OpenAI's Official ChatGPT API and Whisper API – Pricing & Access Tutorial

23 comments

1. I'm going to have to watch this video again. I seem to think this creates a use case that I didn't think was possible earlier.

   Today I attended an online webinar that lasted for 90 minutes. Can I access whisper in the playground and use that microphone feature you showed to record and then transcribe the webinar?

   Once I have the transcription can I use one of the large language models to create notes or summaries of the transcription. Will the length of the transcription cause any issues?

   If this is possible it would be wonderful if you could create a demo video showing the steps necessary.

   I know that's asking for a lot so if you are not able to show the steps could you please just answer the questions above?

   And thank you so much for these demonstrations!

2. Where is the colab notebook??

3. I didn't realized they increased the price 10x 😛 previously davinci was 0.002 / 1k token

4. i will have to start practising speaking eloquently lol

5. i talked to claude using POE on my Iphone X . its fairly stupid and limited. i am not sure why the Anthropic are aiming to loose the AI battle, it cannot help you with coding at all, neither in copywriting,

   Anthropic has placed some severe limitations citing safety, they have censored the chatbot from providing helpful coding solutions by citing that its PROBABLY not safe as the code could be buggy and can lead to malicious outcome. and i was asking basic python scripts for automating some of my trivial tasks . >

   all it does is to inform you about GK and weather ? who uses AI to ask about weather ? is that the most used feature of a large language models ?

6. hey bro can you make a video on how these technologies leading to easy solution will effect our brain ? we should start anticipating the effect of this technology before it becomes obvious , we know from neuroplasticity that when we don't use it we loose it. how is it gonna impact how we think since we will get used to ASKING instead of EXPLORING it ourselves , i mean if it would all boil down to asking good questions wouldn't that imply that our centres of learning involving problem solving will shrink ?

7. Great content! How can I contact you?

8. I think the reason is that they want to make the most out of their first-mover advantage. They're trying to become the default AI data supplier before others come to the market with similar products, trusting in the stickiness of their product.
   If they pull it off, particularly with businesses that will be less likely to jump suppliers, they're set for some time. They'll find other ways to monetize later. I am sure there'll be many. One of those ways may be giving the premium model to the highest paying companies. If I were them I'd make hedge funds compete against each other in open bids until I reach peak pricing (provided they're not being super undermined in price at that time by Google and others). Another is to associate with companies in the financial sector and get a cut of their profits. Another is to partner with job boards. They have a myriad ways to monetize this thing if they're smart enough.
   But if they don't, they'll be commoditized out of the market and such AI viewed like we view hosting companies: something we all have to use but it's nothing special and their only differentiating factor is price (downwards)

9. Really exciting! I have been using a reverse engineered API for the past few weeks, and I am thinking about moving to this one, although, that one is FREE 🙂

10. Feel very much like, just can not program, if there is a ready-made application can be debugged that would be great

11. the price could be larger for ada, curie, davinci than turbo because turbo cannot be fined tuned like them afaik currently so they are more general than turbo guy 🙂

12. I agree the costs have a strong taste. Maybe they don't want you to fine-tune your own model. Something that should be concerning to OpenM$: If you make it too cheap, people can start to systematically extract the knowledge from the model by smart prompting to grow your own foundation model.

13. I was thinking about you when i saw their release statement. "I hope 1littlecoder will soon release a tutorial on this".😀

14. great summary. thanks!

15. How do you add chatgpt API

16. Now they need to drop Ada-002 price. Crazy that embeddings cost more now.

17. Can we set the temperature for GPT 3.5?

18. Will we be able to finetune it? I mean, can we use the gpt-index with that? Are you planning a google collab soon? Thanks, mate!

19. Thanks for this overview. Exciting times…

20. I'm using this Python code to access the API:

    import openai
    
    openai.api_key = "My Api Key Here"

    MODEL = "gpt-3.5-turbo"
    
    response = openai.ChatCompletion.create(
    
    model=MODEL,
    
    messages=[
    
    {"role": "system", "content": "You are a helpful assistant."},
    
    {"role": "user", "content": "Oi"},
    
    {"role": "assistant", "content": "Oi, em que posso ajudar?"},
    
    {"role": "user", "content": "Qual é o seu nome?"},
    
    ],
    
    temperature=0.7,
    
    max_tokens=2800,
    
    )

    print(response['choices'][0]['message']['content'])

    However, even defining MODEL = "GPT-3.5-Turbo" the output is generated using the model "GPT-3.5-Turbo-0301" instead of "GPT-3.5-Turbo". Do you know why this happens?
    
    "GPT-3.5-Turbo-0301" is also 10 times cheaper?

    Also, do you have any suggestions for me to improve this code?

21. First YouTuber talking about it

22. Interesting. I wonder how long until someone makes a custom interface to get ChatGPT plus performance with PayG pricing? (Instead of $20/mo fixed)

23. how would one train for ecommerce like instacart ? can you make a sample on that? do i need a vector database lke pinecone? or index-gpt?

Comments are closed.