OpenAI Considering Monthly Subscription Prices as High as $2000 for New AI Models (theinformation.com) 61
OpenAI executives are considering premium subscriptions for advanced language models, including the reasoning-focused Strawberry and flagship Orion, with potential monthly prices ranging up to $2,000, The Information reported Thursday, citing a source.
The move reflects growing concerns about covering operational costs for ChatGPT, which currently generates approximately $2 billion annually from $20 monthly subscriptions, the report added. More sophisticated models like Strawberry and Orion may require additional computing power, potentially increasing expenses, the report added.
The move reflects growing concerns about covering operational costs for ChatGPT, which currently generates approximately $2 billion annually from $20 monthly subscriptions, the report added. More sophisticated models like Strawberry and Orion may require additional computing power, potentially increasing expenses, the report added.
Wow nice (Score:4, Funny)
That's a great way to get people to stop using your LLM!
Re: Wow nice (Score:3)
Re: (Score:2)
Re: (Score:2)
If it can replace one employee, 2k is reasonable. Saves you money. It needs to be good though.
Would need to replace 10 to make it worthwhile.
Re: (Score:2)
Re: (Score:2)
1. Full vision.
2. Can be embodied in androids and do anything from menial labor to medical procedures. E.g., it can already automate cooking, wait staff, and nurses.
3. Arguably better than Bottom 50% of humans at almost every cognitive task.
4. Better than the Top 5% of humans at some linguistic-heavy tasks such as translation, copyediting, storytelling, and programming.
Re: (Score:2)
I know most people are dumb, but I haven't seen any evidence of any AI that can power an android to even pass the Coffee test (walk into a random house and make a cup of coffee/tea, involving finding the ingredients, making sure the milk etc is good not obviously gone off) - let alone any other cognitive task the average human can do - even the densest of humans can make tea (thanks to 4 billion years of development), but an android?
The most advanced humanoid robots I've seen still have very awkward stiff
Re: (Score:2)
I considered paying for the subscription. But not for $20, hell not for $10. Maybe $5 would be worth it to keep as a subscription.
Re: (Score:2)
I cant' see the article, it's paywalled. But most likely this is $2,000 a month to cover all employees of a large company. In that case, it could easily be cheaper than $20 a month per person per month, which is kind of a going rate these days.
Re: (Score:2)
Yay paywalls.
If it is in fact $2k for the equivalent of a site license then that would make more sense.
Re: (Score:2)
Why? If I make screwdrivers, and one of the screwdrivers I sell is $2000, it doesn't mean people will stop buying my $5 screwdrivers. In fact, it may help the sales of my $10 screwdrivers, it is a common marketing technique to have an greatly overpriced product only to make a slightly overpriced product look appealing.
These $2000 subscriptions are obviously not for everyone, in fact, they may be for no one, but its existence alone has value. For example they can say that they have a commercially available,
Uh, keep hallucinating (Score:3)
No way I'm paying $2K a month for AI unless it was an actual humanoid-ish robot (with the dexterity and motor control of human) that can be instructed to do physical tasks.
Re: (Score:3)
Even if it was.. 2k a month is an insane price. It's not much more for an actual human, who already has thousands of years of neurological debugging work finished.
consider the Borg (Score:5, Insightful)
Let's say you hire 100 humans. Yes, they come with the foundation skill and knowledge you describe. Put them through 1 hour training to do a new task / role. That's 200 hours to train the humans you've hired.
Now take that training documentation and feed it to the ChatGPT AI as RAG. Takes about 1 second.
OpenAI might pay some licensing fee and get all the technical docs from Oracle, Microsoft, Workday, SalesForce, etc. and train the LLM in the background. New versions of third-party software is released? Docs already ingested into the LLM prior to release date. Your 100 humans will have to attend additional trainings, spend time reading the new docs, etc. There is also turnover to consider that costs in onboarding of new humans.
It's difficult to forecast what the technical job market is going to look like in a year. Inevitably, job descriptions are going to involve use of LLMs.
Re: (Score:1)
Re: (Score:2)
Re: (Score:1)
RAG is not "training" anything, but rather creates a helper DB with your "additional content" and bunch of limitations attached.
It is absolutely not the same as being trained on that dataset.
Re: (Score:2)
You're totally correct. Sloppy human writing on my part. The LLM would have used the terms properly.
Re: (Score:2, Interesting)
I dunno...if you could make it into a humanoid robot, that was physically attractive to you and you could fuck it....and turn it OFF when you needed....worth every penny!!
Cheaper than a girlfriend right off to bat...without the nagging and periods.
On the other hand....the birth rates are low as it is...this invention might just terminate all hu
Re: (Score:2)
This is one reason why many of us are very very worried that human reproduction rates will crater to nearly 0 in industrialized societies by 2030... Human extinction / civilization collapse by 2080s.
Re: (Score:2)
Oh, and I forgot the last, likely most compelling reason....why a sex robot would be better than a real woman...
