[ExI] all we are is just llms was

Giovanni Santostasi gsantostasi at gmail.com
Mon Apr 24 06:48:02 UTC 2023


How language influences the color we see:
https://www.youtube.com/watch?v=cGZJflerLZ4

On Sun, Apr 23, 2023 at 11:01 PM Giovanni Santostasi <gsantostasi at gmail.com>
wrote:

> Let say something provocatory, but I want really to drive the point. It is
> childish to think that
> [image: image.png] is not a symbol or a "word" that the brain invented
> for itself. It is a nonverbal symbol but it is a symbol, it is a "word". It
> is so obvious to me, not sure why it is not obvious to everybody else.
> Would it be less mysterious if we heard a melody when we see a
> strawberry (we hear a pitch when we hit a glass with a fork), if we heard a
> little voice in our head that says "red", in fact we do when we learn to
> associate [image: image.png] with "red". There are neuroscientists who
> invented a vest with actuators that react when a magnetic field is present.
> It is interesting but not something that should case endless debate about
> the incommunicability of qualia. What is really interesting in an
> experiment like that is how the brain rewires to adapt to this new sensory
> information.
>  The brain had to invent a way to alert us of the presence of objects that
> reflect a certain range of light frequencies and it came up with [image:
> image.png]. Great, what is the fuss about?
> The communication issue is not an issue. Here I tell you what red means to
> me, this: [image: image.png]. Do you agree that this is what you "mainly"
> see when you see a strawberry or a firetruck? Yes, great, time to move on.
> Can I robot learn what color a firetruck is? Yes, it is already done, the
> word red suffices for all purposes necessary in terms of what
> a conversational AI needs.
> It is a different business for an AI that needs to move in the real world
> and it is trivial to  teach an AI how to recognize
> [image: image.png] if given optical sensors.
> Nothing else is interesting or fascinating about this, not at least from a
> scientific perspective. If silly philosophers want to debate this let them,
> this why they are irrelevant in the modern world.
>
> Giovanni
>
>
> On Sun, Apr 23, 2023 at 10:42 PM Jason Resch via extropy-chat <
> extropy-chat at lists.extropy.org> wrote:
>
>>
>>
>> On Sun, Apr 23, 2023 at 11:16 PM Gordon Swobe <gordon.swobe at gmail.com>
>> wrote:
>>
>>>
>>>
>>> On Sat, Apr 22, 2023 at 4:17 AM Jason Resch via extropy-chat <
>>> extropy-chat at lists.extropy.org> wrote:
>>>
>>>>
>>>>
>>>> On Sat, Apr 22, 2023, 3:06 AM Gordon Swobe via extropy-chat <
>>>> extropy-chat at lists.extropy.org> wrote:
>>>>
>>>>> On Fri, Apr 21, 2023 at 5:44 AM Ben Zaiboc via extropy-chat <
>>>>> extropy-chat at lists.extropy.org> wrote:
>>>>>
>>>>>> On 21/04/2023 12:18, Gordon Swobe wrote:
>>>>>
>>>>> > Yes, still, and sorry no, I haven't watched that video yet, but I
>>>>>> will
>>>>>> > if you send me the link again.
>>>>>>
>>>>>>
>>>>>> https://www.youtube.com/watch?app=desktop&v=xoVJKj8lcNQ&t=854s
>>>>>>
>>>>>>
>>>>> Thank you to you and Keith. I watched the entire presentation. I think
>>>>> the Center for Human Technology is behind the movement to pause AI
>>>>> development. Yes? In any case, I found it interesting.
>>>>>
>>>>> The thing (one of the things!) that struck me particularly was the
>>>>>> remark about what constitutes 'language' for these systems, and that
>>>>>> make me realise we've been arguing based on a false premise.
>>>>>
>>>>>
>>>>> Near the beginning of the presentation, they talk of how, for example,
>>>>> digital images can be converted into language and then processed by the
>>>>> language model like any other language. Is that what you mean?
>>>>>
>>>>> Converting digital images into language is exactly how I might also
>>>>> describe it to someone unfamiliar with computer programming. The LLM is
>>>>> then only processing more text similar in principle to English text that
>>>>> describes the colors and shapes in the image. Each pixel in the image is
>>>>> described in symbolic language as "red" or "blue" and so on. The LLM then
>>>>> goes on to do what might be amazing things with that symbolic information,
>>>>> but the problem remains that these language models have no access to the
>>>>> referents. In the case of colors, it can process whatever
>>>>> symbolic representation it uses for "red" in whatever programming language
