364
The Extreme Cost of Training AI Models.
(cdn.statcdn.com)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 23 Sep 2024
364 points (96.7% liked)
Technology
59346 readers
5268 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
Considering the hype and publicity GPT-4 produced, I don't think this is actually a crazy amount of money to spend.
Yeah, I'm surprised at how low that is, a software engineer in a developed country is about 100k USD per year.
So 40M USD for training ChatGPT 4 is the cost of 400 engineers for one year.
They say cost of salaries could make up to 50% of the total, so the total cost is 800 engineers for one year.
That doesn't seem extreme.
100k USD per engineer assumes they're exclusively hiring from US and Switzerland, that's not a general "developed country" thing. US is an outlier.
US and Switzerland are way over 100k. For Netherlands and Germany 100k is a good approximation for the company costs for a senior SWE.
I did already back up the claim with a source, but okay:
US: Senior 128k USD, mid-level 94k USD
CH: Senior 118k CHF (~139k USD), mid-level 95k CHF (~112k USD)
DE: Senior 72k EUR (~80k USD), mid-level 58k EUR (~65k USD)
NL: Senior 69k EUR (~77k USD), mid-level 52k EUR (~58k USD)
Yes, US and Switzerland are outliers.
Yeah, 80k gross for the worker creates close to 100k costs for the employer.
I'm talking about the cost of the engineer for the company, not the salary, which is less relevant here. In some EU countries, the salaries may be lower, but the taxes are higher to pay for the social system, so the cost for the company is similar.
Yes. Also, Europeans work fewer hours per year. There are big differences between EU countries, though. https://en.wikipedia.org/wiki/List_of_countries_by_average_annual_labor_hours
This is just the estimates to train the model, so it's not accounting for the cost to develop the system for training, collecting the data, etc. This is just pure processing cost, which is staggeringly large numbers.
Comparitively speaking, a lot less hype than their earlier models produced. Hardcore techies care about incremental improvements, but the average user does not. If you try to describe to the average user what is "new" about GPT-4, other than "It fucks up less", you've basically got nothing.
And it's going to carry on like this. New models are going to get exponentially more expensive to train, while producing less and less consumer interest each time, because "Holy crap look at this brand new technology" will always be more exciting than "In our comparitive testing version 7 is 9.6% more accurate than version 6."
And for all the hype, the actual revenue just isn't there. OpenAI are bleeding around $5-10bn (yes, with a b) per year. They're currently trying to raise around $11bn in new funding just to keep the lights on. It costs far more to operate these models (even at the steeply discounted compute costs Microsoft are giving them) than anyone is actually willing to pay to use them. Corporate clients don't find them reliable or adaptable enough to actually replace human employees, and regular consumers think they're cool, but in a "nice to have" kind of way. They're not essential enough a product to pay big money for, but they can only be run profitably by charging big money.
The latest releases ChatGPT 4o costs $600/hr per instance to run based on the discussion I could find about it.
If OpenAI is running 1k of those models to service the demand (they're certainly running more since queries can take 30+ seconds) then that's 200M/yr just keeping the lights on.
That's a lot, but what's their revenue?
3.4bn is their gross - we have no idea what their operating costs are since they refuse to share them.
Some estimates say they're burning 8 billion a year.