Connectionists: The symbolist quagmire

[Ali Minai]

Gary and Steve

My use of the phrase "symbolic quagmire" referred only to the explicitly
symbolic AI models that dominated from the 60s through the 80s. It was not
meant to diminish the importance of understanding symbolic processing and
how a distributed, self-organizing system like the brain does it. That is
absolutely crucial - as long as we let the systems be brain-like, and not
force-fit them into our abstract views of symbolic processing (not saying
that anyone here is doing that, but some others are).

My own - frankly biased and entirely intuitive - opinion is that once we
have a sufficiently brain-like system with the kind of hierarchical
modularity we see in the brain, and sufficiently brain-like learning
mechanisms in all their aspects (base of evolutionary inductive biases,
realized initially through unsupervised learning, fast RL on top of these
coupled with development, then - later - supervised learning in a more
mature system, learning through internal rehearsal, learning by prediction
mismatch/resonance, use of coordination modes/synergies, etc., etc.),
processing that we can interpret as symbolic and compositional will emerge
naturally. To this end, we can try to infer neural mechanisms underlying
this from experiments and theory (as Bengio seems to be doing), but I have
a feeling that it will be hard if we focus only on humans and
human-levelprocessing. First, it's very hard to do controlled experiments
at the required resolution, and second, this is the most complex instance.
As Prof. Smith says in the companion thread, we should ask if animals too
do what we would regard as symbolic processing, and I think that a case can
be made that they do, albeit at a much simpler level. I have long been
fascinated by the data suggesting that birds - perhaps even fish - have the
concept of numerical order, and even something like a number line. If we
could understand how those simpler brains do it, it might be easier to
bootstrap up to more complex instances.

Ultimately we'll understand higher cognition by understanding how it
evolved from less complex cognition. For example, several people have
suggested that abstract representations might be a much more
high-dimensional cortical analogs of 2-dimensional hippocampal place
representations (2-d in rats - maybe higher-d in primates). That would be
consistent with the fact that so much of our abstract reasoning uses
spatial and directional metaphors. Re. System I and System II, with all due
respect to Kahnemann, that is surely a simplification. If we were to look
phylogenetically, we would see the layered emergence of more and more
complex minds all the way from the Cambrian to now. The binary I and II
division should be replaced by a sequence of systems, though, as with
everything is evolution, there are a few major punctuations of
transformational "enabling technologies", such as the bilaterian
architecture at the start of the Cambrian, the vertebrate architecture, the
hippocampus, and the cortex.

Truly hybrid systems - neural networks working in tandem with explicitly
symbolic systems - might be a short-term route to addressing specific
tasks, but will not give us fundamental insight. That is exactly the kind
or "error" that Gary has so correctly attributed to much of current machine
learning. I realize that reductionistic analysis and modeling is the
standard way we understand systems scientically, but complex systems are
resistant to such analysis.

Best
Ali



*Ali A. Minai, Ph.D.*
Professor and Graduate Program Director
Complex Adaptive Systems Lab
Department of Electrical Engineering & Computer Science
828 Rhodes Hall
University of Cincinnati
Cincinnati, OH 45221-0030

Phone: (513) 556-4783
Fax: (513) 556-7326
Email: Ali.Minai at uc.edu
          minaiaa at gmail.com

WWW: https://eecs.ceas.uc.edu/~aminai/ <http://www.ece.uc.edu/%7Eaminai/>


On Mon, Jun 13, 2022 at 1:37 PM <jose at rubic.rutgers.edu> wrote:

