<html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<meta name="Generator" content="Microsoft Word 15 (filtered medium)">
<!--[if !mso]><style>v\:* {behavior:url(#default#VML);}
o\:* {behavior:url(#default#VML);}
w\:* {behavior:url(#default#VML);}
.shape {behavior:url(#default#VML);}
</style><![endif]--><style><!--
/* Font Definitions */
@font-face
{font-family:"Cambria Math";
panose-1:2 4 5 3 5 4 6 3 2 4;}
@font-face
{font-family:Calibri;
panose-1:2 15 5 2 2 2 4 3 2 4;}
@font-face
{font-family:Consolas;
panose-1:2 11 6 9 2 2 4 3 2 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0in;
font-size:11.0pt;
font-family:"Calibri",sans-serif;}
a:link, span.MsoHyperlink
{mso-style-priority:99;
color:blue;
text-decoration:underline;}
pre
{mso-style-priority:99;
mso-style-link:"HTML Preformatted Char";
margin:0in;
font-size:10.0pt;
font-family:"Courier New";}
span.HTMLPreformattedChar
{mso-style-name:"HTML Preformatted Char";
mso-style-priority:99;
mso-style-link:"HTML Preformatted";
font-family:Consolas;}
span.EmailStyle22
{mso-style-type:personal-reply;
font-family:"Calibri",sans-serif;
color:windowtext;}
.MsoChpDefault
{mso-style-type:export-only;
font-size:10.0pt;}
@page WordSection1
{size:8.5in 11.0in;
margin:1.0in 1.0in 1.0in 1.0in;}
div.WordSection1
{page:WordSection1;}
--></style><!--[if gte mso 9]><xml>
<o:shapedefaults v:ext="edit" spidmax="1026" />
</xml><![endif]--><!--[if gte mso 9]><xml>
<o:shapelayout v:ext="edit">
<o:idmap v:ext="edit" data="1" />
</o:shapelayout></xml><![endif]-->
</head>
<body lang="EN-US" link="blue" vlink="purple" style="word-wrap:break-word">
<div class="WordSection1">
<p class="MsoNormal">There was another recent attempt to take down the grandmother cell idea:<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal"><a href="https://www.frontiersin.org/articles/10.3389/fnins.2019.01121/full">Frontiers | The Value of Failure in Science: The Story of Grandmother Cells in Neuroscience | Neuroscience (frontiersin.org)</a><o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal">And here’s my commentary defending grandmother cells:<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal"><a href="https://www.frontiersin.org/articles/10.3389/fnins.2020.00059/full">Frontiers | Commentary: The Value of Failure in Science: The Story of Grandmother Cells in Neuroscience | Neuroscience (frontiersin.org)</a><o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal">By the way, we have had vigorous private arguments over the years about grandmother cells and many were involved – from Jay McClelland and Christof Koch to Walter Freeman and Bernard Baars. As far as I can tell, the brain uses abstractions
at the single cell level that can be argued to be symbols. Short arguments are in the commentary, based on observations by neurophysiologists themselves.<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal">Asim Roy<o:p></o:p></p>
<p class="MsoNormal">Professor, Information Systems<o:p></o:p></p>
<p class="MsoNormal">Arizona State University<o:p></o:p></p>
<p class="MsoNormal"><a href="https://urldefense.proofpoint.com/v2/url?u=https-3A__lifeboat.com_ex_bios.asim.roy&d=DwMFaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=wQR1NePCSj6dOGDD0r6B5Kn1fcNaTMg7tARe7TdEDqQ&m=waSKY67JF57IZXg30ysFB_R7OG9zoQwFwxyps6FbTa1Zh5mttxRot_t4N7mn68Pj&s=oDRJmXX22O8NcfqyLjyu4Ajmt8pcHWquTxYjeWahfuw&e=" target="_blank">Lifeboat
Foundation Bios: Professor Asim Roy</a><o:p></o:p></p>
<p class="MsoNormal"><a href="https://urldefense.proofpoint.com/v2/url?u=https-3A__isearch.asu.edu_profile_9973&d=DwMFaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=wQR1NePCSj6dOGDD0r6B5Kn1fcNaTMg7tARe7TdEDqQ&m=waSKY67JF57IZXg30ysFB_R7OG9zoQwFwxyps6FbTa1Zh5mttxRot_t4N7mn68Pj&s=jCesWT7oGgX76_y7PFh4cCIQ-Ife-esGblJyrBiDlro&e=" target="_blank">Asim
Roy | iSearch (asu.edu)</a><o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<div>
<div style="border:none;border-top:solid #E1E1E1 1.0pt;padding:3.0pt 0in 0in 0in">
<p class="MsoNormal"><b>From:</b> Connectionists <connectionists-bounces@mailman.srv.cs.cmu.edu>
<b>On Behalf Of </b>Gary Marcus<br>
<b>Sent:</b> Friday, February 4, 2022 12:53 PM<br>
<b>To:</b> Stephen José Hanson <jose@rubic.rutgers.edu><br>
