There are lots of different ways to approach this topic as there are many areas of concerns to cover. But I've chosen to discuss from the angle of “If open source succeeds at consistently matching private AI in capabilities, what the future looks like for private AI companies.”
I was reading a post on Reddit and it occurred to me that we may just be at the stage where private AI companies have to become aggressive with pricing out competitors.
It is worth highlighting that the recent massive rise of AI data centers has created extreme scarcity for computer chips, RAM, and graphics cards. High-end parts are being heavily hoarded by tech conglomerates. Because components have become exponentially more expensive, tech analysts often comment that "the era of affordable, home-built personal PCs is dying" because consumers are being priced out.
We could argue that recent race by AI companies to secure funding towards aggressively building more infrastructure for AI has been to price out competitors and that has effectively also led to consumer demands for personal compute to not be met.
Now, it is unlikely that personal compute becomes outlawed, especially in a place like the United States, even with recent uproar of “AI risks to humanity,” the best that can be achieved is that consumer access to compute is limited, this is already a thing anyways since computing above certain power levels would attract regulations and at the same time, if private AI companies can consistently fund expansions of infrastructure building for AI, they can consistently buy out supplies before they are even made.
AI is getting cheap
Open weights are making AI cheap and accessible. It would not be an exaggeration to say that open source AI is a threat to private AI companies.
At the end of the day, this is all business for profits and if profit cannot be made, the AI becomes a liability.
So while trying to cure cancer and usher in abundance, the AI companies will need to achieve profitability and open source AI directly attacks their ability to monetize their models, even the most capable ones after each release.
A couple of people have suggested that this could be part of the reasons why top developers are calling for regulations to slow down AI developments.
The monetizable business becomes infrastructure
It's so comical that at the end of the day, what these companies could begin to monetize might really not be the AI they spend so much money building.
It is inevitable that training and running AI becomes cheap, just as AI has made doing several other things cheap.
This reality will force these companies into having to sell infrastructure and not capable models because those will always have an open source alternative free and accessible to anyone.
Open source could consistently kill the pricing power of private AI companies, the business model may shift from selling models to selling infrastructure, data, distribution, or services.
Posted Using INLEO
