It’s possible Anthropic can devote orders of magnitude more compute with more capable internal models
Not really. Intel was famous for employing all the bells and whistles. That was ultimately a major contributing factor to their undoing at 10nm.
Micron is basically saying they have visibility on 3D DRAM and it’s not worth the impact to yields and effective capacity on the changeover to migrate to 4F2 for a couple generations. Very similar to their decision to…
Micron is regarded as a very technologically adept. The have a strong track record of delivering leading/performant dram without the manufacturing bells and whistles used by competitors. The CEO was one of the important…
Can you not feel it’s growing presence?
Ya but the same would be true no matter what kind of industrial facility was across the street.
Are corn ethanol and feedstock alfalfa (the two largest agricultural consumers of water) essential for the survival of the species?
Or if people can easily move. Or if you want the next generation of startups to operate in your state.
The doctor who killed himself was an Australian man living and working in Australia. Australia has public, universal healthcare.
Exactly! I suspect the parent commenter does not interact much with the US healthcare system.
Doesn’t seem to be the case as the price per token is not increasing…
The pace of change is accelerating! If this is an S curve, we are still near the bottom…
The critical back end step for HBM is thermal compression bonding - where they actually compress 13 very thin die separated by solder balls. Cracking is a major issue.
Easy to find yourself in literally 15 seconds.
Is that sort of research unavoidably impaired by a more crowded night sky? Or do we just have to spend more to collect the same quality of data from more or better terrestrial observatories?
I am a science and astronomy fan, but I am sorry, in this case progress is more important. If we regret our decision, the LEOs will fall out of the sky by themselves in a few years and it will be ok.
This comment may have been true 6-9 months ago, but not today
To me it is crazy that you are being downvoted. My experience in academia was that an incredible amount of time was devoted to data cleansing analysis, coding, etc., which were completely non-core to the actual…
Orders of magnitude less compute for pre- and post-training than the the frontier labs.
Intel made bad engineering decisions. There was not under-investment in fabs or R&D.
This model is going to start miles behind the frontier and the gap will only grow.
That is not even close to correct.
It's very easy to find out when the new fabs come online. Try asking Claude or ChatGPT.
I spend much less than that at a chain theater in Manhattan.
AI slop article
It’s possible Anthropic can devote orders of magnitude more compute with more capable internal models
Not really. Intel was famous for employing all the bells and whistles. That was ultimately a major contributing factor to their undoing at 10nm.
Micron is basically saying they have visibility on 3D DRAM and it’s not worth the impact to yields and effective capacity on the changeover to migrate to 4F2 for a couple generations. Very similar to their decision to…
Micron is regarded as a very technologically adept. The have a strong track record of delivering leading/performant dram without the manufacturing bells and whistles used by competitors. The CEO was one of the important…
Can you not feel it’s growing presence?
Ya but the same would be true no matter what kind of industrial facility was across the street.
Are corn ethanol and feedstock alfalfa (the two largest agricultural consumers of water) essential for the survival of the species?
Or if people can easily move. Or if you want the next generation of startups to operate in your state.
The doctor who killed himself was an Australian man living and working in Australia. Australia has public, universal healthcare.
Exactly! I suspect the parent commenter does not interact much with the US healthcare system.
Doesn’t seem to be the case as the price per token is not increasing…
The pace of change is accelerating! If this is an S curve, we are still near the bottom…
The critical back end step for HBM is thermal compression bonding - where they actually compress 13 very thin die separated by solder balls. Cracking is a major issue.
Easy to find yourself in literally 15 seconds.
Is that sort of research unavoidably impaired by a more crowded night sky? Or do we just have to spend more to collect the same quality of data from more or better terrestrial observatories?
I am a science and astronomy fan, but I am sorry, in this case progress is more important. If we regret our decision, the LEOs will fall out of the sky by themselves in a few years and it will be ok.
This comment may have been true 6-9 months ago, but not today
To me it is crazy that you are being downvoted. My experience in academia was that an incredible amount of time was devoted to data cleansing analysis, coding, etc., which were completely non-core to the actual…
Orders of magnitude less compute for pre- and post-training than the the frontier labs.
Intel made bad engineering decisions. There was not under-investment in fabs or R&D.
This model is going to start miles behind the frontier and the gap will only grow.
That is not even close to correct.
It's very easy to find out when the new fabs come online. Try asking Claude or ChatGPT.
I spend much less than that at a chain theater in Manhattan.
AI slop article