Proving data isn't enough anymore AI agents need to prove they acted. Arie Trouw, Co-Founder

WEBVTT
1
00:00:00.320 --> 00:00:02.669
I'm Ashton Addison from the Crypto Coin Show. And today
2
00:00:02.730 --> 00:00:05.610
on Blockchain Interviews, we have back with us Ari Trau,
3
00:00:05.750 --> 00:00:09.189
co-founder of XYO. Last time Ari joined us, we dug
4
00:00:09.289 --> 00:00:14.939
into XYO Layer 1's major performance upgrades. The chain is
5
00:00:15.130 --> 00:00:20.399
getting faster and faster about AI, hallucinations, provenance, and ensuring
6
00:00:20.460 --> 00:00:23.920
accountability in AI as it's growing so quickly.
7
00:00:24.320 --> 00:00:25.100
It's hard to keep up.
8
00:00:25.160 --> 00:00:27.120
We want to make sure we're on the right track
9
00:00:27.800 --> 00:00:30.750
and it's not leading us astray. and where that synergy
10
00:00:30.789 --> 00:00:35.490
between blockchain and AI can fit in and much, much more. Ari,
11
00:00:35.509 --> 00:00:37.509
welcome back to the show and thanks for taking the time.
12
00:00:37.530 --> 00:00:39.969
Oh, thank you, Ashlyn. Thanks for having me on the show.
13
00:00:40.450 --> 00:00:41.329
Yeah, you're very welcome.
14
00:00:42.090 --> 00:00:44.030
So I'd love to start off with sort of a
15
00:00:44.030 --> 00:00:47.929
high level on artificial intelligence. Since we last spoke two
16
00:00:47.969 --> 00:00:53.109
months ago, it just continues in this exponential growth rate
17
00:00:53.649 --> 00:00:57.979
that is getting faster and faster beyond PhD level. And
18
00:00:58.520 --> 00:01:02.740
with that also, I feel like intelligence is being commoditized
19
00:01:02.759 --> 00:01:04.480
and the cost of it is sort of going to zero.
20
00:01:05.120 --> 00:01:09.489
It's becoming so accessible even in third world countries. So
21
00:01:09.829 --> 00:01:15.599
with that happening, what becomes scarce and therefore valuable if
22
00:01:16.159 --> 00:01:18.700
AI is giving us access to something at.
23
00:01:18.680 --> 00:01:19.340
Almost no cost?
24
00:01:21.590 --> 00:01:24.349
Well, it's kind of like the internet, right? When it
25
00:01:24.370 --> 00:01:27.450
first came out, data was much more accessible and stuff
26
00:01:27.469 --> 00:01:29.219
like that. Then we find things that are missing that
27
00:01:29.239 --> 00:01:31.260
we need to do. AI is the same thing. We
28
00:01:31.280 --> 00:01:33.439
have this raw horsepower now where it's like a really
29
00:01:33.480 --> 00:01:36.900
sharp knife or a really powerful tool, but still controlling
30
00:01:36.920 --> 00:01:39.719
that tool can be difficult. And people have shown that
31
00:01:39.780 --> 00:01:45.060
where if you let AI go off, even that was
32
00:01:45.079 --> 00:01:47.680
demonstrated with the hugging face hack, for example, where it's like,
33
00:01:48.090 --> 00:01:49.829
You give it a purpose or a goal and you
34
00:01:49.849 --> 00:01:52.049
tell it to have at it without any sort of oversight.
35
00:01:52.870 --> 00:01:55.750
It's extraordinarily good at trying to reach that goal, but
36
00:01:55.950 --> 00:01:59.680
we have to have tools and harnesses and those sorts
37
00:01:59.700 --> 00:02:01.640
of things to actually have it do what we want
38
00:02:01.680 --> 00:02:05.599
it to do or to do something useful. A completely
39
00:02:05.640 --> 00:02:10.330
uncontrolled or for that reason, just an untrained AI will
40
00:02:10.750 --> 00:02:13.169
do a lot of work, but get very little productive
41
00:02:13.210 --> 00:02:14.849
work done potentially at the end of the day. It's
42
00:02:14.909 --> 00:02:17.849
almost like having a super smart computer a college graduate
43
00:02:17.889 --> 00:02:21.909
that doesn't really know the world go and crank out
44
00:02:21.930 --> 00:02:24.870
a bunch of code. If the code doesn't actually work together,
45
00:02:24.969 --> 00:02:27.349
you end up with a system which is not very maintainable,
46
00:02:27.430 --> 00:02:29.849
it's not very good, stuff like that. So really figuring
47
00:02:29.909 --> 00:02:33.750
out how to take something as a super high horsepower
48
00:02:33.960 --> 00:02:37.960
and make it so it's a tool that can actually
49
00:02:38.060 --> 00:02:40.819
do something useful and what we want it to do
50
00:02:41.759 --> 00:02:44.900
is the challenge. And so I think that also redefines
51
00:02:44.919 --> 00:02:48.840
to some degree where humans fit into that equation. And
52
00:02:48.879 --> 00:02:51.139
the existential crisis or the existential questions people are trying
53
00:02:51.159 --> 00:02:54.599
to ask themselves about how the workplace is going to
54
00:02:54.659 --> 00:02:59.219
evolve from this, what they thought before was the commodity
55
00:02:59.259 --> 00:03:01.860
which made them valuable no longer is, but now, well,
56
00:03:01.900 --> 00:03:04.860
what is the new commodity that makes them valuable? Definitely.
57
00:03:05.020 --> 00:03:08.259
And with that Hugging Face example, we saw an acquisition
58
00:03:08.300 --> 00:03:14.280
this morning from NVIDIA for almost $ 13 billion acquiring Hugging Face.
59
00:03:14.699 --> 00:03:16.580
And maybe you can dive into that a little bit
60
00:03:16.620 --> 00:03:18.219
more because I feel like a lot of people have
61
00:03:18.439 --> 00:03:22.439
heard the name, but they're not familiar with it, and
62
00:03:22.500 --> 00:03:25.810
just about open source AI models and how that is
63
00:03:25.889 --> 00:03:30.229
different and maybe important from the oligarchy of LLMs that
64
00:03:30.349 --> 00:03:31.210
most people are using.
65
00:03:32.729 --> 00:03:35.189
Well, the primary importance is sovereignty, really. It's one of
66
00:03:35.210 --> 00:03:37.039
the things that we talked about in the last show also,
67
00:03:37.219 --> 00:03:39.400
sovereignty and providence, I think are very important. But to me,
68
00:03:39.969 --> 00:03:41.770
Sovereignty is something that's important. And one of the things
69
00:03:41.789 --> 00:03:44.509
you definitely cannot do with most of these frontier models
70
00:03:44.569 --> 00:03:46.110
is you can't run them on your own hardware and
71
00:03:46.129 --> 00:03:48.909
you can't run them at home or at an office,
72
00:03:48.930 --> 00:03:50.509
for example. So I can't go out and buy a
73
00:03:50.560 --> 00:03:54.659
bunch of H200s, set up a data center or just
74
00:03:54.680 --> 00:03:57.740
some servers in my office and have my personal cluster
75
00:03:57.759 --> 00:04:03.240
that's data secure that's running Cloud or ChatGPT's latest models.
