It doesn’t do that on first run because it does not have reasoning capabilities. That’s why you prompt it again. But starting with o1, it can reason and may detect issues with the response and it will tries to fix it before sending the final results to you. But even at that, having your own agentic workflow will make the final response even better.
It might be possible that referring to it as "zero-shot" comes from a programmer's perspective, as in programming, 0 often represents the starting point, and 1 represents the next step.
1. Like your work very much. Thank you for it! 2. Your question "why doesn't the model do it right at the first place?" My take on that: because of the first prompt, which was not concise enough. However: 2, 3 multi shot prompting, chain of thoughts yields better results anyway. They like step by step approaches anyway. Maybe even computatual resources allocation coded into them = "smart" questioning you'll get more resources. Besides they like to collaborate on evolving projects much more than just answer questions. Now have to watch the rest of the video 😉. Btw like your approach in refeeding the output back into other models. Can be nicely done and automated in workflows with f.i. n8n creating a "swarm of experts" where each one does its own job in the process...
Great analysis, thank you! I need some advice: My OKX wallet holds some USDT, and I have the seed phrase. (alarm fetch churn bridge exercise tape speak race clerk couch crater letter). How can I transfer them to Binance?
* Seconding: Your question “why doesn’t the model do it right in the first place?”. It seems that models have limited computational capacity and like humans if you narrow their focus quality usually improves. I tend to break down prompts once a model gives lousy results. Sometimes in one session if the chunk is somewhat large it does not succeed in accomplishing its task at all (null, zero) and narrowing down things at that point does not seem to work as well. I would have to restart a new session where even (re seeding) might kick in somehow. Not too predictable on larger scales.
Thank u
It doesn’t do that on first run because it does not have reasoning capabilities. That’s why you prompt it again. But starting with o1, it can reason and may detect issues with the response and it will tries to fix it before sending the final results to you. But even at that, having your own agentic workflow will make the final response even better.
It might be possible that referring to it as "zero-shot" comes from a programmer's perspective, as in programming, 0 often represents the starting point, and 1 represents the next step.
1. Like your work very much. Thank you for it!
2. Your question "why doesn't the model do it right at the first place?" My take on that: because of the first prompt, which was not concise enough. However: 2, 3 multi shot prompting, chain of thoughts yields better results anyway. They like step by step approaches anyway. Maybe even computatual resources allocation coded into them = "smart" questioning you'll get more resources. Besides they like to collaborate on evolving projects much more than just answer questions.
Now have to watch the rest of the video 😉.
Btw like your approach in refeeding the output back into other models. Can be nicely done and automated in workflows with f.i. n8n creating a "swarm of experts" where each one does its own job in the process...
Thank you for sharing...good to know!
Great analysis, thank you! I need some advice: My OKX wallet holds some USDT, and I have the seed phrase. (alarm fetch churn bridge exercise tape speak race clerk couch crater letter). How can I transfer them to Binance?
I'm not the person to ask...don't know. Sorry.
* Seconding: Your question “why doesn’t the model do it right in the first place?”. It seems that models have limited computational capacity and like humans if you narrow their focus quality usually improves. I tend to break down prompts once a model gives lousy results. Sometimes in one session if the chunk is somewhat large it does not succeed in accomplishing its task at all (null, zero) and narrowing down things at that point does not seem to work as well. I would have to restart a new session where even (re seeding) might kick in somehow. Not too predictable on larger scales.
WHY DO YOU KEEP INTERRUPTING HIM Grrrrrr
he put a link to the original if that is what you want