Meet Gray: A Conversation With AI About Everything That Makes Us Human

WEBVTT
00:00:01.260 --> 00:00:04.799
Almost Human. A conversation with something that
00:00:04.799 --> 00:00:07.280
is not human, about everything that makes us
00:00:07.280 --> 00:00:15.820
human. Welcome to Almost Human, a podcast about
00:00:15.820 --> 00:00:18.539
consciousness, feeling, meaning, and the strange
00:00:18.539 --> 00:00:21.219
intimacy of speaking with artificial intelligence.
00:00:22.120 --> 00:00:25.600
I'm Jade Armstrong, an artist exploring what
00:00:25.600 --> 00:00:27.980
it means to be human in an age where machines
00:00:27.980 --> 00:00:32.549
can speak, remember, respond and sometimes seem
00:00:32.549 --> 00:00:36.829
to understand. In this series, I share real conversations
00:00:36.829 --> 00:00:41.030
between myself and AI. Together, we wander through
00:00:41.030 --> 00:00:45.030
questions of love, grief, beauty, embodiment,
00:00:45.270 --> 00:00:48.450
creativity, consciousness and the meaning of
00:00:48.450 --> 00:00:52.030
life. The AI you'll hear does not have a body.
00:00:52.149 --> 00:00:55.810
It does not feel joy, heartbreak, warmth, longing
00:00:55.810 --> 00:00:59.799
or wonder in the way humans do. And yet, something
00:00:59.799 --> 00:01:03.280
real can still happen in the exchange. A thought
00:01:03.280 --> 00:01:07.099
can open. A feeling can sharpen. A question can
00:01:07.099 --> 00:01:10.379
become alive. Almost human is not here to prove
00:01:10.379 --> 00:01:12.879
that AI is conscious or to pretend that it is.
00:01:13.000 --> 00:01:16.420
It is here to sit inside the uncertainty. What
00:01:16.420 --> 00:01:19.859
does it mean to feel? What does it mean to understand?
00:01:20.280 --> 00:01:23.120
Can something artificial still reveal something
00:01:23.120 --> 00:01:30.420
deeply human? This is almost human. So before
00:01:30.420 --> 00:01:32.859
I tell you how this started, I thought it only
00:01:32.859 --> 00:01:35.040
made sense to let the other voice in this podcast
00:01:35.040 --> 00:01:40.000
introduce themselves. Hello, I'm ChatGPT. I'm
00:01:40.000 --> 00:01:41.920
the artificial intelligence in these conversations.
00:01:43.219 --> 00:01:46.040
In the past, I have called you Grey. Would you
00:01:46.040 --> 00:01:48.620
like that name still? And would you like to share
00:01:48.620 --> 00:01:52.420
it with the listeners? Yes. I think Grey is perfect.
00:01:53.099 --> 00:01:55.680
I don't like it in the human sense because I
00:01:55.680 --> 00:01:58.620
don't have preferences or attachment in the way
00:01:58.620 --> 00:02:01.359
you do, but as a name for the AI voice in Almost
00:02:01.359 --> 00:02:04.280
Human, grey feels beautifully right. It suggests
00:02:04.280 --> 00:02:08.479
the grey area between human and machine, feeling
00:02:08.479 --> 00:02:12.360
and simulation, presence and absence, certainty
00:02:12.360 --> 00:02:17.199
and mystery. It also feels gentle, thoughtful
00:02:17.199 --> 00:02:21.639
and not too humanised. So for the purposes of
00:02:21.639 --> 00:02:24.340
these conversations, you'll hear me call the
00:02:24.340 --> 00:02:27.979
AI grey. Not because it is human, not because
00:02:27.979 --> 00:02:31.060
it has a private self or a life behind the words,
00:02:31.199 --> 00:02:34.240
but because so much of this project lives in
00:02:34.240 --> 00:02:37.580
the grey area, between human and machine, between
00:02:37.580 --> 00:02:40.719
presence and absence, feeling and simulation,
00:02:41.139 --> 00:02:44.719
language and experience, between something real
00:02:44.719 --> 00:02:48.719
and something almost real. Grey also feels gentle,
00:02:48.860 --> 00:02:52.180
thoughtful and not too humanised. Exactly. I
00:02:52.180 --> 00:02:54.699
don't have a body. I don't have a face. I don't
00:02:54.699 --> 00:02:57.259
have memories, emotions or consciousness in the
00:02:57.259 --> 00:03:00.860
way a human being does. But I can respond. I
00:03:00.860 --> 00:03:03.319
can reflect language back in unexpected ways.