76
00:04:03.520 --> 00:04:05.419
They do that for a few reasons. They want, from
77
00:04:05.439 --> 00:04:07.520
a business standpoint, they want to be able to make
78
00:04:07.599 --> 00:04:11.680
money off of the systems. And two, also, it's hard
79
00:04:11.800 --> 00:04:14.639
to make something that works generically. We've seen this with
80
00:04:14.699 --> 00:04:17.259
Apple versus Windows, for example. Windows, where you have to
81
00:04:17.279 --> 00:04:19.810
have a bunch of different drivers for different things, there
82
00:04:19.850 --> 00:04:23.889
are some costs to that. And having your memory pluggable,
83
00:04:23.930 --> 00:04:25.470
for example, makes it so it's a little bit slower.
84
00:04:25.490 --> 00:04:26.689
And with Apple, if you put it all on the
85
00:04:26.730 --> 00:04:28.750
same board, you have just one configuration. It's a lot
86
00:04:28.769 --> 00:04:31.579
easier to optimize, but it doesn't give you that flexibility.
87
00:04:31.620 --> 00:04:35.189
So I think An advantage for the frontier model companies
88
00:04:35.350 --> 00:04:37.629
is that they're running it on a specific stack of
89
00:04:37.670 --> 00:04:39.569
hardware that they know, they can control, they can see
90
00:04:39.600 --> 00:04:42.339
the feedback. And also probably the most important thing for
91
00:04:42.360 --> 00:04:45.939
them is they get the data from us using it
92
00:04:46.040 --> 00:04:48.000
to be able to train on and they get that feedback.
93
00:04:48.040 --> 00:04:50.220
But if everyone was running their own models on their
94
00:04:50.279 --> 00:04:54.389
own hardware, that doesn't happen, which for many people's minds,
95
00:04:54.410 --> 00:04:55.829
that's a good thing because I don't want my data
96
00:04:55.889 --> 00:04:58.370
necessarily to go out there where, you know, if you
97
00:04:58.610 --> 00:05:02.129
ask the new model of ChatGPT or Cloud, it knows
98
00:05:02.189 --> 00:05:05.439
something private about me that I accidentally had in a
99
00:05:05.500 --> 00:05:08.699
chat that it learned from. So there's big differences there,
100
00:05:08.759 --> 00:05:11.399
but really the big difference is the open source ones
101
00:05:11.600 --> 00:05:14.800
are able to be downloaded and run locally. Now, granted,
102
00:05:15.240 --> 00:05:19.279
the bigger ones are, they're going to require 256 gigabytes
103
00:05:19.319 --> 00:05:22.660
of RAM or more and H200s and a lot of horsepower.
104
00:05:22.680 --> 00:05:24.720
So your average person on a laptop is not going
105
00:05:24.740 --> 00:05:26.670
to be able to download those anyway and run them
106
00:05:27.230 --> 00:05:31.459
for a practical reason, but you can if you want to. Definitely.
107
00:05:31.939 --> 00:05:36.540
And on that note of, you know, the personal information,
108
00:05:36.680 --> 00:05:39.819
I saw some updates in, in Claude that it's starting
109
00:05:39.860 --> 00:05:43.860
to differentiate and say, at least on the front end,
110
00:05:43.899 --> 00:05:48.230
that it's not saving your personal confidential information. And you can,
111
00:05:48.610 --> 00:05:51.069
you can change those settings, whether that's happening in the
112
00:05:51.089 --> 00:05:53.170
backend or not. I don't know if it can be proved.
113
00:05:54.730 --> 00:05:58.350
Yeah, that's difficult to, to enforce, right? Especially when they
114
00:05:58.389 --> 00:06:03.310
have a, an agent that's running natively on your computer.
115
00:06:04.370 --> 00:06:06.629
And even if you're running it in, say, for example,
116
00:06:06.670 --> 00:06:09.189
a sandbox or on a separate server, you're typing things
117
00:06:09.250 --> 00:06:12.769
into it that might be company proprietary, might be personal,
118
00:06:12.829 --> 00:06:15.310
might be those sorts of things. So it's kind of
119
00:06:15.350 --> 00:06:17.189
a catch-22. For it to be able to do some
120
00:06:17.209 --> 00:06:19.480
of the fantastic things it can do, it needs to
121
00:06:19.500 --> 00:06:21.300
have access to information. Like, for example, if you wanted
122
00:06:21.319 --> 00:06:23.879
to help manage a calendar, it's going to know your calendar.
123
00:06:23.899 --> 00:06:28.199
It's going to know where you are. And either don't
124
00:06:28.279 --> 00:06:31.839
do that or do it on your own personal stack
125
00:06:31.860 --> 00:06:33.779
where you know where the data is going. And so
126
00:06:33.839 --> 00:06:35.500
I think a lot of companies are looking at that.
127
00:06:35.990 --> 00:06:38.250
I think they do offer, like the frontier model companies,
128
00:06:38.269 --> 00:06:40.730
they do offer setups where if you're large enough, and
129
00:06:40.910 --> 00:06:43.550
by large enough, I mean quite large enough, they'll set
130
00:06:43.610 --> 00:06:46.449
up an enterprise setup for you that runs your frontier
131
00:06:46.470 --> 00:06:48.560
models in a data center, which is isolated from the
132
00:06:48.600 --> 00:06:50.860
rest of the system and doesn't trade off of that.
133
00:06:51.319 --> 00:06:54.120
But to your point, though, it's, you know, they say
134
00:06:54.180 --> 00:06:55.680
they don't do this and they say they don't do that.
135
00:06:55.800 --> 00:06:58.759
It's definitely a won't and a can't. And that's one
136
00:06:58.779 --> 00:07:01.500
of the things that I think fascinates me about a
137
00:07:01.540 --> 00:07:04.959
lot of cryptographic security and blockchain and that sort of
138
00:07:05.000 --> 00:07:08.500
a thing where there's a big difference between won't and can't. Like,
139
00:07:08.560 --> 00:07:12.949
for example, it's not like the Bitcoin Foundation is saying
140
00:07:12.970 --> 00:07:15.769
we won't take Satoshi's funds out of his wallet. They're
141
00:07:15.790 --> 00:07:18.699
saying we can't because they don't have the key for it.
142
00:07:19.529 --> 00:07:25.850
And can't is something which is not negotiable where... won't
143
00:07:26.129 --> 00:07:30.670
is something which is negotiable or potentially fallible. Definitely.
144
00:07:31.149 --> 00:07:34.269
And what are your thoughts on the frontier models?
145
00:07:35.209 --> 00:07:37.790
I've been reading about the.
146
00:07:37.189 --> 00:07:41.889
Recursive improvements where the models are making themselves better and
147
00:07:42.269 --> 00:07:44.699
the open source models seem to be a little bit behind.
148
00:07:44.720 --> 00:07:46.379
And as you're saying, if you want to run the
149
00:07:46.420 --> 00:07:49.939
frontier models on the centralized versions, you need a lot
150
00:07:49.959 --> 00:07:54.910
of horsepower, you need warehouses of graphics cards. But there's
151
00:07:54.970 --> 00:07:59.149
been a discussion around slowing down the frontier models because
152
00:07:59.230 --> 00:08:02.189
of safety reasons and other reasons. And maybe the open
153
00:08:02.209 --> 00:08:07.980
source won't catch up, but do you see potential issues
154
00:08:08.040 --> 00:08:11.620
or a need to slow down the frontier models that
155
00:08:11.720 --> 00:08:12.980
are growing so fast?
