Yes and no. American companies have been following OpenAI’s strategy, which is simply scaling up as quickly as possible. From massive data centers swallowing our drinking water for coolant to coal and natural gas power plants to keep it running, it’s been all about pouring as much money and resources as possible in order to scale up to improve their models.
What DeepSeek has done is prove that that’s the wrong way to go about it, and now suddenly, all these companies that have been a massive money sink without any clear path to profitability already have to completely pivot their strategy. Most will probably die before they can. Investors are already selling off their stock.
So AI will become closer to actually being practical/profitable, but I imagine most of the companies who reach that goal won’t be the companies that exist today, and the AI bubble itself will probably collapse from this pivot, if we’re lucky.
Since DeepSeek is also open source, we might even see free competitors that can be run locally pop up that can go toe to toe with the likes of ChatGPT, which would be a real stake through the heart for these massive companies.
I wouldn’t consider DeepSeek open source. A few weeks ago when it would discuss this subject freely with me (it doesn’t anymore), it described keeping some of the most important parts private but the rest being open source. It’s not really open source if it’s only partial because you can’t reproduce it yourself in the same way.
Someone suggested the term ‘open weight’ might be a more honest term. “Open source” has really caught on though.
Deepseek’s massive efficiency improvement ought to improve profitability
Depends on if you’re the AI provider, or the user.
For institutions that had to pay massive fees to use cloud-based AI services, now they might be able to pull it off in house or with far less costs involved. It will save money.
For those selling AI, it’ll get very competitive, and they can’t charge hundreds or thousands of dollars anymore. It will be less profitable or not at all.
AI was already unprofitable
And the silver lining was that all those American companies wasted hundreds of millions, if not, billions on developing the tech. Good for them for wasting all that money.
And good for China for making Deepseek open source as an added “fuck you” to AI capitalists.
For those selling AI, it’ll get very competitive, and they can’t charge hundreds or thousands of dollars anymore. It will be less profitable or not at all.
It is already not at all profitable. Competition will drive prices down, but probably not by as much as the efficiency increase. AI companies could go from having high prices and even higher costs to having low prices and even lower costs. Or they could go under, and be replaced by the competition.
This does the opposite of that. AI was already unprofitable; Deepseek’s massive efficiency improvement ought to improve profitability
Yes and no. American companies have been following OpenAI’s strategy, which is simply scaling up as quickly as possible. From massive data centers swallowing our drinking water for coolant to coal and natural gas power plants to keep it running, it’s been all about pouring as much money and resources as possible in order to scale up to improve their models.
What DeepSeek has done is prove that that’s the wrong way to go about it, and now suddenly, all these companies that have been a massive money sink without any clear path to profitability already have to completely pivot their strategy. Most will probably die before they can. Investors are already selling off their stock.
So AI will become closer to actually being practical/profitable, but I imagine most of the companies who reach that goal won’t be the companies that exist today, and the AI bubble itself will probably collapse from this pivot, if we’re lucky.
Since DeepSeek is also open source, we might even see free competitors that can be run locally pop up that can go toe to toe with the likes of ChatGPT, which would be a real stake through the heart for these massive companies.
I wouldn’t consider DeepSeek open source. A few weeks ago when it would discuss this subject freely with me (it doesn’t anymore), it described keeping some of the most important parts private but the rest being open source. It’s not really open source if it’s only partial because you can’t reproduce it yourself in the same way.
Someone suggested the term ‘open weight’ might be a more honest term. “Open source” has really caught on though.
Depends on if you’re the AI provider, or the user.
For institutions that had to pay massive fees to use cloud-based AI services, now they might be able to pull it off in house or with far less costs involved. It will save money.
For those selling AI, it’ll get very competitive, and they can’t charge hundreds or thousands of dollars anymore. It will be less profitable or not at all.
And the silver lining was that all those American companies wasted hundreds of millions, if not, billions on developing the tech. Good for them for wasting all that money.
And good for China for making Deepseek open source as an added “fuck you” to AI capitalists.
It is already not at all profitable. Competition will drive prices down, but probably not by as much as the efficiency increase. AI companies could go from having high prices and even higher costs to having low prices and even lower costs. Or they could go under, and be replaced by the competition.