Hacker News Reader

Comment by jumploops | Hacker News Reader

> the idea that I can run a complex prompt and have key details of how that prompt was evaluated hidden from me feels like a big step backwards.

As a developer, this is highly concerning, as it makes it much harder to debug where/how the “reasoning” went wrong. The pricing is also silly, because I’m paying for tokens I can’t see.

As a user, I don’t really care. LLMs are already magic boxes and I usually only care about the end result, not the path to get there.

It will be interesting to see how this progresses, both at OpenAI and other foundation model builders.

View on HN

> As a user, I don’t really care.

Tell me: Just how is it fair for a user to pay for the reasoning tokens without actually seeing them? If they are not shared, the service can bill you anything they want for them!

OpenAI’s answer to this would be, “Okay then, don’t use it.”

The simple answer is: I don't care. I'll statistically figure out what the typical total cost per call is from experience, and that's what matters. Who cares if they lie about it, if the model's cost per call fits my budget?

If it starts costing $1 per call, and that's too high, then I just won't use it commercially. Whether it was $1 because they inflated the token count or because it just actually took a lot of tokens to do its reasoning isn't really material to my economic decision.

The thing is it might increase in cost after you've decided to use it commercially, and have invested a lot of time and resources in it. Now it's very hard to move to something else, but very easy for OpenAI to increase your cost arbitrarily. The statistics you made are not binding for them.

Same applies to every other API in the world, yes.

No, S3 pricing for example is predictable, and written in a contract. There's no way for AWS to charge you 3x amount of dollars for 1GB tomorrow. They need to announce it in advance, and give you time to exit the contract if you disagree with the new price. It's really not the same. OpenAI can just tell you your prompt from tomorrow used up 20x times reasoning tokens. There's no advance warning or predictability. I really don't understand how you can claim the situations are identical.

The API returns how many tokens were used in reasoning, so it would be easy to see any average change in reasoning token consumption. And token prices in general have been extremely deflationary over the last 18 months.

This is experimental, frontier stuff, obviously it comes with risks. Building on GPT-4 in March of 2023 was like that as well, but now you can easily switch between a few models of comparable quality made by different companies (yay capitalism and free markets!). You can risk and use just released stuff right now, or, most likely, come back in 6-12 months (probably earlier) and get several different providers with very similar APIs.

Everything that OpenAI does with LLMs has already been done and validated in the open source community well before OpenAI gets around to it. OpenAI is not an innovator. simbianai/taskgen on github is an example of one such project, although there are others too that don't come to mind right now.

As such, I would never call their work "frontier stuff", but they do bring it to the masses with their commercial service.

The person you are replying to points this out. They make a distinction between developers and users. An end user on a monthly subscription plan doesn’t care about how much compute happens for their chat.

OpenAI could have also figured out the average number of extra output tokens, and put a markup in overall API costs. As a user, I wouldn’t care either, because the price would mostly be the same.

Yeah it is fair. You don't pay a lawyer for 40s of work expecting to see all the research between your consult and the document. You don't pay a cook for a meal and expect to sit and interrogate all the ingredients and the oven temperature.

Actually, if a lawyer is billing you by the minute, then yes, you are entitled to a detailed breakdown. If the lawyer is billing you by the job, then no.

If the output alone is high enough quality, it's worth paying extra.

Pricing for many things in life is abstrated away.

More opportunity for competitors to differentiate.

OpenAI doesn't really have a moat. This isn't payments or SMS where only Stripe or Twilio were trying to win the market. Everybody and their brother is trying to build an LLM business.

Grab some researchers, put some compute dollars in, and out comes a product.

Everyone wants this market. It's absurdly good for buyers.

> As a user, I don’t really care.

People should understand and be able to tinker with the tools they use.

The tragedy of personal computing is that everything is so abstracted away that users use only a fraction of the power of their computer. People who grew up with modern PCs don't understand the concept of memory, and younger people who grew up with cellphones don't understand the concept of files and directories.

Open-weight AI models are great because they let normal users learn how they can make the model work for their particular use cases.

"trust us, we're using your tokens as efficiently as possible"

> As a user, I don’t really care.

As a user, whether of ChatGPT or of the API, I absolutely do care, so I can modify and tune my prompt with the necessary clarifications.

My suspicion is that the reason for hiding the reasoning tokens is to prevent other companies from creating a big CoT reasoning dataset using o1.

It is anti-competitive behavior. If a user is paying through the nose for the reasoning tokens, and yes they are, the user deserves to be able to see them.

>My suspicion is that the reason for hiding the reasoning tokens is to prevent other companies from creating a big CoT reasoning dataset using o1.

I mean...they say as much

Once again true to their name.

Reply on news.ycombinator.com