Message:

<p>[QUOTE=&quot;geekpryde, post: 25722818, member: 36248&quot;]I&#039;m not sure I can agree with this statement.&nbsp; &nbsp; First off, there are several major different types of Machine Learning, and only some of them use Labels / Weights / Buckets or whatever you want to call Humans teaching AI pre-existing ideas by training it with examples.&nbsp; &nbsp; Yes, that would be the most common method for getting the ML to grade and give us results that match the current expected human grade. </p><p><br /></p><p>But what about types of ML that don&#039;t attempt to &quot;train&quot; or structure the learning?&nbsp; Or types of ML that you give it more than zero info, but essentially zero?</p><p><br /></p><p>Example, you give it a many millions of coin images and the ONLY thing you tell it is how much the coin sold for, which is a proxy for how desirable the coin is.</p><p><br /></p><p>You don&#039;t teach it human numeric grades, or give it rarity data, nor mintages, nor concepts like wear, nor idea of PMD, nor luster, nor anything that humans can know about a coin.&nbsp; &nbsp; You don&#039;t even define what a &quot;coin&quot; is.</p><p><br /></p><p>You give the ML images and dollar figures.</p><p><br /></p><p>You don&#039;t think that sophisticated Machine Learning will teach itself the idea of &quot;luster&quot; and that it&#039;s more appealing on some coins that others?&nbsp; &nbsp; Machine learning will likely be able to teach us Humans characteristics of coins that we have NEVER identified that makes coins desirable.&nbsp; Concepts that we don&#039;t have names for yet in numismatics.</p><p><br /></p><p>I am confident that ML will be able to judge emotionally ideas like &quot;desirableness&quot; and &quot;eye-appeal&quot; as well as more set-in-stone Numeric grading based on wear with a lot less structured input data than most people think.</p><p><br /></p><p>I also believe that we will find beauty in areas of numismatics that we humans we learn from a completely unsupervised ML that are not bound by human ideas.&nbsp;&nbsp; In short, the Machines might be teaching us about coins soon enough, not the other&nbsp; way around. </p><p><br /></p><p>An example of this beauty might be for instance Chess Engines that value things differently than human players, are not bound by human ideas, and yet create gameplay that shows beauty and elegance.</p><p><br /></p><p>Just my thoughts, not really looking for an argument with anyone pro or against AI.&nbsp; &nbsp; But I think there is a misconception about exactly what AI is, or what AI needs from humans in order to be able to &quot;learn&quot;.&nbsp; </p><p><br /></p><p>Regarding Chess, that is one are that I think normal people can understand that ML engines have moved way beyond human teachers with weight and measures and pre-cooked ideas from human programmers.</p><p><br /></p><p>AI coin grading might be at where Chess engines were at in the late 90&#039;s era of chess.&nbsp; &nbsp;&nbsp; Imagine Coin grading engines with 20 more years of active development, newer algorithms, like Chess engines have had.</p><p><br /></p><p><img src="styles/default/xenforo/clear.png" class="mceSmilieSprite mceSmilie13" alt=":android:" unselectable="on" unselectable="on" /></p><p><br /></p><p>p.s.&nbsp; Keep in mind, I am no AI &quot;fanboy&quot;.[/QUOTE]</p><p><br /></p>

Your name or email address:

Do you already have an account?

• No, create an account now.
• Yes, my password is:
• Forgot your password?

Stay logged in