Prompt to ask chatgpt
- give me an overview in table form of the 20 most established findings in field/subfield x. Give a 10 sentence description per finding.
Pieter retweeted
Stop saying: AI will replace humans.
Start saying: humans who know how to use AI at work will replace those who don’t.
Pieter retweeted
*Two days* after #stablediffusion’s release, we’re already seeing things like this. Gonna be a wild ride. 🔥 – at San Jose, CA
Gonna ship a Figma plugin to go from prompts + simple shapes to design ideas using #stablediffusion #aiart
Pieter retweeted
A photograph of a mannequin wearing a jacket made of moss, studio lighting #dalle
Nature already found a way to solve AGI, and in the process, it made organisms who want to do the same
1
I see AI alignment as impossible if we humans will have to tell it what to do. Instead, it will have to understand us deeply, how our need structure is organised, what we are driven by, and its ingenuity will be really called upon to reconcile the many different human interests
1
'Computers are still so dumb compared to humans' or 'computers could never match human creativity' say a lot about our own intelligence - they wouldn't be so dumb if we were a little smarter
1
A measure of how important you are to someone else is the latency it takes for that person to take the time for you when you need them for something important. Applies to all areas of life
1
1
The aim of AI research endeavors should be not to create an AI lawyer, doctor or scientist, but an AI-baby that could become any one of them if subject to particular domain knowledge
1
A task to test for high level semantics in NLP could be: a large database of arguments in the form of full essays about any kind of topic and a database of quotes and anecdotes. Goal is to find the quotes or anecdotes that most closely match the meaning of a given essay(argument)
1
Dreams are the ultimate VR experience - fully immersed in an imaginary world without knowing you’re in one
1
2
Decision theory is about, or finding out how much you value or like each option, or finding out the probability of each option happening, or both. Although getting good probability estimates can be hard, the nature of how much you value something is much more elusive
1










