Revolut says that in less than a year it’s managed to slash its [code generation] costs by roughly 90 percent by making self-hosted models its default instead of relying primarily on commercial services like Claude.
mlg@lemmy.worldAbout time this hit enterprise.
Been saying the downfall of OpenAI and Anthropic will be local models that can accomplish the same thing without needing a dstacenter worth of GPUs to run.
Plus you have no total token limit on self hosted models. You can let a slop agent run for weeks without ballooning costs or getting rate limited.
And no model censoring, so you can go build that nuclear bomb you always wanted lol.
Both AI giants are already losing a ton of money because their pricing isn’t viable for the amount of training they’re doing trying to be on top. If they readjust, they’re gonna suffocate against much cheaper solutions.
jld@lemmy.zip
If AI isnʻt going to go away, it needs to be cheap, sustainable, open source, and self hosted/ hostable.
dustyData@lemmy.world 7 hours ago
And ethically trained. People should be financially compensated if their data was used to train AI without their consent.
bumblebeer@lemmy.world 6 hours ago
I don’t think it’s really possible to do this.
I mean speaking just logistically, how would you determine who gets how much? If it’s based on quality, how do you determine what is of high quality? Otherwise, does everyone who has every written anything that finds its way into a training dataset get like a tenth of a cent?
Alternative proposal: all AI models must be released as open weights, available to everyone.
AI, through its use of training data, has benefited from the sum total of human knowledge, therefore the results of this training should be given back for the use of all humanity.
That seems fair to me.
eleitl@lemmy.zip 6 hours ago
That soup is already cooked. And obtaining consent is prohibitive, so your data set will be tiny.
cecilkorik@piefed.ca 4 hours ago
Open weight models cannot do this, they do not have any funding. If you demand this, you are just demanding that there will never be any open models, that proprietary AI models controlled by massive trillion dollar corporations will be the only models available, and frankly the cure is worse than the disease.
It’s also impractical, like closing the gate after the horses have already gone. Open-weight, unethically trained models are already out there. They are not going away. You cannot put the genie back in the bottle. They will never earn any money, so they will never be able to compensate for the training that was done to create them, unless you consider the fact that they exist, and they are open to anybody, as at least some measure of compensation.
It’s not enough compensation, but it is something. The training was certainly unethical, but so is an economic system that requires intellectual property to be transactional. Why fix one unethical thing while keeping and reinforcing another?
The only sensible and reasonable way to resolve the issue in the long term is for significant government support and patronage of the arts and content creators (and ideally all people) to be established. People who do this for a living are providing valuable service to all of humanity and that deserves to be recognized not by transactional, metered capitalism that defines each piece of art or content in terms of its raw economic value (a system easily and frequently corrupted anyway if you look at the prevalence of theft, forgery, fraud and money-laundering in the high art world), but by guaranteeing a comfortable and dignified life that allows creative people to freely create the art or other works they love and feel passionate about without worrying that they’re going to starve to death or be forced into a dead-end job if they’re not economically successful.
jobbies@lemmy.zip 3 hours ago
It is, but the issue is the right time to buy hardware was 2 years ago.
Its like 1/ Mooreʻs law.
The right time to buy hardware was always 2 years ago.