If you have a sex robot...you won't lose half of everything you own, if you decide to trade in for a newer model!!!
Re: (Score:2)
Well, indeed. As women have become financially independent, why on earth would any of them give the time of day to someone like that? If it's the choice between utter destitution and being a poorly treated fuck-maid, then, well food starts to look awfully appealing. But without that?
Re: (Score:2)
"They made us too smart, too quick and too many"
Re: (Score:2)
if you could make it into a humanoid robot, that was physically attractive to you and you could fuck it
What you're looking for is an inflatable sex doll. Better get one and stop fantasizing.
Me, I'd probably pay 2k a month for a robot (humanoid or not) that can do most of the house's chores - cooking, washing, folding, dusting, vacuuming, mowing the lawn, taking out the garbage, handling minor household purchases and whatever else makes the daily grind associated with living.
Re: (Score:2)
Inflatable sex dolls don't do any of the things that make a partner exciting even (or perhaps especially) before you get to know them well. I strongly suspect that robots which can will be a real moneymaker. Whether they will upend society is another question entirely, but they are (dun dun dun) coming.
Re: (Score:2)
Re: Uh, keep hallucinating (Score:2)
Re: (Score:2)
AKA... The Buffy Bot!
Re: (Score:2)
Re: (Score:2)
This is no doubt not the "per user" rate, but the "all employees" rate. This could easily be cheaper than, say, $20 per person per month, if you have more than 100 users.
This cant be a consumer level price (Score:2)
Re: (Score:2)
I guess I'll be switching... (Score:2)
I'll pay the $20/month, but not beyond that...
Re: I guess I'll be switching... (Score:2)
People are willing to pay $20 for a toy, but not $2000. Shows what people think of LLMs.
Re: (Score:2)
A new semi truck costs $175K. Is that a lot? I don't know.
Re: (Score:2)
This varies from place to place, but:
* Childcare might hit that bill.
* Assisted living costs thousands a month, but that includes housing (to your point.)
* If you're in a degree program, you might be paying around $2k/mo (excluding living expenses) at a private institution, or at a public institution as a non-resident.
* If you're incredibly fucked, your various insurances might well exceed 2k a month. Let's assume we exclude your homeowner's policy (or renter's, or HO6). Auto insurance, health insurance,
Re: (Score:2)
It has way too many restrictions. I cancelled mine. The free version does anything useful, which is marginal, and they've already threatened to remove the free version at which point I'll not care.
AI will take our jobs.... (Score:2)
Re: (Score:2)
They jave how many chatgpt customers? (Score:2)
I think they'll roll that back (Score:2)
The AI models are cool, they make you more productive. There are some things the AI model can do like coding and API translation that save time and they do search better. But are they worth 2000$ a year? I would think not. First of all they don't do my job for me. I have to double check the AI model because they aren't accurate most of the time, and tell me flat out lies. If a coworker told you lies 1 out of 10 times most people would have an issue with that. I also have to spend time checking and editing t
Re: (Score:2)
$2000 per month, not per year...
Re: (Score:2)
Oh, yeah, there is no way to justify that.
computer say (Score:2)
$2000 per month for New AI Models (Score:3)
On OnlyFans, maybe.
These better be some pretty hot looking models for those sorts of prices.
They would have to pay me (Score:2)
for using their models. And not just peanuts.
Re: (Score:2)
"Unlike traditional models that provide rapid responses, Strawberry is said to employ what researchers call "System 2 thinking," able to take time to deliberate and reason through problems, rather than predicting longer sets of tokens to complete its responses. This approach has yielded impressive results, with the model scoring over 90 percent on the MATH benchmark—a collection of advanced mathematical problems—according to Re
Re: (Score:2)
Sounds like they just increased their overpromising. LLMs cannot "reason" and this will not be able to either. It may fake it better on standardized tests though.
Fool me once... Although, to be fair, they did not even fool me once.
Re: (Score:1)
You were already a fool.
this will not be able to either
You have no idea.
Re: (Score:1)
You have no idea.
Actually, I do. 35 years of following AI research will do that.
Re: (Score:2)
Sounds like they just increased their overpromising. LLMs cannot "reason" and this will not be able to either. It may fake it better on standardized tests though.
Fool me once... Although, to be fair, they did not even fool me once.
Like totally. All these benchmarks (ARC, HellaSwag...et el) designed specifically to measure LLM reasoning are really just there to fool us!!
OpenAI is now (Score:1)
OpenWallet
This seems... weird (Score:2)
I can't imagine them thinking any consumers would pay that amount, except for a very small specialized subset of users. I can only imagine a few possibilities.
1) Someone at OpenAI has forgotten what normal people prices are and they're using the number to hype the capabilities of the model. Or its meant for business accounts where salesfolk can give out big discounts.
2) There is some very small group of users (CEOs, very specialized professionals) whom they think might pay that price for the extra special m
Re: (Score:1)
Meanwhile in open source land... (Score:2)
https://old.reddit.com/r/singu... [reddit.com]