Speaker: 00:00:00

One of the things they have to do is write

down, here's the rules of the game,

00:00:04,170 --> 00:00:08,383

and here's what, doing a certain

good thing pays you, right?

00:00:08,383 --> 00:00:10,218

They call it a payoff function.

00:00:10,218 --> 00:00:13,221

And so if you, you know, if you take the,

the opponent's pawn little piece,

00:00:13,555 --> 00:00:14,222

here's the payoff.

00:00:14,222 --> 00:00:16,558

Or if you take their queen,

it's a bigger payoff.

00:00:16,558 --> 00:00:20,895

And so, if you can define that in chess,

you can actually define that.

00:00:21,771 --> 00:00:24,566

You can make a system

that that does very well.

00:00:24,566 --> 00:00:28,737

But think about a dating relationship

or a marriage like.

00:00:29,154 --> 00:00:32,240

And we pretty quickly

realize the our attempts

00:00:32,240 --> 00:00:35,243

to quantify and define value.

00:00:36,077 --> 00:00:38,872

It's we're in a different layer

of abstraction.

00:00:38,872 --> 00:00:41,666

We you can't the these don't go together.

00:00:47,297 --> 00:00:49,174

So today in Anabaptist perspectives,

00:00:49,174 --> 00:00:53,511

I'm joined by Ben Harris,

and we're going to be diving into

00:00:53,595 --> 00:00:56,765

the ontological limits

of artificial intelligence.

00:00:58,808 --> 00:00:59,684

Ben, you want to start with a

00:00:59,684 --> 00:01:02,771

little introduction,

and we'll jump into the topic after that.

00:01:03,646 --> 00:01:04,230

Sure, Marlin.

00:01:04,230 --> 00:01:06,191

Good to be with you. So,

my name is Ben Harris.

00:01:06,191 --> 00:01:07,484

Hi, everyone.

00:01:07,484 --> 00:01:11,071

So I'm a professor up at Sattler

College in Boston, Massachusetts.

00:01:11,071 --> 00:01:13,364

I coordinate the business program up here,

00:01:13,364 --> 00:01:15,825

but my my background

comes out of the engineering world.

00:01:15,825 --> 00:01:20,747

I spent more than a decade working in

machine learning, artificial intelligence,

00:01:20,747 --> 00:01:22,749

just many of the classical engineering

disciplines.

00:01:22,749 --> 00:01:25,752

So, I would.

00:01:26,461 --> 00:01:29,339

I think it's hard to define

what an expert in the field is, but,

00:01:29,339 --> 00:01:33,218

I spent a long time thinking about it

and continue to do so because it affects

00:01:33,510 --> 00:01:37,514

not only back in, in the technical world,

but also in academia.

00:01:37,514 --> 00:01:39,015

We contend with AI every day.

00:01:39,015 --> 00:01:41,518

So, Marlin, it's good to be with you.

00:01:41,518 --> 00:01:44,521

Yes. Thanks for coming on, Ben.

00:01:44,771 --> 00:01:47,607

Excited to have the kind of engineering,

computer background,

00:01:47,607 --> 00:01:48,608

that you bring to it.

00:01:49,734 --> 00:01:51,069

So to

00:01:51,069 --> 00:01:55,949

start

with some of the the drama around AI,

00:01:55,949 --> 00:01:58,952

a little over a year ago, May of 2023,

00:01:59,953 --> 00:02:03,998

there's this famous statement on AI

risk, signed by,

00:02:03,998 --> 00:02:09,504

you know, a bunch of the big names,

and their short version was mitigating

00:02:09,504 --> 00:02:13,007

the risk of extinction from

AI should be a global priority

00:02:14,008 --> 00:02:17,971

alongside other societal scale risks

such as pandemics and nuclear war.

00:02:18,763 --> 00:02:21,641

And so that got a bunch of press.

00:02:21,641 --> 00:02:22,684

I found it interesting.

00:02:22,684 --> 00:02:25,687

A few weeks later,

there was a CNN report on a,

00:02:26,521 --> 00:02:29,524

a survey from a number of CEOs.

00:02:30,066 --> 00:02:32,402

they got 119 responses.

00:02:32,402 --> 00:02:35,613

And out of those, 50 of them said,

00:02:37,490 --> 00:02:41,327

yeah, AI could potentially destroy humanity

in the next 5 to 10 years.

00:02:42,078 --> 00:02:44,581

And the other 69 were unconcerned.

00:02:45,915 --> 00:02:47,250

So I guess we

00:02:47,250 --> 00:02:50,253

can start with, where are you

at on that question?

00:02:50,253 --> 00:02:53,256

Your view of artificial intelligence

and that kind of.

00:02:54,048 --> 00:02:57,468

Yeah,

that kind of dramatic fear statements.

00:02:58,720 --> 00:03:01,973

Oh, that's a that's a classic question

for anybody who thinks about AI.

00:03:02,390 --> 00:03:05,768

I, I am not on the

what we call an AI doomer.

00:03:05,768 --> 00:03:07,770

Those are

those are probably what you put those,

00:03:07,770 --> 00:03:09,898

you know,

the Geoffrey Hinton's of the world

00:03:09,898 --> 00:03:13,484

that would you know, he's very, he’s

had a much longer background here.

00:03:13,484 --> 00:03:14,736

This is his view.

00:03:14,736 --> 00:03:19,449

I do not see AI as within that short

period of time, 5 to 10 years,

00:03:20,283 --> 00:03:23,286

as being an existential threat

to humanity.