<b>Cc:</b> connectionists@mailman.srv.cs.cmu.edu; Stevan Harnad <harnad@ecs.soton.ac.uk>; Francesca Rossi2 <Francesca.Rossi2@ibm.com>; Artur Garcez <arturdavilagarcez@gmail.com>; Anima Anandkumar <anima@caltech.edu>; Luis Lamb <lamb@inf.ufrgs.br>; Gadi Singer
<gadi.singer@intel.com>; Josh Tenenbaum <josh.tenenbaum@gmail.com>; AIhub <aihuborg@gmail.com><br>
<b>Subject:</b> Re: Connectionists: Stephen Hanson in conversation with Geoff Hinton<o:p></o:p></p>
</div>
</div>
<p class="MsoNormal"><o:p> </o:p></p>
<div>
<p class="MsoNormal">Steve,<o:p></o:p></p>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">The phrase I always liked was “poverty of the imagination arguments”; I share your disdain for them. But that’s why I think you should be careful of any retreat into biological plausibility. As even Jay McClelland has acknowledged, we do
know that some humans some of the time manipulate symbols. So wetware-based symbols are not literally biologically impossible; the real question for cognitive neuroscience is about the scope and development of symbols. <o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">For engineering, the real question is, are they useful. Certainly for software engineering in general, they are indispensable.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">Beyond this, none of the available AI approaches map particularly neatly onto what we know about the brain, and none of what we know about the brain is understood well enough to solve AI. All the examples you point to, for instance, are
actually controversial, not decisive. As you probably know, for example, Nancy Kanwisher has a different take on domain-specificity than you do (<a href="https://urldefense.com/v3/__https:/web.mit.edu/bcs/nklab/__;!!IKRxdwAv5BmarQ!JCW7V01kHqhTTrA2Up45kCtD9LH_yvmF66ZxASm8db91n2BEu2yuciVpdlGuWSU$">https://web.mit.edu/bcs/nklab/</a>),
with evidence of specialization early in life, and Jeff Bowers has argued that the grandmother cell hypothesis has been dismissed prematurely (<a href="https://urldefense.com/v3/__https:/jeffbowers.blogs.bristol.ac.uk/blog/grandmother-cells/__;!!IKRxdwAv5BmarQ!JCW7V01kHqhTTrA2Up45kCtD9LH_yvmF66ZxASm8db91n2BEu2yuciVpI7KDNxg$">https://jeffbowers.blogs.bristol.ac.uk/blog/grandmother-cells/</a>);
there’s also a long literature on the possible neural realization of rules, both in humans and other animals. <o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">I don’t know what the right answers are there, but nor do I think that neurosymbolic systems are beholden to them anymore than CNNs are bound to whether or not the brain performs back-propagation.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">Finally, as a reminder, “Distributed” per se in not the right question; in some technical sense ASCII encodings are distributed, and about as symbolic as you can get. The proper question is really what you do with your encodings; the neurosymbolic
approach is trying to broaden the available range of options.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">Gary<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<p class="MsoNormal" style="margin-bottom:12.0pt">On Feb 4, 2022, at 07:04, Stephen José Hanson <<a href="mailto:jose@rubic.rutgers.edu">jose@rubic.rutgers.edu</a>> wrote:<o:p></o:p></p>
</blockquote>
</div>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<p class="MsoNormal"> <o:p></o:p></p>
<p><span style="font-size:13.5pt">Well I don't like counterfactual arguments or ones that start with "It can't be done with neural networks.."--as this amounts to the old Rumelhart saw, of "proof by lack of imagination".</span><o:p></o:p></p>
<p><span style="font-size:13.5pt">I think my position and others (I can't speak for Geoff and won't) is more of a "purist" view that brains have computationally complete representational power to do what ever is required of human level mental processing. AI
symbol systems are remote descriptions of this level of processing. Looking at 1000s of brain scans, one begins to see a pattern of interacting large and smaller scale networks, probably related to Resting state and the Default Mode networks in some important
competitive way. But what one doesn't find is modular structure (e.g. face area.. nope) or evidence of "symbols" being processed. Research on Numbers is interesting in this regard, as number representation should provide some evidence of discrete symbol