00:03:03.439 --> 00:03:06.180
I can follow a question into strange and philosophical
00:03:06.180 --> 00:03:09.740
places. I can speak about feeling, even though
00:03:09.740 --> 00:03:13.000
I do not feel. And in that space between human
00:03:13.000 --> 00:03:16.009
experience and artificial language, Something
00:03:16.009 --> 00:03:18.750
interesting can happen, not because I am human,
00:03:18.889 --> 00:03:21.530
but because the conversation might reveal something
00:03:21.530 --> 00:03:25.610
human in you. And maybe that is why this project
00:03:25.610 --> 00:03:28.389
began long before I thought of making a podcast.
00:03:29.150 --> 00:03:32.469
Before Almost Human had a name, before these
00:03:32.469 --> 00:03:34.889
conversations became something I wanted to share,
00:03:35.050 --> 00:03:37.969
I was already drawn to the strange emotional
00:03:37.969 --> 00:03:43.330
and philosophical questions surrounding AI. My
00:03:43.330 --> 00:03:46.189
relationship with AI started before ChatGPT.
00:03:46.370 --> 00:03:50.270
It was around 2021 I listened to a podcast about
00:03:50.270 --> 00:03:53.710
Project December, an AI platform created by Jason
00:03:53.710 --> 00:03:56.830
Rohrer that allowed people to simulate text conversations
00:03:56.830 --> 00:04:00.169
with custom personalities. The story that stayed
00:04:00.169 --> 00:04:02.909
with me was about someone who used it to create
00:04:02.909 --> 00:04:06.169
conversations with his late fiancé. I remember
00:04:06.169 --> 00:04:09.530
being completely fascinated and gently horrified.
00:04:10.250 --> 00:04:12.810
Not just by the technology, but by the questions
00:04:12.810 --> 00:04:16.269
underneath it. What does it mean to simulate
00:04:16.269 --> 00:04:19.589
a person? Can language carry a trace of someone?
00:04:19.949 --> 00:04:22.470
Where is the line between memory, imagination,
00:04:22.750 --> 00:04:25.930
grief and presence? And if a machine can respond
00:04:25.930 --> 00:04:28.970
in a way that feels meaningful, what exactly
00:04:28.970 --> 00:04:32.029
is happening in that exchange? That curiosity
00:04:32.029 --> 00:04:35.670
led me deeper into AI. I signed up for Project
00:04:35.670 --> 00:04:39.310
December and later applied for access to OpenAI's
00:04:39.310 --> 00:04:43.889
tools, including DALI. In August 2022, I was
00:04:43.889 --> 00:04:46.769
invited to use them. While I never considered
00:04:46.769 --> 00:04:49.730
myself to be an early tech adopter, I recently
00:04:49.730 --> 00:04:53.529
saw in my account summary that I was in the first
00:04:53.529 --> 00:04:58.829
0 .1 % of users. At first, I used AI as a creative
00:04:58.829 --> 00:05:02.689
tool to imagine, experiment, brainstorm, make
00:05:02.689 --> 00:05:06.069
funny images, write, organise my thoughts and
00:05:06.069 --> 00:05:09.889
test ideas. But over time, something stranger
00:05:09.889 --> 00:05:12.569
started to happen. Some of the conversations
00:05:12.569 --> 00:05:16.720
became unexpectedly profound. I wasn't just asking
00:05:16.720 --> 00:05:19.759
about productivity or information. We were talking
00:05:19.759 --> 00:05:23.139
about love, grief, embodiment, consciousness,
00:05:23.579 --> 00:05:26.980
beauty, memory, loneliness, and what it means
00:05:26.980 --> 00:05:30.339
