<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
On 24/02/2023 19:15, Adrian Tymes wrote:<br>
<blockquote type="cite"
cite="mid:mailman.145.1677266153.733.extropy-chat@lists.extropy.org">
<table class="moz-header-part1" width="100%" cellspacing="0"
cellpadding="0" border="0">
<tbody>
<tr>
<td><br>
</td>
</tr>
<tr>
<td><br>
</td>
</tr>
<tr>
<td><br>
</td>
</tr>
</tbody>
</table>
<table class="moz-header-part2" width="100%" cellspacing="0"
cellpadding="0" border="0">
<tbody>
<tr>
<td><br>
</td>
</tr>
</tbody>
</table>
<br>
<div class="moz-text-html" lang="x-unicode">
<div dir="auto">
<div>
<div class="gmail_quote">
<div dir="ltr" class="gmail_attr">On Fri, Feb 24, 2023,
8:06 AM Ben Zaiboc via extropy-chat <<a
href="mailto:extropy-chat@lists.extropy.org"
moz-do-not-send="true" class="moz-txt-link-freetext">extropy-chat@lists.extropy.org</a>>
wrote:</div>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
<div> If Big Dog (or whatever the latest iteration of it
is called) had ChatGPT in its head, as well as all the
other stuff it already has, would it be able to build
a canoe and use it to escape from a forest fire,
decide where it was safe to stop, and built a hut?
That would be an interesting experiment.<br>
</div>
</blockquote>
</div>
</div>
<div dir="auto"><br>
</div>
<div dir="auto">Obviously it would not, unless Big Dog without
ChatGPT was already capable of that. ChatGPT is a text
parser: it has no input on 3D representations of objects,
let alone performing any physical action (as distinct from
making text that might describe physical actions).</div>
</div>
</div>
</blockquote>
<br>
Ah, that's interesting. I would have thought it could do reasoning
about 3d objects and various real-life situations, based on what
I've read about it here (never used it myself). My thinking was that
the Big Dog robot would have the ability to actually manipulate
objects, sense the world, and move, and the Chatbot would be able to
tell it what to do, with a feedback and translation mechanism
between the two that looks to the Chatbot like someone typing
questions, and to the robot like a set of high-level instructions on
moving about and manipulating things.<br>
<br>
So your reply suggests another interesting experiment: Ask ChatGPT
questions about 3D objects in the real world, and how they should
behave. See if it can predict that a moving object that passes
behind a larger object will reappear after a while, what things will
look like when rotated, that sort of thing.<br>
<br>
Ben<br>
<br>
</body>
</html>