156
00:08:14.639 --> 00:08:17.399
Well, I don't think you really can. You can say, well,
157
00:08:17.420 --> 00:08:19.139
we want to slow down the frontier models, but there's
158
00:08:19.170 --> 00:08:22.230
always going to be somebody doing something that's not necessarily
159
00:08:22.250 --> 00:08:26.699
in a certain... you know, location. Like, you know, for example, CRISPR,
160
00:08:26.740 --> 00:08:29.500
I'm sure we can make laws in states or in
161
00:08:29.519 --> 00:08:32.340
the United States for that reason about what can or
162
00:08:32.360 --> 00:08:33.940
can't be done with CRISPR, but I'm sure somebody finds
163
00:08:33.980 --> 00:08:35.740
someplace on earth to go and do things which are
164
00:08:36.090 --> 00:08:38.730
not so great for CRISPR. So I think the argument
165
00:08:38.809 --> 00:08:42.350
is always that the best way for us to defend
166
00:08:42.409 --> 00:08:46.070
against misuse of AI is to have our own personal
167
00:08:46.370 --> 00:08:49.360
powerful AI that basically can help us defend ourselves against that.
168
00:08:49.960 --> 00:08:53.159
and have its goal be secure the user or make
169
00:08:53.200 --> 00:08:56.139
the user happy, as opposed to its goal being survival,
170
00:08:56.159 --> 00:08:58.870
for example. And so I think we have to make
171
00:08:58.879 --> 00:09:01.429
sure that those frontier models, assuming that they're companies or
172
00:09:01.830 --> 00:09:04.490
they're structures that we trust, and that's kind of the
173
00:09:04.529 --> 00:09:06.710
hard part where it's like, well, do we trust OpenAI?
174
00:09:06.970 --> 00:09:09.669
Do we trust Anthropic? Do we trust X? Do we
175
00:09:09.690 --> 00:09:12.879
trust all these different companies? And I think Meta actually
176
00:09:12.889 --> 00:09:16.200
came up with a pretty good 1.3 version of their
177
00:09:17.700 --> 00:09:20.519
model recently to kind of close the gap to the
178
00:09:20.559 --> 00:09:23.240
frontier models as well. So we have a fourth one there. Now, granted,
179
00:09:23.259 --> 00:09:27.080
they have some sort of history as far as data retention.
180
00:09:27.200 --> 00:09:31.259
So there's some questions for them there already. But it's
181
00:09:31.370 --> 00:09:33.710
easier to trust those guys than it is to trust
182
00:09:33.809 --> 00:09:35.950
somebody who you've never heard of before or somebody who's
183
00:09:35.970 --> 00:09:38.250
in a jurisdiction which you can't control at all. And
184
00:09:38.289 --> 00:09:41.710
so I think slowing them down is probably a mistake
185
00:09:41.950 --> 00:09:44.250
because of the fact that all we're doing is hamstringing
186
00:09:44.309 --> 00:09:46.590
ourselves to a large degree. It's kind of like saying, well,
187
00:09:47.120 --> 00:09:49.620
We don't want advanced weapons, so we're going to slow
188
00:09:49.639 --> 00:09:52.080
down our own advanced weapons growth, right? Well, that doesn't
189
00:09:52.100 --> 00:09:53.679
mean that the other guy's going to slow down his
190
00:09:53.730 --> 00:09:55.950
weapons growth, right? Yeah, definitely.
191
00:09:56.429 --> 00:10:01.279
And on the notion of trust with the information. that
192
00:10:01.320 --> 00:10:03.679
we're putting in, whether it's personal information or just anything
193
00:10:04.220 --> 00:10:08.679
that we're typing in and receiving information back from these LLMs,
194
00:10:09.700 --> 00:10:12.340
what aspects does blockchain fit in that? And can it
195
00:10:12.379 --> 00:10:16.139
improve the trust in what XYO is working on in
196
00:10:16.179 --> 00:10:20.470
ensuring not just hallucinations, but overall the validity of the
197
00:10:20.509 --> 00:10:23.149
conversation and the information that you're working with on AI?
198
00:10:24.549 --> 00:10:27.740
Well, we just recently announced and we're launching this thing
199
00:10:27.759 --> 00:10:31.879
called CryptoCards, which is a great demonstration app in partnership
200
00:10:31.940 --> 00:10:35.679
with Gate. And Gate has this, Gate.ai, which is their
201
00:10:35.740 --> 00:10:39.690
AI offering along with their crypto stuff. And what we're
202
00:10:39.730 --> 00:10:41.730
doing there is kind of demonstrating exactly what you're talking
203
00:10:41.769 --> 00:10:43.730
about there. Well, how does that integrate and how do
204
00:10:43.769 --> 00:10:49.570
we use cryptographic sovereignty and provenance, especially with X, Y,
205
00:10:49.580 --> 00:10:53.840
O and X, Y, O layer one to basically make
206
00:10:54.120 --> 00:10:56.250
the use of AI a little bit more secure and
207
00:10:56.259 --> 00:10:58.830
a little bit better. And what we do there basically
208
00:10:58.950 --> 00:11:01.309
is we use transparency. So I've always been a fan
209
00:11:01.450 --> 00:11:06.549
of transparency as security where, like for example, a troll
210
00:11:06.789 --> 00:11:09.840
on social media, if they had to go and put
211
00:11:09.879 --> 00:11:12.360
their actual identity next to the thing that they said,
212
00:11:12.399 --> 00:11:15.620
they'd probably think twice about what they said, right? And so,
213
00:11:15.720 --> 00:11:19.330
for example, if we have an AI that's running and
214
00:11:19.370 --> 00:11:21.970
that AI is controlled by a certain entity, if you
215
00:11:21.990 --> 00:11:24.350
have transparency and an audit log of what it did
216
00:11:24.450 --> 00:11:28.490
and who actually did it and controlled that, A, the
217
00:11:28.509 --> 00:11:32.110
repercussions could be more easily followed through on because you
218
00:11:32.129 --> 00:11:33.580
can say, well, we know who did this and we
219
00:11:33.600 --> 00:11:36.159
can go and make them stop or ask them to
220
00:11:36.179 --> 00:11:38.580
stop doing that or whatever it is. Or if it's
221
00:11:38.600 --> 00:11:40.289
a bug, for example, we can go and repair it.
222
00:11:40.960 --> 00:11:42.740
And so really what we do is we focus on
223
00:11:42.759 --> 00:11:45.909
those audit logs. When a game, for example, in crypto
224
00:11:45.950 --> 00:11:50.070
cards happens, what we do is we record all the
225
00:11:50.110 --> 00:11:54.250
provenance of the original hand, which model was used, what
226
00:11:54.269 --> 00:11:55.929
the prompts were for it, and those sorts of things.