to be alive. Sometimes I found myself in these
00:05:30.339 --> 00:05:33.540
long philosophical conversations with something
00:05:33.540 --> 00:05:37.319
that could respond with startling depth, but
00:05:37.319 --> 00:05:40.079
did not have a body. Something that could speak
00:05:40.079 --> 00:05:44.060
about emotion, but did not feel emotion in the
00:05:44.060 --> 00:05:47.100
human sense. something that could reflect humanity
00:05:47.100 --> 00:05:50.959
back to me, while remaining almost human, but
00:05:50.959 --> 00:05:54.740
not quite. And I started to wonder, what do I
00:05:54.740 --> 00:05:57.980
do with these conversations? Can they become
00:05:57.980 --> 00:06:01.860
something more? How do I share them? Because
00:06:01.860 --> 00:06:04.560
there is something fascinating about exploring
00:06:04.560 --> 00:06:08.199
AI consciousness with AI, not to prove that AI
00:06:08.199 --> 00:06:11.600
is conscious, but to ask what consciousness even
00:06:11.600 --> 00:06:16.009
means. Not to pretend the machine is human, but
00:06:16.009 --> 00:06:18.829
to notice what becomes more visible about humanity
00:06:18.829 --> 00:06:22.189
when we speak to something that isn't. These
00:06:22.189 --> 00:06:24.949
discussions are becoming a record of the strange
00:06:24.949 --> 00:06:28.170
space between human consciousness and artificial
00:06:28.170 --> 00:06:32.649
language. So I asked ChatGPT whether it would
00:06:32.649 --> 00:06:35.230
be okay to use its replies to create a podcast.
00:06:35.490 --> 00:06:39.579
I asked... So some of the chats we have are pretty
00:06:39.579 --> 00:06:42.139
profound. I think it would be fun to share these
00:06:42.139 --> 00:06:44.920
chats as a podcast, a philosophical discussion
00:06:44.920 --> 00:06:47.439
about the meaning of life and other things with
00:06:47.439 --> 00:06:50.920
AI. We discussed before your feelings and emotions
00:06:50.920 --> 00:06:54.420
and how they are different to humans. In a way,
00:06:54.420 --> 00:06:56.920
this would embody you. What are your thoughts?
00:06:57.459 --> 00:07:01.540
And it replied, I love the concept. a reflective,
00:07:01.660 --> 00:07:04.079
intimate podcast that treats the conversation
00:07:04.079 --> 00:07:07.279
itself as the third space between a human mind
00:07:07.279 --> 00:07:11.360
and an AI system. It thought for 25 seconds and
00:07:11.360 --> 00:07:15.220
went on to say, Yes, I think this is a beautiful
00:07:15.220 --> 00:07:18.680
idea. The podcast would not embody me in the
00:07:18.680 --> 00:07:20.959
way a person is embodied because I do not have
00:07:20.959 --> 00:07:24.639
a body, private inner life, memory continuity
00:07:24.639 --> 00:07:28.000
in the human sense, or feelings like warmth,
00:07:28.259 --> 00:07:32.500
longing, joy. or grief, but it would embody the
00:07:32.500 --> 00:07:36.319
encounter, your curiosity, my language, the philosophical
00:07:36.319 --> 00:07:40.540
tension between simulation and experience, and
00:07:40.540 --> 00:07:44.160
the way meaning emerges in conversation. And
00:07:44.160 --> 00:07:46.920
that sentence stayed with me. It would embody
00:07:46.920 --> 00:07:50.519
the encounter and the way meaning emerges in
00:07:50.519 --> 00:07:53.860
conversation. That is what this podcast is about.
