4 ms·

If it isn’t already obvious - hardware to run these models is the moat

• leoh 2y ago
  The fascinating thing here that I think caught many of us by surprise— and certainly myself — is realizing that building capable LLMs so far is essentially elastic with respect to capital. I don’t think this was immediately obvious when OpenAI first launched ChatGPT. I also don’t think this is true in general for software (though there are other cases).
• RevEng 2y ago
  That simply means money and that's easy to come by. Proprietary technologies, good will, brand name recognition - those are things that are hard to replicate. If all it takes is buying enough GPUs then there is little preventing a competitor from simply buying their way in to the market.
• singularity2001 2y ago
  when Apple studio comes with a one terabyte integrated ram option the hardware to run even the best llama model will become widely available. For now we are stuck with a medium llama model which is still surprisingly good