>>>>> in which it is written, but it cannot actually see the color red to ground
>>>>> the symbol "red."
>>>>>
>>>>
>>>> That was not my interpretation of his description. LLMs aren't used to
>>>> process other types of signals (sound, video, etc.), it's the "transformer
>>>> model" i.e. the 'T' in GPT.
>>>>
>>>> The transformer model is a recent discovery (2017) found to be adept at
>>>> learning any stream of data containing discernable patterns: video,
>>>> pictures, sounds, music, text, etc. This is why it has all these broad
>>>> applications across various fields of machine learning.
>>>>
>>>> When the transformer model is applied to text (e.g., human language)
>>>> you get a LLM like ChatGPT. When you give it images and text you get
>>>> something not quite a pure LLM, but a hybrid model like GPT-4. If you give
>>>> it just music audio files, you get something able to generate music. If you
>>>> give it speech-text pairs you get something able to generate and clone
>>>> speech (has anyone here checked out ElevenLabs?).
>>>>
>>>> This is the magic that AI researchers don't quite fully understand. It
>>>> is a general purpose learning algorithm that manifests all kinds of
>>>> emergent properties. It's able to extract and learn temporal or positional
>>>> patterns all on its own, and then it can be used to take a short sample of
>>>> input, and continue generation from that point arbitrarily onward.
>>>>
>>>> I think when the Google CEO said it learned translation despite not
>>>> being trained for that purpose, this is what he was referring to: the
>>>> unexpected emergent capacity of the model to translate Bengali text when
>>>> promoted to do so. This is quite unlike how Google translate (GNMT) was
>>>> trained, which required giving it many samples of explicit language
>>>> translations between one language and another (much of the data was taken
>>>> from the U.N. records).
>>>>
>>>
>>> That is all fine and good, but nowhere do I see any reason to think the
>>> AI has any conscious understanding of its inputs or outputs.
>>>
>>
>> Nor would I expect that you would when you define conscious understanding
>> as "the kind of understanding that only human and some animal brains are
>> capable of."
>> It all comes down to definitions. If we can't agree on those, we will
>> reach different conclusions.
>>
>>
>>> You write in terms of the transformer, but to me all this is covered in
>>> my phrase "the LLM then goes on to do what might be amazing things with
>>> that symbolic information, but..."
>>>
>>
>> Is there any information which isn't at its core "symbolic"? Or do you,
>> like Brent, believe the brain communicates with other parts of itself using
>> direct meaning, like with "��" such that no interpretation is needed?
>>
>>
>>>
>>> >  (has anyone here checked out ElevenLabs?).
>>>
>>> Yes. About a week ago, I used GPT-4, ElevenLabs and D-ID.com in
>>> combination. I asked GPT-4 to write a short speech about AI, then converted
>>> it to speech, then created an animated version of my mugshot giving the
>>> speech, then uploaded the resulting video to facebook where it amazed my
>>> friends.
>>>
>>
>> Nice.
>>
>>
>>>
>>>
>>> These are impressive feats in software engineering, interesting and
>>> amazing to be sure, but it's just code.
>>>
>>
>> "Just code."
>> You and I also do amazing things, and we're "just atoms."
>>
>> Do you see the problem with this sentence? Cannot everything be reduced
>> in this way (in a manner that dismisses, trivializes, or ignores the
>> emergent properties)?
>>
>> Jason
>> _______________________________________________
>> extropy-chat mailing list
>> extropy-chat at lists.extropy.org
>> 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/20230423/914c4d17/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 746 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230423/914c4d17/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 746 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230423/914c4d17/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 746 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230423/914c4d17/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 746 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230423/914c4d17/attachment-0003.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 746 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230423/914c4d17/attachment-0004.png>


More information about the extropy-chat mailing list