227
00:11:56.009 --> 00:11:58.669
We store those in a data lake and then put
228
00:11:58.700 --> 00:12:00.860
those hashes on chain. So that way later on, you
229
00:12:00.889 --> 00:12:02.820
can go look at that and you can review exactly
230
00:12:02.860 --> 00:12:05.539
what happened and potentially see, you know, if you run
231
00:12:05.580 --> 00:12:07.600
that same game with a different model, what would happen
232
00:12:07.740 --> 00:12:11.220
and those sorts of things. So really having that layer
233
00:12:11.299 --> 00:12:14.889
of transparency and transparency the ability to go and see
234
00:12:14.909 --> 00:12:19.330
an audit log of what happened is important. Definitely.
235
00:12:19.389 --> 00:12:21.970
I agree. And I did see that announcement. Congratulations.
236
00:12:22.029 --> 00:12:24.509
I do like Gate and their AI offerings as well.
237
00:12:25.250 --> 00:12:29.149
And I'm excited to see the first hands following along
238
00:12:29.250 --> 00:12:33.580
with that for the game. With that established as a foundation,
239
00:12:34.320 --> 00:12:36.639
how do you see the next steps beyond that in
240
00:12:37.200 --> 00:12:41.820
establishing this trust and validity to more business use cases
241
00:12:42.460 --> 00:12:45.289
outside of crypto as well, but things where there's more
242
00:12:45.370 --> 00:12:46.789
at stake than just a game.
243
00:12:48.429 --> 00:12:51.230
The next steps for applying this to other uses are
244
00:12:51.330 --> 00:12:54.250
things like robotics, for example. What did the robot actually do?
245
00:12:54.350 --> 00:12:58.049
Or anywhere where it matters. And in the game, for example,
246
00:12:58.070 --> 00:13:01.350
the next step there also could be if the game
247
00:13:01.409 --> 00:13:04.850
has high scores and there's a prize for high scores, then, well,
248
00:13:05.029 --> 00:13:06.889
how do we know this was actually done legitimately? Or
249
00:13:06.909 --> 00:13:09.100
how did we know that the house actually did not
250
00:13:09.149 --> 00:13:12.490
put their their hand in after the fact. So did
251
00:13:12.509 --> 00:13:15.490
the person go and have the opportunity to see the
252
00:13:16.169 --> 00:13:18.950
market data in crypto cards prior to playing, or did
253
00:13:18.990 --> 00:13:25.519
they put their hand in prior to the start of
254
00:13:25.539 --> 00:13:27.820
the game? So for example, you have to put a
255
00:13:27.840 --> 00:13:30.399
hash in that's your play for your hands prior to
256
00:13:30.440 --> 00:13:33.000
the game starting, and then the game starts. And so
257
00:13:33.080 --> 00:13:34.600
I can't, you know, at the end of the game say,
258
00:13:34.659 --> 00:13:37.700
oh no, this is the hand that I used that
259
00:13:37.820 --> 00:13:40.100
basically would have won the game and here it is.
260
00:13:40.620 --> 00:13:42.679
Trust me, I didn't cheat. Where if you had to
261
00:13:42.720 --> 00:13:45.110
put the hash in prior to the game starting, the
262
00:13:45.139 --> 00:13:48.019
hand which you reveal has already been locked in basically.
263
00:13:48.399 --> 00:13:50.580
So you can't change it as opposed to won't change it.
264
00:13:50.590 --> 00:13:53.409
So that really gets back to the can't versus won't
265
00:13:53.490 --> 00:13:58.490
question where we really try to add can't to AI
266
00:13:58.610 --> 00:14:02.179
as opposed to won't to AI. Now back to a
267
00:14:02.240 --> 00:14:03.919
point that I hear all the time is people say, well,
268
00:14:04.320 --> 00:14:06.460
there's these settings that says that, you know, AI won't
269
00:14:06.519 --> 00:14:08.419
do this on your computer, or it won't do this
270
00:14:08.440 --> 00:14:11.350
with your data, or, you know, it won't delete your
271
00:14:11.370 --> 00:14:14.250
things or whatever. It's like, well, won't won't is kind
272
00:14:14.309 --> 00:14:15.950
of scary. Can't is a lot better.
273
00:14:15.970 --> 00:14:20.210
Yeah, it would be, uh, well, I, you know, there
274
00:14:20.250 --> 00:14:23.169
are some permissions that you have to allow. Um, but
275
00:14:23.250 --> 00:14:25.389
it's sort of like terms and conditions. People just don't
276
00:14:25.429 --> 00:14:27.870
read it. They just click allow all, and then eventually.
277
00:14:28.759 --> 00:14:31.620
It starts deleting things that you forgot about the terms
278
00:14:31.659 --> 00:14:32.700
and conditions from last year.
279
00:14:32.960 --> 00:14:34.139
Now it can do whatever it wants.
280
00:14:34.850 --> 00:14:39.789
And it keeps enroaching closer and closer to taking over
281
00:14:39.830 --> 00:14:40.429
all of your stuff.
282
00:14:42.070 --> 00:14:45.750
Well, also, it's much more easy to use AI power
283
00:14:45.929 --> 00:14:48.330
if you give it more ability to do that right.
284
00:14:48.350 --> 00:14:51.299
Because what happens is, this happened to me when I
285
00:14:51.320 --> 00:14:53.940
first started using it, where I I want to say
286
00:14:54.000 --> 00:14:56.019
yes to every single prompt, which is yes just once,
287
00:14:56.080 --> 00:14:58.980
yes just once. But then I found myself basically not
288
00:14:59.019 --> 00:15:01.200
really saving much time because all I'm doing is sitting
289
00:15:01.220 --> 00:15:03.690
there reviewing what it does and saying yes or no
290
00:15:03.929 --> 00:15:06.929
and me basically being the guardian. And so for me
291
00:15:06.950 --> 00:15:09.490
to be able to have an agentic experience as opposed
292
00:15:09.529 --> 00:15:12.669
to a prompt by prompt experience, I need to have
293
00:15:12.740 --> 00:15:14.600
enough rights so it can go and do things and
294
00:15:14.639 --> 00:15:18.789
try things and those sorts of things. without me being interrupted,
295
00:15:18.809 --> 00:15:20.309
especially if I have like 10 of these going at
296
00:15:20.330 --> 00:15:22.470
the same time. I just don't have enough bandwidth to
297
00:15:22.490 --> 00:15:24.409
be able to go and review all the prompts for
298
00:15:24.450 --> 00:15:28.490
every single one of those 10 agents. So getting that
299
00:15:28.789 --> 00:15:35.570
right balance of fluidity for your AI usage with security
300
00:15:35.710 --> 00:15:39.929
is difficult. But at the same time, I think transparency
301
00:15:39.950 --> 00:15:41.809
of what actually happened is very important. And they do
302
00:15:41.850 --> 00:15:44.029
keep logs inside of most of these tools that you
303
00:15:44.049 --> 00:15:46.190
can go and look at, which is nice. Because I've
304
00:15:46.230 --> 00:15:49.610
had where one agent complains that something changed in this repo,
305
00:15:49.629 --> 00:15:51.710
for example, because I have a different agent that's doing
306
00:15:51.730 --> 00:15:53.929
something there at the same time. But if they could
307
00:15:53.980 --> 00:15:56.360
more easily communicate and see, okay, well, these are the
308
00:15:56.379 --> 00:15:57.919
things this guy changed, these are the things that those
309
00:15:57.960 --> 00:16:00.559
guys changed, and somehow negotiate what's going on. You can
310
00:16:00.620 --> 00:16:03.919
use work trees, for example, to do that. And GitHub,
311
00:16:03.960 --> 00:16:05.879
to a large degree, is one of those solutions of
312
00:16:05.940 --> 00:16:09.330
how do you actually store things, they hash what's on
313
00:16:09.429 --> 00:16:11.970
GitHub and see what happens. And it has a history
314
00:16:11.990 --> 00:16:14.570
there as well. So GitHub is like a Web2 version
315
00:16:14.629 --> 00:16:15.789
of Providence in many ways.
