4 ms·

At the current level of technology, this is quite optimistic... Yes, if the new drug is quite similar to existing ones, the model will probably predict its per

At the current level of technology, this is quite optimistic... Yes, if the new drug is quite similar to existing ones, the model will probably predict its performance and risks. Truly novel drugs, however, require testing as the model's prediction might be random (outside the training set). There is a small but active drug design software business and a small but active quantum chemical software business. The iron is there, so we need to be lucky and have the physics permit good, solid calculations rather than being blocked by entropy. Note: If you're tired of pushing ads and harvesting data, both of these niche specialities would profit from highly-skilled developers. It would be a big win to get the large, somewhat old codes into modern environments and languages such that large-scale clusters can be efficiently used.

chatterhead 4y ago
We are only 60 years past room sized computers. I don't think it's far off at all. In fact, in a sort of prisoners' dilemma situation it behooves us to make sure people believe this technology can do exactly what optimism suggests because those who will benefit from such capabilities would benefit from downplaying their reality. Profit People screw everything up by lacing it with control. The first thread of which is narrative. So, I see your rationally tempered perspective and raise you a wild-eyed vision.
- danielmarkbruce 4y ago
  It was considered a big breakthrough to predict the structure of a single protein. The human body is more complex. Depending on the cell, there are 10's of millions of proteins in a cell, along with other stuff. There are 10's of trillions of cells in the body. If you consider how difficult it would be to accurately simulate that, you start to grasp the problem. Now we might be able to come up with a fairly accurate model which is orders of magnitude smaller, but how you train that model and how small it could be... it seems practically impossible to come up withs something useful. Like, colonizing the sun impossible. This article is quite hand wavy.

chatterhead 4y ago

We are only 60 years past room sized computers. I don't think it's far off at all. In fact, in a sort of prisoners' dilemma situation it behooves us to make sure people believe this technology can do exactly what optimism suggests because those who will benefit from such capabilities would benefit from downplaying their reality. Profit People screw everything up by lacing it with control. The first thread of which is narrative. So, I see your rationally tempered perspective and raise you a wild-eyed vision.

danielmarkbruce 4y ago
It was considered a big breakthrough to predict the structure of a single protein. The human body is more complex. Depending on the cell, there are 10's of millions of proteins in a cell, along with other stuff. There are 10's of trillions of cells in the body. If you consider how difficult it would be to accurately simulate that, you start to grasp the problem. Now we might be able to come up with a fairly accurate model which is orders of magnitude smaller, but how you train that model and how small it could be... it seems practically impossible to come up withs something useful. Like, colonizing the sun impossible. This article is quite hand wavy.