Truly one of the best AI channels on the internet right now. Clear, concise and leagues above the repetitive content most creators put out. Thank you for the insights.
Jason, I sincerely wish you success in all that you do and pray for the health and well-being of you and your family. Your presentations exude a genuine joy in sharing what you have discovered and seeking feedback from those of us who are seeking to utilize this technology for good. That's just from my heart to yours. That said, I must admit that my wish is partly motivated by my hope that you will not stop providing what I find to be the most useful and concisely presented AI content. Thank you for being you and doing you.
i was subscribed since the very beginning, but i cannot believe how far this channel has come (and is going). i've truly gained and learned so much from you and i cannot thank you enough.
just helpful tremendously LOL!!! thank you so much bro as always!! I feel like using the guidance technique is kind of an alternative for fine-tuning. As you said previously, fine tuning is a just way to teach the behavior of the model. I am currently building up a tool for retrieving data from Google Analytics 4, which requires GPT to understand and extract info from users' queries and eventually format corresponding info into code snippets so as to be used later as part of google analytic API. and I was initially trying to do this using fine tuning , but after watched this , i feel like i better give a try for guidance first lol, it just gives crazy control over output format
Bro, how is your English so much easier to understand than many native english speakers! I'm blown away by the clarity and easiness in your explanations making everything so easy to understand. Appreciate you fam.
Hi, Its good video and im having trouble when im trying with my own When I try examples with the GPT-4 and GPT3-5 models, I get responses like this: "The best thing about the beach is {{~gen 'best' temperature=0.7 max_tokens=7}}". Are there other models that can be used, or is there something wrong with these?
Great Video 👍I have on very general question: At the end of the day everything you do in e.g. "guidance" is translated down to a prompt right? So if you define options (e.g. "yes" and "no") that only means that there will be a part of the prompt that asks the model to please stick to those options, right? So if the model actually really does this is not guaranteed, right?
I couldn't find many syntaxes you used in the video on GitHub. Is it just making up syntaxes and the GPT model interprets it? Because I cannot find information on conditional logic in the repository.
For me the quality of response in current open source models seems too poor to build business solutions on. I hope they get better (or that my perception is outdated 😉)
Great video Jason 🔥, am learning a lot from you. I really liked the tools you showed at the end too, will be great if you created a video showing all the amazing tools you use and quick overview how they help you.