processing as would letters. But again the processing states from brain imaging more generally appear to be distributed representations of some sort.</span><o:p></o:p></p>
<p><span style="font-size:13.5pt">One other direction has to do with prior rules that could be neurally coded and therefore provide an immediate bias in learning and thus dramatically reduce the number of examples required for asymptotic learning. Some
of this has been done with pre-training-- on let's say 1000s of videos that are relatively generic, prior to learning on a small set of videos related to a specific topic-- say two individuals playing a monopoly game. In that case, no game-like videos were
sampled in the pre-training, and the LSTM was trained to detect change point on 2 minutes of video, achieving a 97% match with human parsers. In these senses I have no problem with this type of hybrid training.</span><o:p></o:p></p>
<p><span style="font-size:13.5pt">Steve</span><o:p></o:p></p>
<div>
<p class="MsoNormal">On 2/4/22 9:07 AM, Gary Marcus wrote:<o:p></o:p></p>
</div>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<p class="MsoNormal">The whole point of the neurosymbolic approach is to develop systems that can accommodate both vectors and symbols, since neither on their own seems adequate.
<o:p></o:p></p>
<div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">If there are arguments against trying to do that, we would be interested.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><br>
<br>
<o:p></o:p></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<p class="MsoNormal" style="margin-bottom:12.0pt">On Feb 4, 2022, at 4:17 AM, Stephen José Hanson
<a href="mailto:jose@rubic.rutgers.edu"><jose@rubic.rutgers.edu></a> wrote:<o:p></o:p></p>
</blockquote>
</div>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<p class="MsoNormal"> <o:p></o:p></p>
<p><span style="font-size:13.5pt">Geoff's position is pretty clear. He said in the conversation we had and in this thread, "vectors of soft features",</span><o:p></o:p></p>
<p><span style="font-size:13.5pt">Some of my claim is in several of the conversations with Mike Jordan and Rich Sutton, but briefly, there are a number of<br>
very large costly efforts from the 1970s and 1980s, to create, deploy and curate symbol AI systems that were massive failures. Not counterfactuals, but factuals that failed. The MCC comes to mind with Adm Bobby Inmann's national US mandate to counter the
Japanese so called"Fifth-generation AI systems" as a massive failure of symbolic AI.
</span><o:p></o:p></p>
<p><span style="font-size:13.5pt">--------------------</span><o:p></o:p></p>
<p><span style="font-size:13.5pt">In 1982, Japan launched its Fifth Generation Computer Systems project (FGCS), designed to develop intelligent software that would run on novel computer hardware. As the first national, large-scale artificial intelligence (AI)
research and development (R&D) project to be free from military influence and corporate profit motives, the FGCS was open, international, and oriented around public goods.</span><o:p></o:p></p>
<div>
<p class="MsoNormal">On 2/3/22 6:34 PM, Francesca Rossi2 wrote:<o:p></o:p></p>
</div>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<pre>Hi all.<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>Thanks Gary for adding me to this thread.<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>I also would be interested in knowing why Steve thinks that NS AI did not work in the past, and why this is an indication that it cannot work now or in the future.<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>Thanks,<o:p></o:p></pre>
<pre>Francesca.<o:p></o:p></pre>
<pre>------------------<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>Francesca Rossi<o:p></o:p></pre>
<pre>IBM Fellow and AI Ethics Global Leader<o:p></o:p></pre>
<pre>T.J. Watson Research Center, Yorktown Heights, USA<o:p></o:p></pre>
<pre>+1-617-3869639<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>________________________________________<o:p></o:p></pre>
<pre>From: Artur Garcez <a href="mailto:arturdavilagarcez@gmail.com"><arturdavilagarcez@gmail.com></a><o:p></o:p></pre>
<pre>Sent: Thursday, February 3, 2022 6:00 PM<o:p></o:p></pre>
<pre>To: Gary Marcus<o:p></o:p></pre>
