Help
Answers to common questions; use the Contact page if you have other questions.
OpenAI is a company that creates Artificial Intelligence (AI) models. These models, also called LLMs (Large Language Models), are trained on huge volumes of data so that they have, built-into the models, the knowledge to answer questions on a wide variety of topics.
Their first major release was ChatGPT, a website app that can be used to answer many questions posed to it.
SuperChatNow allows you to access more advanced models than the one used by ChatGPT. Once you purchase some credits, you'll be able to ask your chosen model many questions, so the chat process is much like googling, but more centered around answering questions than providing links to websites.
See other questions on this Help FAQ page for more information.
A chat is just a running conversation between you and the AI model. This means a series of questions and answers, or prompts and responses. The implication is that the model will have the context of previous prompts in order to understand the current prompt.
For instance, if you ask "What year was it written?", the model will see this as ambiguous unless you had previously asked something like "Who wrote the play Julius Caesar?".
This information from OpenAI is relevant here:
Which of these [models] you should use depends on the complexity of the tasks you are using the models for. gpt-4 generally performs better on a wide range of evaluations. In particular, gpt-4 is more capable at carefully following complex instructions. By contrast gpt-3.5-turbo is more likely to follow just one part of a complex multi-part instruction. gpt-4 is less likely than gpt-3.5-turbo to make up information, a behavior known as "hallucination". gpt-4 also has a larger context window with a maximum size of 8,192 tokens compared to 4,096 tokens for gpt-3.5-turbo. However, gpt-3.5-turbo returns outputs with lower latency and costs much less per token.OpenAI
ChatGPT stands for 'Chat Generative Pre-Trained Transformer'. It is an LLM (Large Language Model) that generates answers to questions in a human-readable text format.
OpenAI provides many models, but most people use ChatGPT because it's free. For a long time, ChatGPT used their gpt-3 model, which is older and not as powerful as their newer models. Currently, it uses gpt-3.5-turbo, the same as the gpt-3.5-turbo-4k model offered here.
OpenAI also provides "ChatGPT Plus", which has access to gpt-4 and gpt-3.5 in a format similar to ChatGPT, priced at $20/month. (These are probably the gpt-4-8k and gpt-3.5-turbo-4k models.) The problem is that it is very hard to use up $20 worth of chat. A heavy user could use about $3-5 worth in a month; the rest is wasted. This is because chat is so cheap. With SuperChatNow, you can be a heavy user and go a few months on just $12.
Additionly, it would be hard to even be a very heavy user of ChatGPT Plus because it has restrictive rate limits. [Rate limits are restrictions on how many requests you can make in a given time period.] SCN has more lax rate limits.
With SuperChatNow, you can use your credits at your own pace. Plus, you can also use the gpt-3.5-turbo-16k model, which allows for 4x longer conversations than gpt-3.5-turbo-4k. And the gpt-3.5-turbo models are both much cheaper than gpt-4, and are actually faster to respond.
ChatGPT Plus and SuperChatNow both provide image generation. Because image generation is more expensive than chat, you might be able to use up $20 of text and images in a month if you are creating a lot of images. OR, just sign up with SuperChatNow and do your work at your own pace, fast or slow.
Here is how OpenAI explains tokens:
Language models read and write text in chunks called tokens. In English, a token can be as short as one character or as long as one word (e.g., a or apple), and in some languages tokens can be even shorter than one character or even longer than one word. For example, the string "ChatGPT is great!" is encoded into six tokens: ["Chat", "G", "PT", " is", " great", "!"].OpenAI
The total tokens used in a single question/answer (i.e., prompt/response) must be less than the maximum token limit for the model. The ability to have longer chats costs more, and is reflected in the prices.
What is the maximum token limit? Remember the -4k, -16k, and -8k model suffixes? These indicate the length, in tokens, of the longest conversation you can have (where 'k' is 1024, so 4k = 4 * 1024 = 4096 tokens). A conversation, or chat, is a series of back and forth prompts and responses between you and the AI model. The only reason you may want to keep a conversation going is to provide context for the next prompt you enter. The bottom line is that each separate chat is limited in length. Of course, you can start a new chat at any time, though you lose the context of the previous chat.
