Problem 1 - ChatGPT response depends on whole conversation (that is - on stuff you typed and stuff ChatGPT produced), not only the last question. You should start new chat for every question you ask in this test. Problem 2 - if I understand correctly you asked a question, read ChatGPT answer, and THEN wrote your own solution trying to do better? Why not the other way - write your own solution, and then provide it to ChatGPT and ask it to do better than you? Problem 3 - ChatGPT by default answers as "helpfull assistant". It could give much better answers if you start your prompt by "sort of redefining it - by telling ChatGPT to act/respond as a Behavioral Scientist, be creative etc if that's what you were looking for.
Hello Pete! Amazing video and a great topic. Now I see how AI can bring benefits to behavioral science. I loved how you got some real BS challenges and what would be examples of effective/non-effective solutions. Those kinds of videos are very helpful for people like me, that are starting in the behavioral science world.
According to OpenAI CEO Sam Altman, they have progress of 800% each year, the impressive part is the generality of the text-based tasks that it able to do, but this is only the beginning! far more accuracy and far more forms of "intelligence" will come in the future.
I'm experiencing similar responses from ChatGPT: simple, not always useful responses. However, could we get better responses from ChatGPT if we improve the prompts? In other words, is the root cause of the problem linked to ChatGPT's model, training, etc., or is the problem between the chair and the keyboard? Just a thought.
The prompts have a crucial effect on the quality of the result. This is a very important factor with these models, because they are not trained to produce good results, they are trained to produce results that look like what the average person would reply, or the more likely answers (which are not necessarily the best ones).
![CAMPปลิ้น | EP.83[2/2] ประสบการณ์ขนหัวลุกจากตำนานวงการผี!](http://i.ytimg.com/vi/T_kFecQw7Ds/mqdefault.jpg)
Just watched this video! Congrats Pete! Very interesting, educational, and entertaining!
Problem 1 - ChatGPT response depends on whole conversation (that is - on stuff you typed and stuff ChatGPT produced), not only the last question. You should start new chat for every question you ask in this test. Problem 2 - if I understand correctly you asked a question, read ChatGPT answer, and THEN wrote your own solution trying to do better? Why not the other way - write your own solution, and then provide it to ChatGPT and ask it to do better than you? Problem 3 - ChatGPT by default answers as "helpfull assistant". It could give much better answers if you start your prompt by "sort of redefining it - by telling ChatGPT to act/respond as a Behavioral Scientist, be creative etc if that's what you were looking for.
Hello Pete! Amazing video and a great topic. Now I see how AI can bring benefits to behavioral science. I loved how you got some real BS challenges and what would be examples of effective/non-effective solutions. Those kinds of videos are very helpful for people like me, that are starting in the behavioral science world.
Thanks Vicky! That’s really helpful for me to know
Hey, did you prepare your list after reading ChatGPT answer?
6:21 It’s ChatGPT not “ChatGBD”
Generative Pretraining Transformer
Would love to see an updated video where you use GPT4 instead! :D
Or at least comparing the outputs of GPT3.5 vs GPT4
Nice, very nice. Now let's see BingGPT's card.
Omg, it even has a watermark!
According to OpenAI CEO Sam Altman, they have progress of 800% each year, the impressive part is the generality of the text-based tasks that it able to do, but this is only the beginning! far more accuracy and far more forms of "intelligence" will come in the future.
Love your content
I'm experiencing similar responses from ChatGPT: simple, not always useful responses. However, could we get better responses from ChatGPT if we improve the prompts? In other words, is the root cause of the problem linked to ChatGPT's model, training, etc., or is the problem between the chair and the keyboard? Just a thought.
The prompts have a crucial effect on the quality of the result. This is a very important factor with these models, because they are not trained to produce good results, they are trained to produce results that look like what the average person would reply, or the more likely answers (which are not necessarily the best ones).
@@MrTomyCJ Yup, I understand that now. Thanks.
just some feedback, love the idea for the video and the execution. I think the pacing and the thumbnail were lacking
Do you as a behavioral analyst have to know how to program?
your content is so good
whose*
This gave me a lot of hope for my future career as a behavioral scientist. Thank you Pete!
Where do you study behavioral science
Pete well done with another interesting video. Just fyi the name is chatGPT
Lmao y’all acting like an oat milk seating section is a better idea than an oat milk loyalty card is the funniest shit I’ve seen all day
& lemme tell you if a barista tried to talk to me about climate change I would leave just give me my latte