All of those weird prompting tricks (giving tips, threatening the AI) only work sometimes. The truth is that prompting is often more art than science, yet prompting is still very important. I try to reconcile these facts, and give some prompting advice:
@emollick
-
Incentive Systems Fail Outside Games Without Monetary Rewards
By
–
It turns out that those same incentive systems don’t work well outside of games (unless the carrot is money)
-
Games and Crowdsourcing: AI-Powered Scientific Discovery Projects
By
–
Some major games as work efforts that got some actual results:
Folding at home: https://
foldingathome.org
Games with a purpose (sort of became Duolingo): https://
cmu.edu/homepage/compu
ting/2008/summer/games-with-a-purpose.shtml
…
Galaxy Zoo: https://
zooniverse.org/projects/zooke
eper/galaxy-zoo/
…
Project Discovery: https://
eveonline.com/discovery -
OpenAI’s AGI Strategy: Understanding Analyst Misconceptions
By
–
Analysts make errors because they don’t take the motivations seriously, and assume, for example, that many folks at OpenAI are not sincere about building AGI. Not all of them are, but many truly believe it. If you don’t take that seriously, then OpenAI’s strategy is inexplicable.
-
Understanding motivations behind AGI development and AI pause positions
By
–
Having talked to many people in the AI space, I think you can take goals at face value: those who want to build AGI are serious, those who want to pause AI are serious & those who think AI is hype are serious. Motivations may more hidden, but usually or curiosity or ideology.
-
LLMs as Practical Tools for Everyday Task Automation
By
–
LLMs act like Swiss Army knives, especially if you don’t code a lot I keep having GPT-4 solve tiny problems that would have been annoying, like removing line breaks from text that I copy from PDFs, or a little GPT that turns emails into .ics appointments
-
Bostrom’s Information Hazards Framework for AI Safety
By
–
The paper: https://
nickbostrom.com/information-ha
zards.pdf
… -
Generation Time, Compute Requirements, and Commercial Viability
By
–
The usual questions: how long did this take to generate, how many tries were requires, are the compute requirements actually reasonable for commercial use?
-
Impressive AI Demo Shows Remarkable Persistence Complex Shapes
By
–
Honestly this is the most impressive demo I have seen yet and the one I find really hard to wrap my head around. There is persistence of complex shapes and images through multiple angles and distances & thematic consistency from the start to the end with relatively minor issues. https://t.co/NueYEzraRW
— Ethan Mollick (@emollick) 3 mars 2024Honestly this is the most impressive demo I have seen yet and the one I find really hard to wrap my head around. There is persistence of complex shapes and images through multiple angles and distances & thematic consistency from the start to the end with relatively minor issues.