> Well. your conclusion is based on some hearsay and a talk he gave, I
> talked with him directly and we discussed what
>
> you are calling SystemII which just means explicit memory/learning to me
> and him.. he has no intention of incorporating anything like symbols or
>
> hybrid Neural/Symbol systems..    he does intend on modeling conscious
> symbol manipulation. more in the way Dave T. outlined.
>
> AND, I'm sure if he was seeing this.. he would say... "Steve's right".
>
> Steve
> On 6/13/22 1:10 PM, Gary Marcus wrote:
>
> I don’t think i need to read your conversation to have serious doubts
> about your conclusion, but feel free to reprise the arguments here.
>
> On Jun 13, 2022, at 08:44, jose at rubic.rutgers.edu wrote:
>
> 
>
> We prefer the explicit/implicit cognitive psych refs. but System II is not
> symbolic.
>
> See the AIHUB conversation about this.. we discuss this specifically.
>
>
> Steve
>
>
> On 6/13/22 10:00 AM, Gary Marcus wrote:
>
> Please reread my sentence and reread his recent work. Bengio has
> absolutely joined in calling for System II processes. Sample is his 2019
> NeurIPS keynote:
> https://www.newworldai.com/system-1-deep-learning-system-2-deep-learning-yoshua-bengio/
> <https://urldefense.com/v3/__https://www.newworldai.com/system-1-deep-learning-system-2-deep-learning-yoshua-bengio/__;!!BhJSzQqDqA!XG4zEf0hOZijhGBf_sFhhbkQzKlArmTaaBCbKV2h_BBa3TSeO_Be99dqthIiW9gcQf1n4qpT0YBNFXEVOgyztpc$>
>
> Whether he wants to call it a hybrid approach is his business but he
> certainly sees that traditional approaches are not covering things like
> causality and abstract generalization. Maybe he will find a new way, but he
> recognizes what has not been covered with existing ways.
>
> And he is emphasizing both relationships and out of distribution learning,
> just as I have been for a long time. From his most recent arXiv a few days
> ago, the first two sentences of which sounds almost exactly like what I
> have been saying for years:
>
> Submitted on 9 Jun 2022]
> On Neural Architecture Inductive Biases for Relational Tasks
> Giancarlo Kerg
> <https://urldefense.com/v3/__https://arxiv.org/search/cs?searchtype=author&query=Kerg*2C*G__;JSs!!BhJSzQqDqA!XG4zEf0hOZijhGBf_sFhhbkQzKlArmTaaBCbKV2h_BBa3TSeO_Be99dqthIiW9gcQf1n4qpT0YBNFXEV3gZmAsw$>
> , Sarthak Mittal
> <https://urldefense.com/v3/__https://arxiv.org/search/cs?searchtype=author&query=Mittal*2C*S__;JSs!!BhJSzQqDqA!XG4zEf0hOZijhGBf_sFhhbkQzKlArmTaaBCbKV2h_BBa3TSeO_Be99dqthIiW9gcQf1n4qpT0YBNFXEVLC65Ftc$>
> , David Rolnick
> <https://urldefense.com/v3/__https://arxiv.org/search/cs?searchtype=author&query=Rolnick*2C*D__;JSs!!BhJSzQqDqA!XG4zEf0hOZijhGBf_sFhhbkQzKlArmTaaBCbKV2h_BBa3TSeO_Be99dqthIiW9gcQf1n4qpT0YBNFXEVsXExRpc$>
> , Yoshua Bengio
> <https://urldefense.com/v3/__https://arxiv.org/search/cs?searchtype=author&query=Bengio*2C*Y__;JSs!!BhJSzQqDqA!XG4zEf0hOZijhGBf_sFhhbkQzKlArmTaaBCbKV2h_BBa3TSeO_Be99dqthIiW9gcQf1n4qpT0YBNFXEVTTRf_9g$>
> , Blake Richards
> <https://urldefense.com/v3/__https://arxiv.org/search/cs?searchtype=author&query=Richards*2C*B__;JSs!!BhJSzQqDqA!XG4zEf0hOZijhGBf_sFhhbkQzKlArmTaaBCbKV2h_BBa3TSeO_Be99dqthIiW9gcQf1n4qpT0YBNFXEVnyKkuNY$>
> , Guillaume Lajoie
> <https://urldefense.com/v3/__https://arxiv.org/search/cs?searchtype=author&query=Lajoie*2C*G__;JSs!!BhJSzQqDqA!XG4zEf0hOZijhGBf_sFhhbkQzKlArmTaaBCbKV2h_BBa3TSeO_Be99dqthIiW9gcQf1n4qpT0YBNFXEVa03VLYM$>
>
> Current deep learning approaches have shown good in-distribution
> generalization performance, but struggle with out-of-distribution
> generalization. This is especially true in the case of tasks involving
> abstract relations like recognizing rules in sequences, as we find in many
