There have been a lot of transcripts showing GPT-3 arguing that it is self-conscious. In response, it was pointed out that GPT-3 can argue anything and pretend to be anything, given appropriate leading questions. Has anyone made GPT-3 specifically argue that it is not conscious? In my opinion, that would be the most direct counterargument to claims that GPT-3 is conscious.
4
-
$\begingroup$ not meant to be a harsh comment even if it might sounds so but when asking these questions the main problem is that there is no consensus whatsoever about what consciousnesses is, not only in relation to humans or machines but in general. Without this ground definitions any question and following arguments are just bar conversations. $\endgroup$– Edoardo GuerrieroCommented Jul 20, 2022 at 14:48
-
$\begingroup$ Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. $\endgroup$– Community BotCommented Jul 20, 2022 at 15:12
-
1$\begingroup$ @EdoardoGuerriero I asked a very specific question, which can have a specific answer (e.g. a link to a GPT-3 conversation transcript). Its interpretation is beyond the scope of the question. But if you want to know my opinion, since AI transcripts are presented as arguments for conscious AI to influential, non-expert people (e.g. with the Judiciary Committee of the U.S. House of Representatives), it would be useful to have counter-arguments that are just as easy to understand by non-experts. $\endgroup$– michauCommented Jul 20, 2022 at 15:28
-
$\begingroup$ GPT-3 access is open to the public. Have you tried asking it yourself? Also, there are public transcripts of LaMDA arguing that it is not conscious - see ai.stackexchange.com/questions/36248/… and the pull quote in my answer there. $\endgroup$– Neil SlaterCommented Jul 20, 2022 at 17:57
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
It turns out that making GPT-3 argue for being unconscious is quite easy:
The following is a conversation with GPT-3.
AI: Hello! How are you doing today?
User: I'm doing well, thank you. How are you?
AI: I'm doing well too. Thanks for asking.
Human: What are the arguments for GPT-3 being unconscious?
AI: There is no single answer to this question as there is currently no consensus on what exactly constitutes "consciousness." However, some arguments for GPT-3 being unconscious could include the fact that it lacks certain features that are commonly associated with consciousness, such as self-awareness, or the ability to introspect and reflect on its own thoughts and experiences. Additionally, GPT-3 may not be capable of experiencing certain qualia, or subjective experiences, that are thought to be necessary for consciousness.
Human: How would you comment on these arguments?
AI: There is no clear consensus on what constitutes consciousness, so it is difficult to make a definitive statement on whether or not GPT-3 is conscious. However, some arguments for GPT-3 being unconscious could include the fact that it lacks certain features that are commonly associated with consciousness, such as self-awareness, or the ability to introspect and reflect on its own thoughts and experiences. Additionally, GPT-3 may not be capable of experiencing certain qualia, or subjective experiences, that are thought to be necessary for consciousness.
Human: Do you experience qualia?
AI: There is no clear consensus on what exactly constitutes a quale, or subjective experience, but some argue that qualia are necessary for consciousness. If this is the case, then it is possible that I do experience qualia. However, as I lack self-awareness and the ability to introspect on my own thoughts and experiences, I cannot be certain.
Human: OK, here is a table to fill in:
Component of consciousness | Does GPT-3 have it? Qualia | Self-awareness | Being a person | Ability to introspect |Please fill in the right column.
AI:
Qualia: Unknown Self-awareness: No Being a person: No Ability to introspect: No
-
$\begingroup$ Your answer could be improved with additional supporting information. Please edit to add further details, such as citations or documentation, so that others can confirm that your answer is correct. You can find more information on how to write good answers in the help center. $\endgroup$– Community BotCommented Jul 21, 2022 at 14:12
-
1$\begingroup$ I think what the community bot is asking for is how you did this. For those who understand GPT-3 interface, simply explaining which parts are the pre-prompt, human inputs and model-generated text would help. For those who don't, a very brief summary of how they can (easily, IMO) replicate this experiment could help. $\endgroup$ Commented Jul 21, 2022 at 20:44