3 ms·

Your first link doesn't mention machine learning it is quite literally just a fire in a data center.

rvz 3y ago
The point still stands. It shouldn't have to take an entire fleet of data centers to train a AI model after all the investment in finding efficient methods and all it has resulted in was little to no progress but more CO2 wastage. Here's another example which a Google data center recently caught fire and overheated. [0] By now the field of Deep Learning should already had viable and efficient methods of training, inference, fine-tuning without lots of data centers wasting water and overheating and catching fire. Instead we have more of this AI hype and wastage burning the entire planet. [0] https://www.theverge.com/2022/7/19/23270581/google-cloud-oracle-servers-outage-uk-cooling-failure https://www.theverge.com/2022/7/19/23270581/google-cloud-ora...
- jabradoodle 3y ago
  Again, that is a fire in a data center that has no reported mention of machine learning.. you do realise data centers existed before machine learning. Here is a claim from DeepMind they reduced Google's cooling bill using ML 7 years ago https://www.deepmind.com/blog/deepmind-ai-reduces-google-data-centre-cooling-bill-by-40 https://www.deepmind.com/blog/deepmind-ai-reduces-google-dat...
  - rvz 3y ago
    > Again, that is a fire in a data center that has no reported mention of machine learning.. you do realise data centers existed before machine learning. And clearly it has gotten worse as evidenced by the waste of resources caused by the increased popularity of AI as a service products like ChatGPT. My point still stands. By now, there should be far less (not more) data centers needed for viable AI operations instead of building more of them as AI continues to scale up. This tells us that there is little to no progress in creating alternative and efficient methods in training these deep learning models. > Here is a claim from DeepMind they reduced Google's cooling bill using ML 7 years ago That ‘claim’ doesn't mean anything other than complete green washing nonsense from Google, especially when they have collaborated with oil companies to claim to ‘reduce’ their emissions. [0] Other than more greenwashing, little to nothing has changed. [0] https://theintercept.com/2023/02/17/google-cloud-saudi-aramco-oil/ https://theintercept.com/2023/02/17/google-cloud-saudi-aramc...
    - jabradoodle 3y ago
      As a cloud provider they are providing compute to companies that require compute; how is that newsworthy in anyway? > By now, there should be far less (not more) data centers needed for viable AI operations instead of building more of them as AI continues to scale up As there is more research and more products, using more compute is not surprising. There are renewable forms of electricity that can be used; the fact that oil companies held us back from more renewable means is not a reason to complain about people conducting scientific research.
- CyberDildonics 3y ago
  The point still stands. What point? That computers use electricity? I guess people aren't worried about crypto currency mining any more or movie rendering or youtube data centers or anything else now that AI is in the news. Instead we have more of this AI hype and wastage burning the entire planet. What are you basing this on? Where are your comparisons to jet travel, concrete emissions, electricity from air conditioning, combustion for heat, vehicle emissions etc.

rvz 3y ago

The point still stands. It shouldn't have to take an entire fleet of data centers to train a AI model after all the investment in finding efficient methods and all it has resulted in was little to no progress but more CO2 wastage. Here's another example which a Google data center recently caught fire and overheated. [0] By now the field of Deep Learning should already had viable and efficient methods of training, inference, fine-tuning without lots of data centers wasting water and overheating and catching fire. Instead we have more of this AI hype and wastage burning the entire planet. [0] https://www.theverge.com/2022/7/19/23270581/google-cloud-oracle-servers-outage-uk-cooling-failure https://www.theverge.com/2022/7/19/23270581/google-cloud-ora...

jabradoodle 3y ago
Again, that is a fire in a data center that has no reported mention of machine learning.. you do realise data centers existed before machine learning. Here is a claim from DeepMind they reduced Google's cooling bill using ML 7 years ago https://www.deepmind.com/blog/deepmind-ai-reduces-google-data-centre-cooling-bill-by-40 https://www.deepmind.com/blog/deepmind-ai-reduces-google-dat...
- rvz 3y ago
  > Again, that is a fire in a data center that has no reported mention of machine learning.. you do realise data centers existed before machine learning. And clearly it has gotten worse as evidenced by the waste of resources caused by the increased popularity of AI as a service products like ChatGPT. My point still stands. By now, there should be far less (not more) data centers needed for viable AI operations instead of building more of them as AI continues to scale up. This tells us that there is little to no progress in creating alternative and efficient methods in training these deep learning models. > Here is a claim from DeepMind they reduced Google's cooling bill using ML 7 years ago That ‘claim’ doesn't mean anything other than complete green washing nonsense from Google, especially when they have collaborated with oil companies to claim to ‘reduce’ their emissions. [0] Other than more greenwashing, little to nothing has changed. [0] https://theintercept.com/2023/02/17/google-cloud-saudi-aramco-oil/ https://theintercept.com/2023/02/17/google-cloud-saudi-aramc...
  - jabradoodle 3y ago
    As a cloud provider they are providing compute to companies that require compute; how is that newsworthy in anyway? > By now, there should be far less (not more) data centers needed for viable AI operations instead of building more of them as AI continues to scale up As there is more research and more products, using more compute is not surprising. There are renewable forms of electricity that can be used; the fact that oil companies held us back from more renewable means is not a reason to complain about people conducting scientific research.
CyberDildonics 3y ago
The point still stands. What point? That computers use electricity? I guess people aren't worried about crypto currency mining any more or movie rendering or youtube data centers or anything else now that AI is in the news. Instead we have more of this AI hype and wastage burning the entire planet. What are you basing this on? Where are your comparisons to jet travel, concrete emissions, electricity from air conditioning, combustion for heat, vehicle emissions etc.