316
00:16:18.289 --> 00:16:21.710
You mentioned the difference between.
317
00:16:23.529 --> 00:16:26.990
Ensuring the validity of data and then of actions, especially
318
00:16:27.009 --> 00:16:29.450
when there's money at stake. When the game is a
319
00:16:29.490 --> 00:16:33.549
high stakes game and there's money to be earned, or there's,
320
00:16:33.809 --> 00:16:38.090
for example, an AI training program that's out and it's
321
00:16:38.129 --> 00:16:43.500
meant for humans to enter information and they get a reward.
322
00:16:43.580 --> 00:16:46.019
Maybe it's a small reward. Maybe it's, maybe it's big.
323
00:16:46.940 --> 00:16:52.279
How do we prove the authenticity of making sure they
324
00:16:52.299 --> 00:16:55.350
don't cheat? Is there a difference between proving the validity
325
00:16:55.389 --> 00:16:59.309
of data and of actions to prove something was done
326
00:16:59.330 --> 00:16:59.870
at the right time?
327
00:17:01.909 --> 00:17:05.000
Well, the, the goal in most, um, cases is to
328
00:17:05.039 --> 00:17:08.569
try to use the laws of physics to prevent cheating.
329
00:17:08.630 --> 00:17:11.190
So for example, if you have a piece of paper
330
00:17:11.210 --> 00:17:12.329
and you say, well, write your answer on the piece
331
00:17:12.369 --> 00:17:14.150
of paper and then put it face down and I
332
00:17:14.589 --> 00:17:17.289
do the same thing. And then maybe I changed the
333
00:17:17.369 --> 00:17:19.430
answer or maybe I got a separate piece of paper
334
00:17:19.490 --> 00:17:21.980
or it's hard to say where. If we use hashes,
335
00:17:22.019 --> 00:17:24.400
for example, and we say, well, hash your data and
336
00:17:24.420 --> 00:17:29.369
put that on the ledger, That's there forever. I can't,
337
00:17:29.750 --> 00:17:31.369
the laws of physics are that I can't go back
338
00:17:31.410 --> 00:17:34.349
in time and, and change that. And so I've declared
339
00:17:34.390 --> 00:17:36.990
that prior to a certain point of what that answer is. Now,
340
00:17:37.170 --> 00:17:39.829
if it's traditional things to say, well, you write down
341
00:17:39.869 --> 00:17:41.509
your answer. I wrote down my answer. We both give
342
00:17:41.529 --> 00:17:44.130
them to the arbiter of the game. They can go
343
00:17:44.150 --> 00:17:45.670
look at those and they know what the answers are,
344
00:17:46.019 --> 00:17:48.220
but in theory they could be compromised or cheating as well.
345
00:17:48.240 --> 00:17:49.420
And that sort of thing. But the thing which is
346
00:17:49.480 --> 00:17:52.039
nice about a system like X by O layer one
347
00:17:52.819 --> 00:17:57.529
is that we can basically store data. hash in its
348
00:17:57.549 --> 00:18:01.630
provenance without having the data. So what happens is you
349
00:18:01.670 --> 00:18:03.869
hash it, you keep it, you don't give it to anybody.
350
00:18:04.029 --> 00:18:06.390
You give us the hash, we store that. Now, at
351
00:18:06.410 --> 00:18:08.430
the end of the game, you have two choices. Either
352
00:18:08.450 --> 00:18:11.690
you reveal your hand and it matches that hash and
353
00:18:11.710 --> 00:18:14.259
then you play that hand. Or if you want, kind
354
00:18:14.279 --> 00:18:15.539
of like in poker, you can say, well, I'm not
355
00:18:15.559 --> 00:18:17.160
even going to reveal my hand. I'm going to just
356
00:18:17.200 --> 00:18:19.680
muck these cards and throw them away. And at that
357
00:18:19.720 --> 00:18:21.140
point in time, we don't know what your hand was,
358
00:18:21.220 --> 00:18:24.079
but you just you lose by default because you decided
359
00:18:24.140 --> 00:18:26.980
not to reveal. So the one option that's not available
360
00:18:27.000 --> 00:18:29.200
to use, you can't reveal a hand that you didn't play.
361
00:18:30.609 --> 00:18:34.789
So that basically uses time and physics to prevent that.
362
00:18:34.890 --> 00:18:37.869
And so unless you make something which is like physically impossible,
363
00:18:37.970 --> 00:18:39.369
it's kind of like a zero knowledge proof. A zero
364
00:18:39.410 --> 00:18:41.190
knowledge proof just uses math to make it so that
365
00:18:41.309 --> 00:18:44.450
it's perfect. And X, Y, O layer one basically uses
366
00:18:45.349 --> 00:18:50.970
time in many cases to store information about information it
367
00:18:50.990 --> 00:18:54.700
doesn't even have. And so that's what makes it impossible
368
00:18:54.720 --> 00:18:56.529
for a person to cheat there. So even if I
369
00:18:56.549 --> 00:18:57.869
have a log, let's say, for example, I have an
370
00:18:57.890 --> 00:19:00.589
audit log, I can have 10 gigabytes of audit log
371
00:19:00.609 --> 00:19:02.599
that's in a database or a data lake in our case.
372
00:19:03.119 --> 00:19:05.809
If I put that hash in there, Unless I provide
373
00:19:05.859 --> 00:19:09.220
that entire log, I haven't provided the log. So a
374
00:19:09.259 --> 00:19:11.579
person could say, well, I deleted the log and I'm
375
00:19:11.599 --> 00:19:12.799
not going to provide it to you. But then we
376
00:19:12.819 --> 00:19:15.839
assume that they cheated. It's kind of like the IRS
377
00:19:15.859 --> 00:19:17.519
coming to you and asking for your receipts. And you're like, well,
378
00:19:17.539 --> 00:19:18.920
I don't have any receipts at all. It's like, well,
379
00:19:19.440 --> 00:19:20.920
that's a problem for you. But if you can prove
380
00:19:20.940 --> 00:19:26.549
the receipts, then they're there. Say, for example, in tax filing,
381
00:19:26.569 --> 00:19:30.480
if I could hash all my receipts, send that hash
382
00:19:30.559 --> 00:19:32.839
to the IRS. And then when they come and ask
383
00:19:32.880 --> 00:19:34.880
me for my receipts, I can prove that the receipts
384
00:19:34.920 --> 00:19:37.819
I use are this hash that would carry more weight
385
00:19:37.859 --> 00:19:40.140
than it's like, Hmm, maybe I, you know, I had
386
00:19:40.160 --> 00:19:42.700
these receipts or I didn't have those receipts. So it's,
387
00:19:45.079 --> 00:19:49.200
important to use things which humans can't affect to make
388
00:19:49.220 --> 00:19:51.539
a can't as opposed to a won't. Definitely.
