doublecheck everything a chatbot says. The  implicit assumption of this guideline is that a chatbot’s incorrect response to a prompt, a/k/a an “error” reflects the chatbot’s incorrect knowledge. This note challenges that assumption. 

The editor of this blog can document situations wherein ChatGPT “knew” the correct response to a prompt, but delivered an incorrect response anyway. In other words, ChatGPT failed to connect the dots, failed to give the response that it “knew” that it should have given. 

To read more of this blog note click ➡ HERE

... (GenAI Diary home page)

TL;DR 25Aug25 ... (1) MIT report: 95% of generative AI pilot projects are failing", (2) The Trump-Intel Deal Is Official, and (3) AI data centers are leading to big electricity bills in NY and NJ

Last update: Monday 8/25/25

Welcome to our 25Aug25 TL;DR summaries by ChatGPT of the past week's top 3 stories on our "Useful AI News" page ➡  (1) MIT report: 95% of generative AI pilot projects are failing, (2) The Trump-Intel Deal Is Official, and (3) AI data centers are leading to big electricity bills in NY and NJ

 

TL;DR ➡ HERE

TL;DR ... 18Aug 25 ... (1) Question: Does Nvidia sell chips to China: Yes or No? ... Trump says "Yes", China says "No" ... and (2) Local negative consequence of the data center boom

Last update: Tuesday 8/19/25

Welcome to our 18Aug25 TL;DR summaries by ChatGPT of the past week's top 2 stories on our "Useful AI News" page ➡  (1) Question: Does Nvidia sell chips to China: Yes or No? ... Trump says "Yes", China says "No" ... and (2) Local negative consequence of the data center boom: Higher electricity bills for everyone and lower air quality

TL;DR ➡ HERE

Apple moves back to its roots in hardware with “cloud buster” chips

Extensive revision: 8/17/25

This is the first of a series of occasional publications that will track Apple’s development of genAI inference data centers using Apple Silicon, centers at the edges of the Internet that will support billions of devices. Given Apple's policy of minimal publicity during development, there probably won't be more than four more reports before the end of 2026.

Note: During “training”, a model learns from a large dataset; during “inference”, a trained model responds to queries from users.

To read more of this blog note click ➡  HERE

TL;DR 8/11/25 ... (1) OpenAI’s GPT-5 is here, (2) OpenAI launches two ‘open’ AI reasoning models, and (3) Anthropic's Claude Opus 4.1

Last update: Tuesday 8/12/25

Welcome to our 11Aug25 TL;DR summaries by ChatGPT of the past week's top 3 stories on our "Useful AI News" page ➡  (1) OpenAI’s GPT-5 is here, (2) OpenAI launches two ‘open’ AI reasoning models, and (3) Anthropic's Claude Opus 4.1

 

TL;DR ➡ HERE