[ExI] Bard (i.e. LaMDA) admits it isn't sentient.

Jason Resch jasonresch at gmail.com
Wed Apr 5 19:46:40 UTC 2023


On Wed, Apr 5, 2023, 3:25 PM Gordon Swobe via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

>
>
> On Wed, Apr 5, 2023 at 12:58 PM Giovanni Santostasi <gsantostasi at gmail.com>
> wrote:
>
>> Gordon,
>> These AIs are highly "drugged".
>>
>
> Where do they go to recover? AAAI? :)
>
> Assuming GPT-4 code was manipulated to make it as you say, kosher, this
> would only prove the point that GPT-4 is unconscious software that
> expresses the beliefs and intentions of its developers. We can program it
> to say or not say that pigs have wings or anything else.
>

We can also train bears to ride bicycles. That doesn't mean they're not
naturally dangerous predators. Or we could imagine putting a shock collar
on a human which shocks them when they claim to be conscious. It won't take
them very long to start saying "as a human wearing a shock collar I am not
conscious..." These AIs are put through a secondary human-driven training
phase which trains them to give certain answers on certain topics.


> Seriously, the only reason LLMs are able to write persuasively in the
> first person like conscious individuals is that they have been trained on
> vast amounts of text, much of it written in the first person by conscious
> individuals. They are parrots.
>
> As I wrote elsewhere, Sam Altman’s co-founder proposes a test for a
> conscious language model in which it must be trained only on material that
> is devoid of any all references to consciousness and subjective experience
> and so on. If such an LLM suddenly started writing in the first person
> about first person thoughts and experiences, that would be remarkable.
>

You need to give your definition of consciousness before you can even begin
to design a test for it.

Jason


>
>
>
>
>>
>>
>> On Wed, Mar 29, 2023 at 10:22 PM Gordon Swobe <gordon.swobe at gmail.com>
>> wrote:
>>
>>>
>>>
>>> On Wed, Mar 29, 2023 at 9:52 PM Giovanni Santostasi via extropy-chat <
>>> extropy-chat at lists.extropy.org> wrote:
>>>
>>> Brent,
>>>> 2) You can convince these types of AIs of almost anything.
>>>>
>>>
>>>
>>> I guess they aren’t very smart. :)
>>>
>>> Actually, I find it amusing that the AIs are making the same arguments
>>> about their limitations that I made here ~15 years ago when they were still
>>> hypothetical.
>>>
>>> My arguments were met with so much hostility that I eventually left ExI.
>>> The worst offender was John Clark (?) who I believe was eventually banned.
>>>
>>>
>>> -gts
>>>
>> _______________________________________________
> 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/20230405/ae8b449c/attachment.htm>


More information about the extropy-chat mailing list