[ExI] GPT-4 on its inability to solve the symbol grounding problem

Gadersd gadersd at gmail.com
Thu Apr 6 20:29:24 UTC 2023


> Are you sure you are using the right number here? My understanding was that human DNA is approximately 700 MB (it would fit on a CD-ROM). 

Yes, you are correct. Bing lied to me. I guess I shouldn’t trust Bing at face value.

> On Apr 6, 2023, at 3:55 PM, Jason Resch via extropy-chat <extropy-chat at lists.extropy.org> wrote:
> 
> 
> 
> On Thu, Apr 6, 2023, 3:11 PM Gadersd via extropy-chat <extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org>> wrote:
>> You think humans are simple?  Our brains are the most complex thing in the universe and the more psychology tries to understand it, the more complex it becomes.
> 
> I think behind our complexity is a simple idea repeated many times. Just as any finite computation can be composed from a large enough combination of NAND gates, any brain function could in theory be computed by a large enough collection of neurons. At the top level, the result is an object of incredible complexity, but this is a consequence of the scales involved. The parts and the ideas behind them at the base level, are simple.
> 
> 
> The complexity of humans that pushes our abilities beyond that of bacteria for example in entirely due to differences in DNA. The complexity of the human brain is mostly due to our DNA. Our DNA is about 215 million gigabytes of information. 
> 
> Are you sure you are using the right number here? My understanding was that human DNA is approximately 700 MB (it would fit on a CD-ROM). It's further estimated that roughly half our genes relate to the brain. Then the complexity of defining a baby (untrained) brain is capped at 350 MB. This is smaller than the latest version of Microsoft Word (2 GB).
> 
> Of note, amoeba have significantly larger and more complex genomes than humans. I guess one has to learn and accumulate a lot of tricks after living for billions of years.
> 
> The complexity of the brain after developing its neural network of some 100 billion neurons and 7,000 connections per neurons is (assuming 1000 bits to describe each connection) some 87.5 petabytes.
> https://www.google.com/search?q=1000+bits+*+100+billion*+7000+in+gigabytes <https://www.google.com/search?q=1000+bits+*+100+billion*+7000+in+gigabytes>
> 
> Which is a far cry from the 5000 PB (5 exabytes) of the Internet.
> 
> In my former job, we  sold single storage systems to companies having the capacity to backup one or more whole human connectomes (using my estimates above). Due to storage costs getting cheaper, I estimated that during the 2030s one will be able to buy the storage necessary to backup their brain for a few hundred dollars. We'll see if that happens..
> 
> 
> 
> Jason 
> 
> 
> 
> The total information of the internet is about 5 billion gigabytes. The internet therefore is actually much more complex than human DNA.
> 
> Also it should be pointed out that our DNA is not very efficient in its encoding and not all of DNA encodes properties of our brains. The brain is then strictly less complex than DNA.
> 
> GPT has demonstrated good emulation of human thought and reasoning that gives further evidence that the human mind is derived from simple processes and algorithms. GPT4 is estimated to be about 1 trillion parameters.
> 
>> On Apr 6, 2023, at 1:41 PM, William Flynn Wallace via extropy-chat <extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org>> wrote:
>> 
>> Simple???  You think humans are simple?  Our brains are the most complex thing in the universe and the more psychology tries to understand it, the more complex it becomes.  "The more you know the more you realize just how much you don't know."   I dunno who said that.  bill w
>> 
>> On Thu, Apr 6, 2023 at 9:29 AM Gadersd via extropy-chat <extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org>> wrote:
>> > One possible (and very likely, in my opinion) interpretation of all this is not how clever, or intelligent or conscious or whatever, these LLMs are, but how relatively simple, as well as limited, our own vaunted mental processes really are.
>> 
>> This observation can be extended to nearly everything. The universe is much simpler than humans once thought. That the very encompassing known laws of this world can be enumerated on a postcard would have left the ancient Romans starstruck. It should be no surprise that the workings of the mind would be fundamentally simple. It must be simple enough to encode on DNA if not a postcard after all.
>> 
>> > On Apr 6, 2023, at 3:12 AM, Ben Zaiboc via extropy-chat <extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org>> wrote:
>> > 
>> > One possible (and very likely, in my opinion) interpretation of all this is not how clever, or intelligent or conscious or whatever, these LLMs are, but how relatively simple, as well as limited, our own vaunted mental processes really are.
>> > 
>> > Ben
>> > _______________________________________________
>> > extropy-chat mailing list
>> > extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org>
>> > http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat <http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat>
>> 
>> 
>> _______________________________________________
>> extropy-chat mailing list
>> extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org>
>> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat <http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat>
>> _______________________________________________
>> extropy-chat mailing list
>> extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org>
>> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat <http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat>
> 
> _______________________________________________
> extropy-chat mailing list
> extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org>
> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat <http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat>
> _______________________________________________
> extropy-chat mailing list
> extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org>
> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat <http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230406/e2cf03da/attachment-0001.htm>


More information about the extropy-chat mailing list