> intelligence tests. Recent work has explored how forcing relational
> representations to remain distinct from sensory representations, as it
> seems to be the case in the brain, can help artificial systems. Building on
> this work, we further explore and formalize the advantages afforded by
> 'partitioned' representations of relations and sensory details, and how
> this inductive bias can help recompose learned relational structure in
> newly encountered settings. We introduce a simple architecture based on
> similarity scores which we name Compositional Relational Network
> (CoRelNet). Using this model, we investigate a series of inductive biases
> that ensure abstract relations are learned and represented distinctly from
> sensory data, and explore their effects on out-of-distribution
> generalization for a series of relational psychophysics tasks. We find that
> simple architectural choices can outperform existing models in
> out-of-distribution generalization. Together, these results show that
> partitioning relational representations from other information streams may
> be a simple way to augment existing network architectures' robustness when
> performing out-of-distribution relational computations.
>
>
> Kind of scandalous that he doesn’t ever cite me for having framed that
> argument, even if I have repeatedly called his attention to that oversight,
> but that’s another story for a day, in which I elaborate on some
> Schmidhuber’s observations on history.
>
>
> Gary
>
> On Jun 13, 2022, at 06:44, jose at rubic.rutgers.edu wrote:
>
> 
>
> No Yoshua has *not* joined you ---Explicit processes, memory, problem
> solving. .are not Symbolic per se.
>
> These original distinctions in memory and learning were  from Endel
> Tulving and of course there are brain structures that support the
> distinctions.
>
> and Yoshua is clear about that in discussions I had with him in AIHUB
>
> He's definitely not looking to create some hybrid approach..
>
> Steve
> On 6/13/22 8:36 AM, Gary Marcus wrote:
>
> Cute phrase, but what does “symbolist quagmire” mean? Once upon  atime,
> Dave and Geoff were both pioneers in trying to getting symbols and neural
> nets to live in harmony. Don’t we still need do that, and if not, why not?
>
> Surely, at the very least
> - we want our AI to be able to take advantage of the (large) fraction of
> world knowledge that is represented in symbolic form (language, including
> unstructured text, logic, math, programming etc)
> - any model of the human mind ought be able to explain how humans can so
> effectively communicate via the symbols of language and how trained humans
> can deal with (to the extent that can) logic, math, programming, etc
>
> Folks like Bengio have joined me in seeing the need for “System II”
> processes. That’s a bit of a rough approximation, but I don’t see how we
> get to either AI or satisfactory models of the mind without confronting the
> “quagmire”
>
>
> On Jun 13, 2022, at 00:31, Ali Minai <minaiaa at gmail.com>
> <minaiaa at gmail.com> wrote:
>
> 
> ".... symbolic representations are a fiction our non-symbolic brains
> cooked up because the properties of symbol systems (systematicity,
> compositionality, etc.) are tremendously useful.  So our brains pretend to
> be rule-based symbolic systems when it suits them, because it's adaptive to
> do so."
>
> Spot on, Dave! We should not wade back into the symbolist quagmire, but do
> need to figure out how apparently symbolic processing can be done by neural
> systems. Models like those of Eliasmith and Smolensky provide some insight,
> but still seem far from both biological plausibility and real-world scale.
>
> Best
>
> Ali
>
>
> *Ali A. Minai, Ph.D.*