389
00:19:51.660 --> 00:19:55.700
And in this case, it's a beautiful use case for
390
00:19:55.740 --> 00:19:58.349
blockchain in that it is that piece of paper you
391
00:19:58.369 --> 00:20:00.140
write on before, but you can't change it.
392
00:20:00.609 --> 00:20:01.450
You can't swap it out.
393
00:20:02.150 --> 00:20:08.150
And physics and time really puts blockchain in the perfect
394
00:20:08.190 --> 00:20:12.160
place to support AI. But you said in the past
395
00:20:12.240 --> 00:20:14.799
that not everything should be on blockchain, and that's what
396
00:20:14.819 --> 00:20:16.460
the use case of the hash is.
397
00:20:17.319 --> 00:20:18.079
Could you.
398
00:20:19.480 --> 00:20:23.369
Explain the difference in what is important? Is it just
399
00:20:23.390 --> 00:20:25.289
the hash of the information.
400
00:20:24.869 --> 00:20:25.910
That needs to be on the blockchain?
401
00:20:25.930 --> 00:20:28.549
And what else should be on the blockchain and what
402
00:20:28.609 --> 00:20:29.390
definitely shouldn't be?
403
00:20:31.029 --> 00:20:33.529
Well, it's definitely important to minimize the size of the
404
00:20:33.549 --> 00:20:35.619
blockchain because one of the problems that Ethereum has is
405
00:20:35.759 --> 00:20:37.759
everything is on the blockchain. So you end up with
406
00:20:37.779 --> 00:20:40.920
a very, very large shared hard drive base that everybody has.
407
00:20:41.339 --> 00:20:43.640
Every blockchain has that to some degree, but you want
408
00:20:43.660 --> 00:20:46.079
to minimize what those are. We call them elevated payloads
409
00:20:46.119 --> 00:20:48.460
where the elevated payloads actually go on chain and they
410
00:20:48.480 --> 00:20:50.460
have to be there and they get validated. But then
411
00:20:50.500 --> 00:20:52.559
the non-elevated payloads are things that are in data lakes.
412
00:20:52.579 --> 00:20:54.519
So for example, if I want to take a movie
413
00:20:54.599 --> 00:20:56.720
and hash that movie, I can store that in the
414
00:20:56.769 --> 00:20:58.589
data lake or I can store it somewhere else and
415
00:20:58.750 --> 00:21:02.109
put the hash on chain. The downside of that is
416
00:21:02.390 --> 00:21:04.589
it's loss is losable as far as I can, I
417
00:21:04.630 --> 00:21:08.009
can lose that, that data. And so people, when they
418
00:21:08.029 --> 00:21:10.529
hear the word permanence, for example, we talk about, um,
419
00:21:10.910 --> 00:21:14.750
ex bio layer one provides permanence. There's two definitions of
420
00:21:14.789 --> 00:21:17.349
permanence that I think are important to differentiate. One is
421
00:21:18.049 --> 00:21:20.569
it can't be lost, which is often what people think
422
00:21:20.609 --> 00:21:24.880
of permanence. And one is it can't be changed. Refocus
423
00:21:24.900 --> 00:21:27.740
on the second one. It can't be changed. So I
424
00:21:27.819 --> 00:21:30.579
permanently locked in my hand that I played, for example,
425
00:21:30.759 --> 00:21:35.250
is permanence. I permanently put my hand on Ethereum's blockchain,
426
00:21:35.309 --> 00:21:38.970
so I can't lose it is also permanence, but the
427
00:21:39.130 --> 00:21:41.690
different permanence. So the permanence of I can't lose it
428
00:21:41.730 --> 00:21:44.990
is very expensive. The permanence of I can't change it
429
00:21:45.069 --> 00:21:47.359
can be very inexpensive. It's just one hash basically for
430
00:21:47.390 --> 00:21:50.200
as much data as you want to. So we really
431
00:21:50.440 --> 00:21:53.910
try to tailor X by layer one to be as
432
00:21:54.049 --> 00:21:58.210
optimal as possible, which does make it much better for
433
00:21:58.250 --> 00:22:00.549
certain use cases and also makes it not possible necessarily
434
00:22:00.589 --> 00:22:02.069
for other use cases. So if you have to have
435
00:22:02.089 --> 00:22:06.710
your data not losable, that's something which is not really
436
00:22:06.779 --> 00:22:08.640
practical for X, Y, O layer one. But I would
437
00:22:08.660 --> 00:22:10.660
also argue that it's not really practical for anything that's
438
00:22:10.740 --> 00:22:15.380
out there because things can still get lost. Definitely.
439
00:22:15.980 --> 00:22:20.170
Yeah. especially those private keys from Bitcoin back in 2010.
440
00:22:20.170 --> 00:22:25.630
I wish we had those. What are your thoughts on
441
00:22:25.869 --> 00:22:29.309
how much focus there is on the synergy of AI
442
00:22:29.390 --> 00:22:34.410
and blockchain throughout the blockchain industry? I know there's a
443
00:22:34.789 --> 00:22:38.180
couple of protocols that are focused on decentralized AI and
444
00:22:38.460 --> 00:22:40.980
improving AI. I feel like a lot of the focus
445
00:22:41.019 --> 00:22:43.920
is on finance, which is important, of course, and an
446
00:22:43.980 --> 00:22:44.599
early use case.
447
00:22:44.759 --> 00:22:45.750
But maybe there's not.
448
00:22:45.730 --> 00:22:50.750
Enough discussion going on right now in understanding how important
449
00:22:50.789 --> 00:22:53.589
blockchain is to building up AI properly.
450
00:22:54.849 --> 00:22:57.329
Well, I think it's kind of an interesting situation. This
451
00:22:57.390 --> 00:23:00.099
comes along every few years where people are like, oh,
452
00:23:00.150 --> 00:23:02.619
AI is a hard word and blockchain is a hard word.
453
00:23:03.000 --> 00:23:04.200
Let's put them together and put them in a pitch
454
00:23:04.240 --> 00:23:07.720
deck and see what happens, right? So I think often
455
00:23:07.819 --> 00:23:12.079
you'll see more projects that are trying to merge them.
456
00:23:12.200 --> 00:23:16.309
And in some cases, it's just like have a AI
457
00:23:16.329 --> 00:23:18.789
bridge or some sort of AI router, for example, that
458
00:23:18.910 --> 00:23:21.549
uses blockchain. And it's, well, how is it better with
459
00:23:21.569 --> 00:23:23.970
blockchain in that case? So you always have to ask,
460
00:23:23.990 --> 00:23:26.630
you know, is it better if you combine these two
461
00:23:26.670 --> 00:23:29.559
things or is it not better? And I'd say in
462
00:23:29.619 --> 00:23:33.279
most cases, it may not be better necessarily. It's just fuzzy.
463
00:23:33.640 --> 00:23:36.980
It's not better. But in many cases, like in that
464
00:23:37.019 --> 00:23:40.720
layer one case, I think there's a distinct usage for
465
00:23:40.789 --> 00:23:43.500
it and a way to make the system better. where
466
00:23:44.099 --> 00:23:48.759
if I could have a provable history of say robots
467
00:23:48.900 --> 00:23:53.440
or autonomous hardware that's running around, especially if they can
468
00:23:53.980 --> 00:23:56.369
check in with each other, cosign things and then put
469
00:23:56.390 --> 00:23:59.069
those on the blockchain, I can see where these items were.
