AI Credit Usage

I must be totally clueless, as I see no credits whatsoever, nor do I understand what AI is here, and the entire subject.
I'm pretty sure as a Pro subscriber, converted from the previous subscription, I get access to AI feature, and supposedly, :"Lots of credits" but despite all the notes I have seen, I still no less than I did before they even started this thing.

Thanks for the link. This is a little bit unrelated. And a product suggestion to Logos product team. Logos can actually cache the AI results by storing the results in a data lake. When a particular passage is being requested to be summarized, the tool can first fetch what is stored, if it has not been requested. But since the prompts are controlled by Logos, then it can start to store the results. This can save Logos cost.

@Graham Criddle yes, they could absolutely do it ahead of time. However, in the commercial model of most AI usage based on "tokens," it would charge against Logos account without attributing the usage to a user. Technically, when we use the AI summarization, we are using an allocated tokens, and these usage is bundled with the subscription fee. So, what Logos product could do is that if someone runs a summary of a chapter of a book, then Logos should just tag that and store the summary in the data lake. It can be reused by others without using up that AI token because its just a stored data similar to how we would store personal Notes. Now, Logos product team will have to do research to see if this is something that will save money and decrease overall utilization of AI tokens. This is also a data engineering question and it's something that many companies are starting to do. It may work for Logos because Logos has a limited set of data set (the collection of books in its inventory). It doesn't work for internet searches because of the variables (current events, news, etc.).

Just a thought to support Logos and its ongoing iteration of improvement.

Here's an explanation of Tokens:

In artificial intelligence (AI), a "token" is the smallest unit of data used to process and analyze text, essentially breaking down sentences into manageable pieces like words or even parts of words, allowing AI models to understand and generate human language more effectively; it's like how we break down sentences into words when reading, but with more nuanced capabilities to capture context and meaning within a language model.
Key points about tokens in AI:

• Function: Tokens act as the building blocks for AI models, particularly in Natural Language Processing (NLP) tasks, where they enable the analysis of text by breaking it down into smaller units that can be easily processed by the algorithm.
• How they work: When you input text into an AI system, it first "tokenizes" the text, meaning it splits it into individual tokens based on predefined rules (like spaces between words, punctuation marks).
• Example: "The quick brown fox jumps over the lazy dog" would be tokenized as ["The", "quick", "brown", "fox", "jumps", "over", "the", "lazy", "dog"].
• Advanced tokenization: Some AI models use more sophisticated tokenization methods, like "subword tokenization," where words can be broken down into smaller parts (like "un" and "derstand" from "understand") to better handle rare words or complex language.

Why tokens are important:

• Context understanding: By processing tokens together, AI models can grasp the relationships between words and understand the overall meaning of a sentence, which is crucial for tasks like translation, sentiment analysis, and text generation.
• Efficiency: Breaking down text into tokens allows for faster processing and analysis by the AI model compared to handling entire sentences at once.
• Flexibility: Different AI models can use different tokenization strategies depending on the specific task and language they are designed for.

Important considerations with tokens:

• Token limits: Most AI models have a maximum number of tokens they can process in a single input, so users need to be mindful of the length of their queries.
• Cost implications: In some AI services, users are charged based on the number of tokens used.

Thanks. Very helpful. At least as far as locating the info needed.
As far as the AI credits go; Im not sure it's all that. But, thats a whole different issue. I guess some AI software is better than others. Although quite inaccurate too often, MS' Copilot I actually like. It's like talking to a real person; it tends to understand what I am saying, and is relatively cool. Most other bots, are annoying, stupid, and a pain, as they have absolutely no clue (Like CVS, and most retailers automated bots that just need to go away). Since this is too new for me, I am unsure where I file this one. But, so far, it just seems like the same search we had before. So, not sure where the AI comes in. Yet.