Latest / Connor with Honor: AI, Real Estate and Real Life / Stop Wasting AI Tokens: Make ChatGPT Choose the Cheapest Model

Transcript

Stop Wasting AI Tokens: Make ChatGPT Choose the Cheapest Model · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: Information about ChatGPT, and maybe this is something that you've been doing, but you know, token spin's an issue.
• 0:06
  SPEAKER_00: I don't know what you do with your ChatGPT or how much you're paying.
• 0:09
  SPEAKER_00: Same thing with Claude, but you know, when you get up there and and you start burning through like everything in a day or two, trying to do good functional work, okay?
• 0:18
  SPEAKER_00: When you do that, you know, it starts to get a little stressful.
• 0:21
  SPEAKER_00: So, you know, I've I've tried to work with Claude, and maybe I'm not doing it the right way.
• 0:26
  SPEAKER_00: You know, there's there's so many things to learn.
• 0:28
  SPEAKER_00: Just a real quick video.
• 0:29
  SPEAKER_00: On on I'll uh I'll plug in what I put in here, but I put in a prompt into Astra.
• 0:36
  SPEAKER_00: I was working with Astra with with another project.
• 0:38
  SPEAKER_00: I put in a prompt on Astra, and then it popped back, it updated all the systems I'm using.
• 0:44
  SPEAKER_00: It came back and it said, I'll proceed with the local model switch over on Frank Vincent.
• 0:49
  SPEAKER_00: Yeah, I name all my stuff, sorry.
• 0:51
  SPEAKER_00: And then it continues, test the candidates on representative tasks and change the defaults at pass.
• 0:56
  SPEAKER_00: Say, wait, whatever, lip service, because I've gotten this before.
• 0:58
  SPEAKER_00: And then it says, now, after that, model recommendation is so high and appropriate for this integration work.
• 1:05
  SPEAKER_00: Trying to integrate some systems using uh some of the installed large language models here to pick up some of the heavy work off the plate of Claude.
• 1:13
  SPEAKER_00: So, you know, it has to go in there, and I want it to actually configure the Spark and do some different things in there.
• 1:19
  SPEAKER_00: So it comes back with this.
• 1:20
  SPEAKER_00: So, yes, it works.
• 1:22
  SPEAKER_00: So if you're concerned about token spin, then do yourself a favor, just ask it to have it monitor your token spin and look at each and every task without burning out tokens, just looking at every task, but then make a determination as to for this particular workflow, what's the best model choice?
• 1:38
  SPEAKER_00: Is it the the highest end Astra, Astra on Ultra, or is it something else?
• 1:43
  SPEAKER_00: Now, of course, if it keeps saying Astra and Ultra, you know, maybe they're stacking it against us.
• 1:47
  SPEAKER_00: But right now it just switched from Astra and Ultra down to uh Seoul, high, which is okay, there we go.
• 1:54
  SPEAKER_00: And I've had other issues or other issues, other indications that it does the same thing, even recommended, even less expensive models.
• 2:02
  SPEAKER_00: That's it.
• 2:02
  SPEAKER_00: Um yeah, very impressed so far.
• 2:04
  SPEAKER_00: Be well.
• 2:04
  SPEAKER_00: Connor with honor, you need me, let me know.
• 2:06
  unknown: Bye.