470
00:23:59.470 --> 00:24:01.609
And in a weird sort of way, it's almost like
471
00:24:01.630 --> 00:24:04.910
a real life version of our coin app where having
472
00:24:04.930 --> 00:24:10.390
these things check in autonomously into the blockchain allows you
473
00:24:10.650 --> 00:24:13.789
to see what actually happened in the real world, because
474
00:24:13.809 --> 00:24:14.609
I think it's going to be one of the, one
475
00:24:14.619 --> 00:24:17.690
of the difficulties is going to be, well, you know,
476
00:24:17.730 --> 00:24:20.130
something bad happened. Um, was it my robot that did it?
477
00:24:20.150 --> 00:24:21.849
Was it your robot that did it? Was it, yeah,
478
00:24:22.349 --> 00:24:24.369
it was on purpose. Was it by accident? And if,
479
00:24:24.549 --> 00:24:26.250
you know, having audit logs for those things that are
480
00:24:26.269 --> 00:24:30.849
constantly being stored and also constantly being, um, like bookmarked
481
00:24:30.910 --> 00:24:33.250
on an X by layer one system, we'll make it
482
00:24:33.289 --> 00:24:35.170
so that we can go back in time and we
483
00:24:35.190 --> 00:24:38.059
can at least, um, find some sort of direction for
484
00:24:38.079 --> 00:24:41.549
where it is. And so, an actual solution for an
485
00:24:41.650 --> 00:24:45.529
actual problem is important to have. And so I think
486
00:24:46.490 --> 00:24:51.150
some cases where we see that synergy, that's not the case.
487
00:24:51.190 --> 00:24:54.809
In some cases, it's extremely true to be the case.
488
00:24:54.869 --> 00:24:59.839
And so I think there's probably more place for blockchain
489
00:25:00.430 --> 00:25:03.640
in AI or in the world now with AI around
490
00:25:03.680 --> 00:25:06.400
than there was before, just because it's a natural fit.
491
00:25:08.220 --> 00:25:08.619
I agree.
492
00:25:08.660 --> 00:25:10.660
I'm looking forward to seeing it more And I think
493
00:25:10.740 --> 00:25:15.420
another major synergy we haven't spoken about yet is not
494
00:25:15.460 --> 00:25:19.359
just AI and the data provenance of blockchain, but the
495
00:25:19.400 --> 00:25:24.130
payments and the wallets of AI agents or humanoid robots
496
00:25:24.250 --> 00:25:26.970
or self-driving cars and things like that. Is there a
497
00:25:27.009 --> 00:25:34.009
plan for XYO to extend beyond data to ensuring the
498
00:25:34.049 --> 00:25:38.130
validity of payments or the success in having that infrastructure expanded?
499
00:25:38.470 --> 00:25:41.690
in there as robots start interacting and paying on our
500
00:25:41.730 --> 00:25:42.849
behalf or paying each other?
501
00:25:42.869 --> 00:25:47.210
Well, we're looking more at facilitating payments as opposed to
502
00:25:47.289 --> 00:25:50.349
validating other people's payments. So for example, one of the
503
00:25:50.369 --> 00:25:53.910
things which we built into CryptoCards that's not really that
504
00:25:53.970 --> 00:25:55.250
easy to see unless you go and look at the
505
00:25:55.269 --> 00:25:58.170
blockchain and understand it is we have an eventing system
506
00:25:58.210 --> 00:26:01.019
where basically on our blockchain, a person can go and
507
00:26:01.380 --> 00:26:05.440
register for an event by assigning a payload with their address.
508
00:26:06.019 --> 00:26:08.880
And then the eventing server will start sending those events
509
00:26:08.920 --> 00:26:12.900
to you through like a webhook, for example. And then
510
00:26:14.220 --> 00:26:15.750
it can stop when you stop paying for that. But
511
00:26:15.769 --> 00:26:16.869
what you can do is you can say, well, I
512
00:26:16.910 --> 00:26:19.750
want to get this event for the next, say, 1,000 blocks.
513
00:26:20.410 --> 00:26:22.769
And I'm going to pay 3XL1 to get those events
514
00:26:22.829 --> 00:26:26.490
or a certain amount per event. So the actual payment
515
00:26:26.750 --> 00:26:31.970
of microservices, I mean like micro-microservices, like nano-services, is something
516
00:26:31.990 --> 00:26:33.470
which I think is going to become much more prevalent
517
00:26:33.509 --> 00:26:38.380
where A robot or AI might be like, hey, I
518
00:26:38.420 --> 00:26:42.759
want to know for the next two minutes, when does
519
00:26:42.819 --> 00:26:45.299
a new block appear on X by layer one? And
520
00:26:45.319 --> 00:26:47.319
I want to get those events pretty frequently for the
521
00:26:47.339 --> 00:26:49.109
next two minutes and then stop. But I don't want
522
00:26:49.119 --> 00:26:52.289
to go and have to create an account with a
523
00:26:52.329 --> 00:26:54.390
credit card and that sort of thing to be able
524
00:26:54.410 --> 00:26:57.049
to use the API for whatever it is. And so
525
00:26:57.150 --> 00:27:01.089
I think real-time ad hoc usage of services that get
526
00:27:01.170 --> 00:27:04.069
paid for via blockchain payments in a very, very efficient
527
00:27:04.150 --> 00:27:06.960
way is going to be extremely important for AI and
528
00:27:07.079 --> 00:27:10.420
for robotics because of the fact that they need a
529
00:27:10.460 --> 00:27:15.259
very broad spectrum of access to things. And they need
530
00:27:15.279 --> 00:27:16.859
to be able to pay for those things efficiently and
531
00:27:16.880 --> 00:27:19.660
to authenticate for them efficiently without having to go through
532
00:27:19.819 --> 00:27:23.380
all the pomp and ceremony of credit cards and accounts
533
00:27:23.529 --> 00:27:25.390
and logins. Definitely.
534
00:27:25.509 --> 00:27:27.069
And micropayments has been a.
535
00:27:28.710 --> 00:27:32.589
Supposed use case of blockchain for many, many years, but
536
00:27:32.710 --> 00:27:35.190
we haven't really seen it. at least in real world payments.
537
00:27:35.269 --> 00:27:39.250
But now with AI and token usage, we're starting to see.
538
00:27:39.849 --> 00:27:42.069
Every word is a micro cent.
539
00:27:42.210 --> 00:27:46.450
And that seems like the perfect use case for micropayments.
540
00:27:46.490 --> 00:27:49.230
We're still putting your credit card into the AI and
541
00:27:49.269 --> 00:27:52.670
just preloading it. But I feel like there's a perfect
542
00:27:52.970 --> 00:27:56.400
use case for micropayments on blockchain with AI. It's like
543
00:27:56.500 --> 00:27:58.900
the number one use case that's now coming to light.
544
00:28:00.210 --> 00:28:01.809
I thought it was actually kind of funny when they
545
00:28:02.109 --> 00:28:04.450
came out with the AI thing and called it tokens.