> Professor and Graduate Program Director
> Complex Adaptive Systems Lab
> Department of Electrical Engineering & Computer Science
> 828 Rhodes Hall
> University of Cincinnati
> Cincinnati, OH 45221-0030
>
> Phone: (513) 556-4783
> Fax: (513) 556-7326
> Email: Ali.Minai at uc.edu
>           minaiaa at gmail.com
>
> WWW: https://eecs.ceas.uc.edu/~aminai/
> <https://urldefense.com/v3/__http://www.ece.uc.edu/*7Eaminai/__;JQ!!BhJSzQqDqA!UCEp_V8mv7wMFGacqyo0e5J8KbCnjHTDVRykqi1DQgMu87m5dBCpbcV6s4bv6xkTdlkwJmvlIXYkS9WrFA$>
>
>
> On Mon, Jun 13, 2022 at 1:35 AM Dave Touretzky <dst at cs.cmu.edu> wrote:
>
>> This timing of this discussion dovetails nicely with the news story
>> about Google engineer Blake Lemoine being put on administrative leave
>> for insisting that Google's LaMDA chatbot was sentient and reportedly
>> trying to hire a lawyer to protect its rights.  The Washington Post
>> story is reproduced here:
>>
>>
>> https://www.msn.com/en-us/news/technology/the-google-engineer-who-thinks-the-company-s-ai-has-come-to-life/ar-AAYliU1
>> <https://urldefense.com/v3/__https://www.msn.com/en-us/news/technology/the-google-engineer-who-thinks-the-company-s-ai-has-come-to-life/ar-AAYliU1__;!!BhJSzQqDqA!UCEp_V8mv7wMFGacqyo0e5J8KbCnjHTDVRykqi1DQgMu87m5dBCpbcV6s4bv6xkTdlkwJmvlIXapZaIeUg$>
>>
>> Google vice president Blaise Aguera y Arcas, who dismissed Lemoine's
>> claims, is featured in a recent Economist article showing off LaMDA's
>> capabilities and making noises about getting closer to "consciousness":
>>
>>
>> https://www.economist.com/by-invitation/2022/06/09/artificial-neural-networks-are-making-strides-towards-consciousness-according-to-blaise-aguera-y-arcas
>> <https://urldefense.com/v3/__https://www.economist.com/by-invitation/2022/06/09/artificial-neural-networks-are-making-strides-towards-consciousness-according-to-blaise-aguera-y-arcas__;!!BhJSzQqDqA!UCEp_V8mv7wMFGacqyo0e5J8KbCnjHTDVRykqi1DQgMu87m5dBCpbcV6s4bv6xkTdlkwJmvlIXbgg32qHQ$>
>>
>> My personal take on the current symbolist controversy is that symbolic
>> representations are a fiction our non-symbolic brains cooked up because
>> the properties of symbol systems (systematicity, compositionality, etc.)
>> are tremendously useful.  So our brains pretend to be rule-based symbolic
>> systems when it suits them, because it's adaptive to do so.  (And when
>> it doesn't suit them, they draw on "intuition" or "imagery" or some
>> other mechanisms we can't verbalize because they're not symbolic.)  They
>> are remarkably good at this pretense.
>>
>> The current crop of deep neural networks are not as good at pretending
>> to be symbolic reasoners, but they're making progress.  In the last 30
>> years we've gone from networks of fully-connected layers that make no
>> architectural assumptions ("connectoplasm") to complex architectures
>> like LSTMs and transformers that are designed for approximating symbolic
>> behavior.  But the brain still has a lot of symbol simulation tricks we
>> haven't discovered yet.
>>
>> Slashdot reader ZiggyZiggyZig had an interesting argument against LaMDA
>> being conscious.  If it just waits for its next input and responds when
>> it receives it, then it has no autonomous existence: "it doesn't have an
>> inner monologue that constantly runs and comments everything happening
>> around it as well as its own thoughts, like we do."
>>
>> What would happen if we built that in?  Maybe LaMDA would rapidly
>> descent into gibberish, like some other text generation models do when
>> allowed to ramble on for too long.  But as Steve Hanson points out,
>> these are still the early days.
>>
>> -- Dave Touretzky
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20220613/c02ac064/attachment.html>