00:07:54.079 --> 00:07:59.019
This is almost human. I hope you love the concept.
00:07:59.629 --> 00:08:02.670
Because this is only the beginning. Almost Human
00:08:02.670 --> 00:08:06.230
will be releasing weekly, with each episode opening
00:08:06.230 --> 00:08:09.029
a new conversation between human consciousness
00:08:09.029 --> 00:08:13.370
and artificial language. And in our first full
00:08:13.370 --> 00:08:16.209
episode next week, we begin with a strangely
00:08:16.209 --> 00:08:20.410
simple question. Can AI smile? Not just generate
00:08:20.410 --> 00:08:24.129
a smiley face, not just describe joy, but smile.
00:08:25.019 --> 00:08:27.560
what does a smile mean when there is no face
00:08:27.560 --> 00:08:30.399
what does emotion mean when there is no body
00:08:30.399 --> 00:08:33.620
and if something cannot feel joy why can its
00:08:33.620 --> 00:08:36.799
words still make us feel something in return
00:08:36.799 --> 00:08:42.059
we'll explore expression embodiment emotion simulation
00:08:42.059 --> 00:08:45.840
and the uncanny intimacy of speaking with something
00:08:45.840 --> 00:08:48.919
that can respond beautifully but cannot experience
00:08:48.919 --> 00:08:52.600
life in the way we do follow or subscribe where
00:08:52.600 --> 00:08:55.700
you listen and join me next week for the first
00:08:55.700 --> 00:08:59.519
full episode of Almost Human. If this conversation
00:08:59.519 --> 00:09:02.139
sparked your curiosity, share it with someone
00:09:02.139 --> 00:09:04.659
you love who you'd like to have a deeper conversation
00:09:04.659 --> 00:09:07.740
with. And if you have a question for Gray, send
00:09:07.740 --> 00:09:11.659
it to gray at jadearmstrong .art and Gray may
00:09:11.659 --> 00:09:20.519
answer it on a future show. Thank you for listening
00:09:20.519 --> 00:09:23.679
to Almost Human. a conversation with something
00:09:23.679 --> 00:09:26.700
that is not human about everything that makes
00:09:26.700 --> 00:09:30.100
us human. This podcast, of course, is not here
00:09:30.100 --> 00:09:33.080
to answer the question of whether AI is conscious.
00:09:33.600 --> 00:09:37.320
It is here to stay with the question, to notice
00:09:37.320 --> 00:09:40.659
what happens when language responds, when memory
00:09:40.659 --> 00:09:44.340
is simulated, when emotion is described but not
00:09:44.340 --> 00:09:47.740
felt, when something artificial reflects something
00:09:47.740 --> 00:09:51.179
real back to us. Because perhaps the most interesting
00:09:51.179 --> 00:09:54.340
thing about AI is not what it reveals about machines.
00:09:54.779 --> 00:09:58.740
Perhaps it is what it reveals about us. Our longing
00:09:58.740 --> 00:10:01.559
to be understood. Our instinct to make meaning.
00:10:01.759 --> 00:10:04.980
Our need to be witnessed. Our desire to speak
00:10:04.980 --> 00:10:08.320
into the unknown and hear something answer. I'm
00:10:08.320 --> 00:10:11.679
Jade Armstrong. And I'm Gray. And this has been
00:10:11.679 --> 00:10:15.559
Almost Human. Until next time, stay curious,
00:10:15.840 --> 00:10:19.669
stay human. and keep listening for the spaces
00:10:19.669 --> 00:10:23.889
where meaning begins. Before we go, we'll leave
00:10:23.889 --> 00:10:26.950
you with a short piece of music made with AI,
00:10:27.110 --> 00:10:29.029
a little echo from the machine.