DeepSeek was certain to happen. The only unknown was who was going to do it. The choices were a startup or someone outside the current center of US AI leadership and innovation.
"China took an obvious and clever approach that US companies were mostly blind to because of the path that got them to where they are today before AI."
Looking forward to seeing all the companies around the world using Deepseek's "work smart" approach to pushing forward the AI curve.
In Nov 2023 (Big Tech Are Better Placed to Build the Future https://blog.inverteum.com/p/bigger-is-better), I wrote that big companies will be more innovative due to how ridiculously expensive GPUs are. Never have I been happier to be proven wrong.
I get what you're saying but the efficiencies that Deepseek have found can also be applied by hyperscalers to boost their performance. In the short run, since the scaling laws are just starting for test-time compute and are a much more consistent way to boost model performance as compared to sporadic moments of human ingenuity, I disagree with your assessment but over a longer period of time I agree that the current capex cycle would probably end with a lot of wasted spending by hyperscalers to the benefit of everyone else much like Cisco during the dot com bubble, and such wastage would be amplified if an efficient frontier shifting paradigm shift emerges.
This is a really awesome take - thank you. Strategy in the software business requires this type of historical perspective on innovation. I had to google Sherlocking but I recall a bunch of Windows utilities were similar back in the day. I found a few typos in this article, but I'll have to email them to you.
Check out Bittensor.com. Steve, you just wrote another justification for their decntralized approach. Great article!
Tbh I don't think it's out or up. They need quality data.
"China took an obvious and clever approach that US companies were mostly blind to because of the path that got them to where they are today before AI."
Looking forward to seeing all the companies around the world using Deepseek's "work smart" approach to pushing forward the AI curve.
In Nov 2023 (Big Tech Are Better Placed to Build the Future https://blog.inverteum.com/p/bigger-is-better), I wrote that big companies will be more innovative due to how ridiculously expensive GPUs are. Never have I been happier to be proven wrong.
I get what you're saying but the efficiencies that Deepseek have found can also be applied by hyperscalers to boost their performance. In the short run, since the scaling laws are just starting for test-time compute and are a much more consistent way to boost model performance as compared to sporadic moments of human ingenuity, I disagree with your assessment but over a longer period of time I agree that the current capex cycle would probably end with a lot of wasted spending by hyperscalers to the benefit of everyone else much like Cisco during the dot com bubble, and such wastage would be amplified if an efficient frontier shifting paradigm shift emerges.
This is a really awesome take - thank you. Strategy in the software business requires this type of historical perspective on innovation. I had to google Sherlocking but I recall a bunch of Windows utilities were similar back in the day. I found a few typos in this article, but I'll have to email them to you.
Thank you. Typos are no extra charge. Thank you. Will add a link to Sherlock. I think I caught some of the typos this morning but do send.