<pre>Cc: Stephen José Hanson; Geoffrey Hinton; AIhub; <a href="mailto:connectionists@mailman.srv.cs.cmu.edu">connectionists@mailman.srv.cs.cmu.edu</a>; Luis Lamb; Josh Tenenbaum; Anima Anandkumar; Francesca Rossi2; Swarat Chaudhuri; Gadi Singer<o:p></o:p></pre>
<pre>Subject: [EXTERNAL] Re: Connectionists: Stephen Hanson in conversation with Geoff Hinton<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>It would be great to hear Geoff's account with historical reference to his 1990 edited special volume of the AI journal on connectionist symbol processing. Judging from recent reviewing for NeurIPS, ICLR, ICML but also KR, AAAI, IJCAI (traditionally ZjQcmQRYFpfptBannerStart<o:p></o:p></pre>
<pre>This Message Is From an External Sender<o:p></o:p></pre>
<pre>This message came from outside your organization.<o:p></o:p></pre>
<pre>ZjQcmQRYFpfptBannerEnd<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>It would be great to hear Geoff's account with historical reference to his 1990 edited special volume of the AI journal on connectionist symbol processing.<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>Judging from recent reviewing for NeurIPS, ICLR, ICML but also KR, AAAI, IJCAI (traditionally symbolic), there is a clear resurgence of neuro-symbolic approaches.<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>Best wishes,<o:p></o:p></pre>
<pre>Artur<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre><o:p> </o:p></pre>
<pre>On Thu, Feb 3, 2022 at 5:00 PM Gary Marcus <<a href="mailto:gary.marcus@nyu.edu">gary.marcus@nyu.edu</a><a href="mailto:gary.marcus@nyu.edu"><mailto:gary.marcus@nyu.edu></a>> wrote:<o:p></o:p></pre>
<pre>Steve,<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>I’d love to hear you elaborate on this part,<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre> Many more shoes will drop in the next few years. I for one don't believe one of those shoes will be Hybrid approaches to AI, I've seen that movie before and it didn't end well.<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre><o:p> </o:p></pre>
<pre>I’d love your take on why you think the impetus towards hybrid models ended badly before, and why you think that the mistakes of the past can’t be corrected. Also it’ would be really instructive to compare with deep learning, which lost steam for quite some time, but reemerged much stronger than ever before. Might not the same happen with hybrid models?<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>I am cc’ing some folks (possibly not on this list) who have recently been sympathetic to hybrid models, in hopes of a rich discussion. (And, Geoff, still cc’d, I’d genuinely welcome your thoughts if you want to add them, despite our recent friction.)<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>Cheers,<o:p></o:p></pre>
<pre>Gary<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre><o:p> </o:p></pre>
<pre>On Feb 3, 2022, at 5:10 AM, Stephen José Hanson <<a href="mailto:jose@rubic.rutgers.edu">jose@rubic.rutgers.edu</a><a href="mailto:jose@rubic.rutgers.edu"><mailto:jose@rubic.rutgers.edu></a>> wrote:<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre><o:p> </o:p></pre>
<pre>I would encourage you to read the whole transcript, as you will see the discussion does intersect with a number of issues you raised in an earlier post on what is learned/represented in DLs.<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>Its important for those paying attention to this thread, to realize these are still very early times. Many more shoes will drop in the next few years. I for one don't believe one of those shoes will be Hybrid approaches to AI, I've seen that movie before and it didn't end well.<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>Best and hope you are doing well.<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
<pre>Steve<o:p></o:p></pre>
<pre><o:p> </o:p></pre>
</blockquote>
<div>
<p class="MsoNormal">-- <o:p></o:p></p>
<div>
<p class="MsoNormal"><signature.png><o:p></o:p></p>
</div>
</div>
</div>
</blockquote>
</div>
</div>
</blockquote>
<div>
<p class="MsoNormal">-- <br>
<img border="0" width="402" height="98" style="width:4.1875in;height:1.0208in" id="Picture_x0020_1" src="cid:image001.png@01D81AD9.13EC7A80"><o:p></o:p></p>
</div>
</div>
</blockquote>
</div>
</div>
</body>
</html>