Now here's the tricky part: when submitting a question to the OpenAI model, you include all the previous prompts and responses so that the model has the context for the conversation. Essentially, it forgets the conversation each time you ask a new question. So this means (the tricky part) that each question (or prompt) is an entire chat! So it is each prompt and response that has the limit of 4k tokens or 8k tokens or whatever.
On the advanced panel on the chat page, you can specify the max_tokens that you want in a response. This is only useful for limiting the output, as OpenAI wants the total number of tokens in the prompt plus the max_token value to be less than the maximum chat length (which comes from that model suffix, like -4k means 4096 tokens maximum). This limits how big max_tokens can be (and SuperChatNow (SCN) makes sure that whatever max_tokens is set to, the value submitted to OpenAI is never too big to violate their rule).
However, you can set the max_token limit to a smaller value, say 200, if you want to constrain the response to be short. So it is really only useful for downsizing responses.
See the FAQ "Why are my credits being used so fast?" for information on how SCN will summarize a chat so that you're not paying for so many tokens in each prompt.
OpenAI charges by which model you pick and how many multiples of 1000 tokens you use. Additionally, they (on some models), charge a different amount for the prompt than for the response.
Using a credit system makes it so you don't really have to care about tokens and how they charge differently in different cases. You can just buy a block of credits and start asking questions.
Also, this system means you can use your credits whenever you wish; you are not tied to using a certain amount per month, for example. If you want a subscription, it still will only renew when your credits run low, and not at a specific time in the future.
You can also switch between models whenever you like (credits will be transferred to the new model proportionally), and you can buy more credits whenever you like.
This relates to the "What are tokens?" question above.
OpenAI charges for both prompt and response text. In a given chat, the prompt also includes all past prompts and responses. This is so the AI model has context on the running conversation. There are 2 ways to help with this situation:
1) Start a new chat if you don't need the context from the previous conversation.
2) SuperChatNow summarizes the 5 previous prompt/responses, so once you have more than 5, the number of credits used for the prompt portion will remain relatively constant. This 'summarization' number (5) will be exposed in a later revision as a value you can change. Summarizing the last 5 means that context prior to that won't be included in the current question, which may or may not be ok, depending on your circumstance.
Including previous prompts/responses in the current prompt is how OpenAI's models work (how they get context), and not something SCN imposes.
- Write clear instructions - provide all details and context that you can.
- Ask the model to adopt a persona - this can cause the output to be much more creative and interesting.
- Tell the model how you want the output formatted. You can tell it: to include an abstract, summary, conclusion; to include bulleted lists; to write the output as a novel, a poem, a song, a news article, etc; to include delimiters to demarcate sections that may be in a different format, like XML or JSON.
- Give the model specific steps to follow to get to an output.
- Give the model examples of what the output format or style should be like.
- Specify the desired length of the output, like "write a short poem" or "write a discussion of current events using 5 paragraphs and no more than 2000 words."
- Give the model references, tell it to use the references in creating its answer. You can also have it provide citations to reference text that it might use.
- Split big, complex tasks into subtasks. Give it information on how to complete each subtask.
- If the model needs to do math or otherwise figure things out and compare this to some other given solution, tell it to first work out its own solution before doing the comparison.
- Ask the model follow up questions, like: "is there more recent information?" or "are there more relevant examples?".
Yes! On the chat page, click the green "image" button (under the blue "Aa" button) to start image mode. The gear icon to the left of the image icon can be used to configure the size and quality of the image you want. Then, just ask the model to generate an image, for instance: "Generate an image of a dog".
The pricing uses the credit model just like chat, where an image costs some number of credits; the amount depends on the plan you're on. Various image sizes will be available. Higher quality and larger images are more expensive.
You have full personal and commercial rights to the images you create. There are no usage restrictions.
SuperChatNow stores up to around 160 medium size question/answer pairs. This equates to around 0.25 MB per user. Once you go over the limit, the oldest chats will be deleted first. In the future, we hope to offer an option to buy more storage space.
We support all major credit cards, debit cards, Paypal, CashApp, Venmo, ApplePay and GooglePay. There are many non-US payment methods as well.
JSON is typically delivered as just text, so to have it formatted, add this sentence to your prompt:
In your answer, precede any JSON parts with the characters: ```json.