Sunny

← All weeks

What I’ve been reading (W18, 2023)

Google’s AI strategy and Prime Video’s engineering costs

Notes from leaked Google AI document

  • Open-source models are faster, more customizable, more private, and pound-for-pound more capable. They are doing things with $100 and 13B params that we struggle with at $10M and 540B. And they are doing so in weeks, not months. While our [Google] models still hold a slight edge in terms of quality, the gap is closing astonishingly quickly.
  • A tremendous outpouring of innovation followed by Meta’s LLaMA mode release. Many of the new ideas are from ordinary people. The barrier to entry for training and experimentation has dropped from the total output of a major research organization to one person, an evening, and a beefy laptop.
  • Large models aren’t more capable in the long run if we can iterate faster on small models. Focusing on maintaining some of the largest models on the planet actually puts us at a disadvantage.
  • OpenAI doesn’t matter. They are making the same mistakes we are in their posture relative to open source, and their ability to maintain an edge is necessarily in question. Open-source alternatives can and will eventually eclipse them unless they change their stance. In this respect, at least, we can make the first move

Scaling up the Prime Video audio/video monitoring service and reducing costs by 90% + Monoliths are the future

  • Microservices and serverless components are tools that do work at a high scale, but whether to use them over monolith has to be made on a case-by-case basis.
  • Moving our service to a monolith reduced Amazon’s infrastructure cost by over 90%.
  • Monoliths are the future because the problem people are trying to solve with microservices doesn’t really line up with reality.
  • We’re going to break it up and somehow find the engineering discipline we never had in the first place. That means you went from writing bad code to building bad infrastructure.

Cool URIs don’t change

  • Keeping URIs so that they will still be around in 2, 20 or 200, or even 2000 years is clearly not as simple as it sounds. However, all over the Web, webmasters are making decisions which will make it really difficult for themselves in the future.
  • This happens because many are using tools whose task is seen as to present the best site at the moment, and no one has evaluated what will happen to the links when things change.
  • The message here is that many things can change, and your URIs can and should stay the same.

Vicious Traps: When two little things combine to form one huge thing

  • Patience and confidence. They both sound great. But mixed together, they often form stubbornness, which is a disaster. Confidence that you’re right gives you permission to ignore signs that you’re wrong, and patience gives you permission to extend that denial indefinitely.
  • Curiosity and boldness. They are wonderful on their own but combined can easily create impulsiveness.
  • Humility and ambition. Excellent traits, but together they can create successfully disguised arrogance. When you’re humble (“I am fortunate, and could not have achieved this without other peoples’ help”) but ambitious (“I am talented, and can achieve more than I already have”), the mental conflict often settles with hidden arrogance (“I am special, but I can’t say that out loud).

Supplements for healthy ageing

  • Calcium for bone strength
  • Vitamin D for immunity and bone strength
  • Probiotics for gut health
  • Magnesium for mood (Magnesium glycinate)
  • Multivitamins to cover the bases
Substack edition

View the original post

Notes from leaked Google AI document

  • Open-source models are faster, more customizable, more private, and pound-for-pound more capable. They are doing things with $100 and 13B params that we struggle with at $10M and 540B. And they are doing so in weeks, not months. While our [Google] models still hold a slight edge in terms of quality, the gap is closing astonishingly quickly.

  • A tremendous outpouring of innovation followed by Meta’s LLaMA mode release. Many of the new ideas are from ordinary people. The barrier to entry for training and experimentation has dropped from the total output of a major research organization to one person, an evening, and a beefy laptop.

  • Large models aren’t more capable in the long run if we can iterate faster on small models. Focusing on maintaining some of the largest models on the planet actually puts us at a disadvantage.

  • OpenAI doesn’t matter. They are making the same mistakes we are in their posture relative to open source, and their ability to maintain an edge is necessarily in question. Open-source alternatives can and will eventually eclipse them unless they change their stance. In this respect, at least, we can make the first move

Vicious Traps: When two little things combine to form one huge thing

  • Patience and confidence. They both sound great. But mixed together, they often form stubbornness, which is a disaster. Confidence that you’re right gives you permission to ignore signs that you’re wrong, and patience gives you permission to extend that denial indefinitely.

  • Curiosity and boldness. They are wonderful on their own but combined can easily create impulsiveness.

  • Humility and ambition. Excellent traits, but together they can create successfully disguised arrogance. When you’re humble (“I am fortunate, and could not have achieved this without other peoples’ help”) but ambitious (“I am talented, and can achieve more than I already have”), the mental conflict often settles with hidden arrogance (“I am special, but I can’t say that out loud).

Scaling up the Prime Video audio/video monitoring service and reducing costs by 90% + Monoliths are the future

  • Microservices and serverless components are tools that do work at a high scale, but whether to use them over monolith has to be made on a case-by-case basis.

  • Moving our service to a monolith reduced Amazon’s infrastructure cost by over 90%.

  • Monoliths are the future because the problem people are trying to solve with microservices doesn’t really line up with reality.

  • We’re going to break it up and somehow find the engineering discipline we never had in the first place. That means you went from writing bad code to building bad infrastructure. 

Cool URIs don’t change

  • Keeping URIs so that they will still be around in 2, 20 or 200, or even 2000 years is clearly not as simple as it sounds. However, all over the Web, webmasters are making decisions which will make it really difficult for themselves in the future. 

  • This happens because many are using tools whose task is seen as to present the best site at the moment, and no one has evaluated what will happen to the links when things change. 

  • The message here is that many things can change, and your URIs can and should stay the same. 

 Supplements for healthy ageing 

  • Calcium for bone strength

  • Vitamin D for immunity and bone strength

  • Probiotics for gut health

  • Magnesium for mood (Magnesium glycinate)

  • Multivitamins to cover the bases

Originally on Substack