Mind the ‘Digital Transformation’ Gap! See http://
bit.ly/DigTransformat
ionNews
… I reflect on new #research by @Broadridge which highlights the significant gap emerging between financial organizations which are leading in #DigitalTransformation vis a vis those who are falling behind –
AI
-

Digital Transformation Gap Emerges in Financial Organizations
By
–
-

Who Funds Education Infrastructure for Innovation Today?
By
–
Who is doing the equivalent to Andrew Carnegie’s libraries now? The 1,500 cities that he gave libraries to in the 1900s had 8-13% higher patenting rates than similar cities without them. Plus, their kids stayed in school longer, and they later chose safer & high prestige careers
-
ChatGPT’s Hallucinations and Narcissism Problem
By
–
It's me, hi
I'm the problem, it's me, ChatGPT Hallucinations, everybody agrees Did you hear my covert narcissism I disguise as erudition? I'm the problem, it's me, ChatGPT -
Llama-3 8B Model Impresses with Local Laptop Performance Potential
By
–
I'm spending the weekend playing with the smaller version of Llama-3 (8bn) that will run on most laptops. It is really impressive, and it makes me think that extremely powerful models will run locally in future.
-

GPT-4 Outperforms Humans in Emotional Reappraisal Tasks
By
–
GPT-4 helped reappraise a difficult emotional situation better than 85% of humans, beating human advice-givers on the effectiveness, novelty, and empathy of their reappraisal. Paying humans to do better did not change the results.
-
Prompt Injection Security: Why Reducing Attack Success Rate Isn’t Enough
By
–
That's the key challenge with prompt injection: reducing to a tiny probability isn't good enough because this is a security vulnerability: if only 1/1000 attacks work then an adversarial attacker will find still find the ones that do
-
Output Mimicry vs Process Understanding in AI Systems
By
–
once upon a time, one could have said the same thing (“they mimic us so well, so they are the best model we have”) about photocopiers or video cameras. mimicking output is not the same as mimicking underlying process.
-

Chatlighting in LLMs: RLHF’s Role in Obsequious Responses
By
–
Question. Suppose we call the kind of unrepetant boot-licking (“I apologize for the error”) that LLMs often do “chatlighting” (h/t @Dolimac
). Why exactly is chatlighting so frequent? Is it a consequence of RLHF? Do we see the same chatlighting behavior in “base models” without -
Whisper Training Data Source YouTube Analysis
By
–
I've seen people say that's because Whisper was trained on huge amounts of YouTube, which feels feasibly to me
-

ChatGPT Voice Hallucinations from Background Noise Fragments
By
–
It's quite fun leaving ChatGPT Voice running and later seeing how it hallucinated increasingly weird fragments of conversation based on snippets of background noise