Regarding your footnote. Citibank would be insane if they outsourced their data centers to the cloud. I can think of at least three reasons why that would be bad.
1. Price. Given the likely load it will be massively cheaper for Citibank to have its own data centers (or at least its own dedicated racks in known data centers). Cloud works great for situations where your compute requirements are very dynamic but banking is not, typically, an environment where that is the case
2. Confidentiality. Whether or not there are regulations, short of some extremely expensive options (see 1 above) there are significant risks that data could leak due to the inherent shared nature of cloud services and the required public access. For the most part the insecure s3 bucket is a thing of the past but there are plenty of other ways that data can leak that simply do not apply to hosting in a dedicated data center
3. Lack of clarity about resilience. We have seen large cloud providers break their entire network due to errors during upgrades/routine maintenance. A bank has no way to control for this short of having two cloud providers (which is another example of 1) and even with two cloud providers it is entirely possible that there maybe unexpected dependencies that are not exposed until a failure occurs. I know that, for example, a UK financial institution discovered the hard way that renting fiber from two different telcos didn't mean that the two fibers had entirely separate routes between trading floor and data center. They discovered their error when a backhoe dug up the conduit containing both in the short distance where the two fibers shared a conduit.
Dec 24, 2023·edited Dec 25, 2023Liked by Dave Friedman
Thanks a lot for this Interesting writeup. You may have noticed that, while there is a lot of 'anecdotal' evidence of productivity growing (subjective stuff like surveys etc), there is a dearth of 'hard' evidence on productivity increasing. Our randomized controlled trial of real world (FDA approved etc) autonomous AI showed such evidence. It would be great if we can discuss further, as, IMO, healthcare has been at the forefront for real world implementations (read businessmodels) of autonomous AI for a while now.
If I am reading that graph correctly, one consequence of increasingly powerful AI is going to be a heck of a lot of unemployed displaced workers. Yeah, I know. Classical economic theory always says that new technologies end up creating more jobs than they destroy. But AI isn't just a technology. It is, in itself, a worker. An enormously powerful and productive worker. And eventually it will be able to create newer, more powerful, more productive versions of itself - in other words - an entirely new labor force. Then what?
I agree with your assessment that AI technology is rapidly improving, and it is making many people much more productive at their jobs. Over time this will affect the economy in ways which will surprise us, and which will redound to our benefit. Insightful article thank you.
I'm thinking generative AI is a bit oversold. We're not even seeing the possible benefits of every Knowledge Worker having Excel on their computer yet.
Regarding your footnote. Citibank would be insane if they outsourced their data centers to the cloud. I can think of at least three reasons why that would be bad.
1. Price. Given the likely load it will be massively cheaper for Citibank to have its own data centers (or at least its own dedicated racks in known data centers). Cloud works great for situations where your compute requirements are very dynamic but banking is not, typically, an environment where that is the case
2. Confidentiality. Whether or not there are regulations, short of some extremely expensive options (see 1 above) there are significant risks that data could leak due to the inherent shared nature of cloud services and the required public access. For the most part the insecure s3 bucket is a thing of the past but there are plenty of other ways that data can leak that simply do not apply to hosting in a dedicated data center
3. Lack of clarity about resilience. We have seen large cloud providers break their entire network due to errors during upgrades/routine maintenance. A bank has no way to control for this short of having two cloud providers (which is another example of 1) and even with two cloud providers it is entirely possible that there maybe unexpected dependencies that are not exposed until a failure occurs. I know that, for example, a UK financial institution discovered the hard way that renting fiber from two different telcos didn't mean that the two fibers had entirely separate routes between trading floor and data center. They discovered their error when a backhoe dug up the conduit containing both in the short distance where the two fibers shared a conduit.
Great points.
Thanks a lot for this Interesting writeup. You may have noticed that, while there is a lot of 'anecdotal' evidence of productivity growing (subjective stuff like surveys etc), there is a dearth of 'hard' evidence on productivity increasing. Our randomized controlled trial of real world (FDA approved etc) autonomous AI showed such evidence. It would be great if we can discuss further, as, IMO, healthcare has been at the forefront for real world implementations (read businessmodels) of autonomous AI for a while now.
https://www.nature.com/articles/s41746-023-00931-7.epdf
Thanks for the comment. I’d be happy to discuss your paper--maybe I can make a post out of it. What’s the best way for me to contact you?
Great! abramoff@digitaldiagnostics.com or michael-abramoff@uiowa.edu
If I am reading that graph correctly, one consequence of increasingly powerful AI is going to be a heck of a lot of unemployed displaced workers. Yeah, I know. Classical economic theory always says that new technologies end up creating more jobs than they destroy. But AI isn't just a technology. It is, in itself, a worker. An enormously powerful and productive worker. And eventually it will be able to create newer, more powerful, more productive versions of itself - in other words - an entirely new labor force. Then what?
There’s definitely some concern about this. I’m not yet clear on how this will all work out.
I agree with your assessment that AI technology is rapidly improving, and it is making many people much more productive at their jobs. Over time this will affect the economy in ways which will surprise us, and which will redound to our benefit. Insightful article thank you.
I'm thinking generative AI is a bit oversold. We're not even seeing the possible benefits of every Knowledge Worker having Excel on their computer yet.