Open Source AI seems to be setting Silicon Valley up to fail. While they pour hundreds of billions into closed AI systems in the hope they’ll get a ‘Unicorn’ that will dominate the market, at every step Open Source AI equals or exceeds them. If this goes on long enough, eventually the Venture Capitalists are going to lose.
Is the same about to happen with robotics? This announcement is not the first time a Chinese group has open-sourced a robotics model. The US is desperate to slow Chinese technological advancement. Is this all part of Chinese counter-measures? If it isn’t, is it just a coincidence it will severely hamper how Silicon Valley functions?
Why would VC money go into open source?
The entire point of VC is getting a return on investment, where is the return on making something you give away for free?
That’s why your link is about a nonprofit.
Are you really asking when wealthy capitalist will just start giving their wealth away out of the goodness of the heart for the betterment of humanity?
Cuz…
It’s gonna be a while.
oh they wont, however, reverse engineering by open source companies or competing companies like the ones that created R1 have demonstrated that open source AI is not only able to keep up with large AI companies, but also able to offer more or slightly less at certain development points with lesser hardware through more efficient coding. also as AI progresses the ability to use AI, to make AI progresses as well.
open source AI is undermining the industry whenever it can, and i love seeing those corpos lose their shit.
demonstrated that open source AI is not only able to keep up with large AI companies,
Mostly because the hard part is training, not coding…
The VC is what pays for that training, with the hopes that once it’s trained, it will be good enough to make money.
Its great there’s an open source code for an AI, but it’s not what you think is happening.
well if you have an open source model even somewhat on par with top models, curated for your own personal use, scraping information from sources you personally request, or even providing previously censored answers at your request. it evens the playing field for the future open source users drastically (and whatever they may be up against)
also R1 was trained on lesser hardware and and can run on lesser hardware (at home) with more efficient code than chatgpt could at the time of its release, with similar if not better benchmarks in certain aspects.
you can train more efficiently, with more efficient code. and the more training the bot undertakes, the better potential for more advanced efficient code comes down the line. this is central as a function to AGI and ASI self improvement.
Sam Altman, beforehand, was getting massive investments simply because he believed, or at least wanted investors to believe, that they needed more efficient hardware, as that was the only major bottle neck he could communicate that would justify heavy investment. then R1 came out, and Nvidias price collapsed as result, because what sam altman said, wasnt entirely true.
training is essential, integral even, but its made easier with more efficient code. just like a video game that gets pushed out before its fully functional, like a new COD without proper optimization. with massive bloat code that could have been cut down drastically by more efficient coding, if given time and effort. except…deadlines exist.
instead they care about speed of production and benchmark performance being marginally better. they want to throw money and hardware at the issue, while stealing our data, because to them its the easiest route. but as far as has been proven, efficiency through code not only cuts costs for the maker, but also for the consumer.
if everyone could run a top of the line model at home with a few GPUs, customized to their needs. then a subscription model becomes less desirable or relevant, with its limitations. hence why openai and sam altman attacked R1 and its makers, and continued to push the hardware myth. not only for themselves but also for stock prices in companies like Nvidia which were seeing drastic increases in stock prices before R1 crashed them.
the training data exists, and is available, legally and illegally. the code is the AI. cant train an AI with a room full of GPUs and data if the AI doesnt exist. if you had the same amount of hardware for two seperate models and one is coded more efficiently, then obviously that one will win the race. but to get one to do the same thing as the other, on par, with even less hardware, thats arguably much more impressive than a large portfolio of investors throwing cash at a front mans promises of results through hardware and training data alone. especially if the AI cant discern quality training data from slop. which again, would be possible with better coding.
yes please
Open source … HA!