Forums
New posts
Search forums
What's new
Articles
Latest reviews
Author list
Classifieds
Log in
Register
What's new
Search
Search
Search titles and first posts only
Search titles only
By:
New posts
Search forums
Menu
Log in
Register
Install the app
Install
Forums
Hunting
Long Range Hunting & Shooting
How fast did weapons and ammo technology really advance and when did it happen?
JavaScript is disabled. For a better experience, please enable JavaScript in your browser before proceeding.
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an
alternative browser
.
Reply to thread
Message
<blockquote data-quote="DNADave" data-source="post: 1871126" data-attributes="member: 110467"><p>People often overstate the potential of AI/machine learning. The amount of data required to adequately train an algorithm to act accordingly within a situation is rather large. Often, the algorithm will be trained to do the right thing only within a very narrow set of cases and will fail spectacularly when operating outside of those bounds. This is why Google/Uber/Tesla/and others are constantly training their self driving cars. The car can only operate within situations it has seen before and the software needs to encounter all possible situations in order to know what to do.</p><p></p><p>I train these algorithms within the health care arena to predict whether you respond to a drug, or not, or to predict whether you will have an adverse event (bad side effect), or not, when taking a drug. It is incredibly difficult. We can often find an algorithm that works well in the data sets we have access to, only to find that it fails when we try it on a new data set that the algorithm hasn't seen.</p><p></p><p>I'm not saying that we'll never get to a point where AI/machine learning isn't more pervasive than it is, and useful, only that its uses are often overstated today. In areas where the outcome is less controlled by external forces/environment, AI /machine learning will certainly be more useful. In biology/health care, we have a much longer road to travel than for machining processes and other processes such as ballistic calculations.</p></blockquote><p></p>
[QUOTE="DNADave, post: 1871126, member: 110467"] People often overstate the potential of AI/machine learning. The amount of data required to adequately train an algorithm to act accordingly within a situation is rather large. Often, the algorithm will be trained to do the right thing only within a very narrow set of cases and will fail spectacularly when operating outside of those bounds. This is why Google/Uber/Tesla/and others are constantly training their self driving cars. The car can only operate within situations it has seen before and the software needs to encounter all possible situations in order to know what to do. I train these algorithms within the health care arena to predict whether you respond to a drug, or not, or to predict whether you will have an adverse event (bad side effect), or not, when taking a drug. It is incredibly difficult. We can often find an algorithm that works well in the data sets we have access to, only to find that it fails when we try it on a new data set that the algorithm hasn't seen. I'm not saying that we'll never get to a point where AI/machine learning isn't more pervasive than it is, and useful, only that its uses are often overstated today. In areas where the outcome is less controlled by external forces/environment, AI /machine learning will certainly be more useful. In biology/health care, we have a much longer road to travel than for machining processes and other processes such as ballistic calculations. [/QUOTE]
Insert quotes…
Verification
Post reply
Forums
Hunting
Long Range Hunting & Shooting
How fast did weapons and ammo technology really advance and when did it happen?
Top