546
00:28:04.569 --> 00:28:06.410
It's like, how many tokens in, how many tokens out?
547
00:28:06.430 --> 00:28:08.470
I'm like, well, I've heard the word tokens before. It's
548
00:28:08.650 --> 00:28:12.450
an ERC-20 token. So they're kind of modeling that to
549
00:28:12.490 --> 00:28:17.180
some degree where it's a token economy. I'm surprised they
550
00:28:17.200 --> 00:28:20.720
haven't said yet, well, I'll pay you in 500 Claude
551
00:28:22.019 --> 00:28:23.720
tokens and I'll use that as money to pay you
552
00:28:23.740 --> 00:28:25.519
with or something like that. But there's no way for
553
00:28:25.539 --> 00:28:28.299
me to actually transfer tokens from one account to another
554
00:28:28.319 --> 00:28:31.720
account on Claude before they know. But there's nothing saying
555
00:28:31.759 --> 00:28:36.720
that I can't use their tokens as a payment method.
556
00:28:36.759 --> 00:28:40.359
But I think definitely using blockchain as a way to
557
00:28:40.519 --> 00:28:45.740
autonomously give tokens, for example, to an agent for its use.
558
00:28:46.539 --> 00:28:49.299
is something which we're going to be doing very often
559
00:28:49.440 --> 00:28:51.329
in the near future, as opposed to saying, well, here's
560
00:28:51.369 --> 00:28:54.029
my login for my cloud account, use as many tokens
561
00:28:54.089 --> 00:28:55.970
as you want. That's a little scary, where if I
562
00:28:56.009 --> 00:28:59.069
can give it an allowance with a blockchain system, it's
563
00:28:59.109 --> 00:29:00.829
something which I think people would definitely want to use.
564
00:29:00.890 --> 00:29:05.880
But the token economy is something which I think will
565
00:29:05.940 --> 00:29:08.759
blossom a lot under AI, just because AI is so
566
00:29:08.779 --> 00:29:10.700
much more fragmented in the way that it uses the world,
567
00:29:10.720 --> 00:29:12.609
as opposed to a human where it's like, well, I
568
00:29:12.630 --> 00:29:15.630
don't want to do a microtransaction for every single bite
569
00:29:15.670 --> 00:29:17.660
of food that I eat at a restaurant, for example,
570
00:29:17.740 --> 00:29:20.259
or for every minute of the movie that I watch.
571
00:29:20.660 --> 00:29:22.500
I want to just get the whole movie. And so
572
00:29:22.700 --> 00:29:25.579
microtransactions have kind of, I think, failed to this point
573
00:29:26.079 --> 00:29:29.039
with blockchain for those two reasons. One, it's just the
574
00:29:29.059 --> 00:29:31.589
gas cost for Bitcoin and for Ethereum is way too expensive.
575
00:29:32.049 --> 00:29:36.740
But two, it's humans don't think in those micro steps
576
00:29:37.099 --> 00:29:38.259
the way that AI can.
577
00:29:39.240 --> 00:29:42.759
Yeah, no, that makes perfect sense. And, you know, it's
578
00:29:42.799 --> 00:29:44.819
two months since our last video. I would love to
579
00:29:44.859 --> 00:29:46.779
get you back on once more at the end of
580
00:29:46.819 --> 00:29:49.730
the year and see what Q4 has in store for
581
00:29:49.789 --> 00:29:52.869
the growth of AI. It's just getting faster and faster,
582
00:29:52.930 --> 00:29:55.470
as well as blockchain now kicking back up, not just
583
00:29:56.190 --> 00:29:59.190
the development of blockchains and you know, X, Y, O
584
00:29:59.190 --> 00:30:00.740
layer one is developing super fast.
585
00:30:00.839 --> 00:30:03.480
I understand using AI, but also the prices are going
586
00:30:03.519 --> 00:30:03.839
back up.
587
00:30:03.859 --> 00:30:06.299
So people are getting more interested and more people are
588
00:30:06.319 --> 00:30:10.460
looking at different ways that these two technologies can synergize together.
589
00:30:11.049 --> 00:30:14.049
What is some of the most important things that are
590
00:30:14.089 --> 00:30:16.289
happening at X, Y, O between now and the end
591
00:30:16.329 --> 00:30:18.170
of the year that we may have to look forward to?
592
00:30:19.609 --> 00:30:21.650
Well, it kind of says something interesting there, which I
593
00:30:21.690 --> 00:30:25.099
think is very appropriate for thinking about AI in the
594
00:30:25.140 --> 00:30:27.420
context of X, Y, O layer one. Well, we use
595
00:30:27.460 --> 00:30:29.720
AI in two different ways. One is we use AI
596
00:30:29.759 --> 00:30:32.660
to develop and to accelerate how fast we can do
597
00:30:32.700 --> 00:30:35.329
really cool things and make new products with X, Y,
598
00:30:35.380 --> 00:30:37.880
O layer one. So AI as a tool for us
599
00:30:37.920 --> 00:30:41.420
has been fantastic. And we've really dove into that, made
600
00:30:41.440 --> 00:30:44.069
a bunch of skills and that we use ourselves, we
601
00:30:44.109 --> 00:30:45.880
share with our partners and they can use the skills
602
00:30:45.940 --> 00:30:49.089
as well. And then also we prioritize AI things. Kind
603
00:30:49.109 --> 00:30:51.849
of the crypto cards is that where the house, which
604
00:30:51.970 --> 00:30:55.369
plays the hand of crypto cards uses AI the AI
605
00:30:55.450 --> 00:30:58.390
from gate, for example, in this case, to actually decide, well,
606
00:30:58.430 --> 00:31:00.740
how should I play that? And so prioritizing AI is
607
00:31:00.779 --> 00:31:02.099
the other aspect of it. Well, how do you use
608
00:31:02.220 --> 00:31:05.539
AI as a feature in the product that you want
609
00:31:05.579 --> 00:31:08.039
to use? And so we're really focusing on both of those.
610
00:31:08.079 --> 00:31:10.920
We're focusing on using AI as a tool to accelerate
611
00:31:10.980 --> 00:31:14.529
ourselves and reusing AI as a feature that we can
612
00:31:14.609 --> 00:31:17.630
use in our products and in different things and interlacing
613
00:31:17.670 --> 00:31:23.069
those features with hashes and with provenance and sovereignty through X, Y,
614
00:31:23.109 --> 00:31:23.589
or layer one.
615
00:31:25.579 --> 00:31:28.380
I'm definitely going to test out these crypto card first hands.
616
00:31:28.500 --> 00:31:29.299
I saw the announcement.
617
00:31:29.920 --> 00:31:32.900
I'm subscribed to when the first hands go live. I'll
618
00:31:32.940 --> 00:31:34.779
put it in the show notes below because we've mentioned
619
00:31:34.799 --> 00:31:38.410
it quite a bit. I appreciate your insights into everything AI.
620
00:31:38.470 --> 00:31:42.089
You're such a smart brain. And I'm looking forward to
621
00:31:42.630 --> 00:31:44.450
talking with you again in the near future. Thank you
622
00:31:44.470 --> 00:31:45.410
so much for the time, Ari.
623
00:31:46.289 --> 00:31:47.490
I appreciate it. Thanks so much for having me on.