00:03:24,078 --> 00:03:26,956

Both for a theological reason,

I think, like God has defined

00:03:26,956 --> 00:03:30,793

the end of humanity for us,

but also that we

00:03:31,169 --> 00:03:36,174

we look at the technology

and what we assume in the forecast of,

00:03:36,174 --> 00:03:39,219

you know, AI improving

or growing to a certain level assumes

00:03:39,219 --> 00:03:42,222

no major hurdles

that we encounter on the way.

00:03:42,347 --> 00:03:44,432

I and I just think we're starting

to see the hurdles.

00:03:44,432 --> 00:03:44,724

Right?

00:03:44,724 --> 00:03:47,936

We we at the moment cannot produce

00:03:47,936 --> 00:03:51,231

data fast enough in order to support

these bigger and bigger AI models.

00:03:51,231 --> 00:03:51,522

We don't.

00:03:51,522 --> 00:03:52,857

We're running out of,

00:03:52,857 --> 00:03:57,195

raw materials for creating training

chips and systems that do this.

00:03:57,654 --> 00:03:59,364

So there

so we're starting to see the hurdles.

00:03:59,364 --> 00:04:00,323

And I think that that's going to

00:04:00,323 --> 00:04:03,326

that's going to skew the forecast

fairly dramatically.

00:04:03,701 --> 00:04:07,747

I, I am not particularly concerned about,

existential issues.

00:04:07,747 --> 00:04:08,581

partly

00:04:08,581 --> 00:04:12,377

because what we're going to talk about

today is the is the ontology of AI, right?

00:04:12,377 --> 00:04:16,422

The movie

versions of AI being a threat to humans,

00:04:17,840 --> 00:04:20,927

in the in the

narrative tend to revolve around a moment

00:04:21,094 --> 00:04:25,098

when I realized what it was

and what it could be, and it has this self

00:04:25,098 --> 00:04:29,352

defensive response or reaction to humanity

trying to shut it down.

00:04:30,395 --> 00:04:33,398

I don't yet think that AI has the

00:04:33,564 --> 00:04:37,193

has an existential understanding

of what it is yet.

00:04:37,485 --> 00:04:40,071

You know, it

probably can give you a text response.

00:04:40,071 --> 00:04:43,825

Yes, I system, system system

am an AI system,

00:04:44,325 --> 00:04:48,997

but that does that does not yet imbue

value in a self defensive reaction.

00:04:48,997 --> 00:04:51,499

Right? The

that we naturally experience as humans.

00:04:51,499 --> 00:04:55,545

So I think there's still a gap

ontologically and as well as what

00:04:55,545 --> 00:04:59,716

we would call artificial

generalized intelligence or AGI, right.

00:04:59,716 --> 00:05:02,719

Systems

that can think and reason for themselves.

00:05:03,970 --> 00:05:05,638

I mean, it's still an ongoing open

00:05:05,638 --> 00:05:08,850

research area in, in what they call

argument mining, right?

00:05:08,850 --> 00:05:13,062

You can present a bit of text to an engine

and say, is this a good argument?

00:05:13,521 --> 00:05:16,482

And it has a really difficult time

determining

00:05:16,482 --> 00:05:19,152

yes or no, whereas

we as humans read that and say, oh, that's

00:05:19,152 --> 00:05:22,196

a terrible argument or a great argument,

and I find it very persuasive.

100

00:05:22,196 --> 00:05:27,410

And so the, there's the that's one

I think about a lot that particular gap.

101

00:05:27,410 --> 00:05:31,122

But there are a number

of them that that in my view,

102

00:05:32,248 --> 00:05:33,833

set they set a

103

00:05:33,833 --> 00:05:37,545

fairly big chasm today between where AI is

104

00:05:37,795 --> 00:05:40,798

and anything that we would need to worry

about from an existential standpoint.

105

00:05:42,008 --> 00:05:43,676

So you're setting a chasm.

106

00:05:43,676 --> 00:05:48,222

And you're saying those time frames

are way over inflated or sorry,

107

00:05:49,390 --> 00:05:50,433

not over inflated.

108

00:05:50,433 --> 00:05:52,143

Opposite of over inflated.

109

00:05:52,143 --> 00:05:54,854

Way too ambitious in a sense.

110

00:05:54,854 --> 00:05:56,647

To push on it a little bit.

111

00:05:56,647 --> 00:05:58,983

I still hear you using words like.

112

00:05:58,983 --> 00:06:04,781

Not yet, implying that the time is coming.

113

00:06:04,781 --> 00:06:06,491

It's just not yet.

114

00:06:06,491 --> 00:06:09,494

and I guess even to push on that

a little bit more,

115

00:06:10,370 --> 00:06:13,373

you know,

we talk about things in computing, like

116

00:06:13,664 --> 00:06:17,585

Moore's Law, which I think is a fairly

specific technical definition for that.

117

00:06:17,585 --> 00:06:20,588

But in a general sense,

118

00:06:22,006 --> 00:06:24,884

you know, computing

power changes very quickly.

119

00:06:26,844 --> 00:06:29,806

I mean, you even noted in an email to me,

it seems like there's new

120

00:06:29,806 --> 00:06:32,809

AI applications

coming out every couple of weeks.

121

00:06:35,311 --> 00:06:35,770

so I guess

122

00:06:35,770 --> 00:06:38,773

how would you respond

to an argument like that that says, well,

123

00:06:39,273 --> 00:06:42,276

you know, computing power

has been doubling

124

00:06:42,693 --> 00:06:46,072

frequently for the last number of years,

and we're just going to see,

125

00:06:47,156 --> 00:06:50,118

you know,

we can't see what's around the corner.

126

00:06:50,118 --> 00:06:51,077

You know, it's a fair question.

127

00:06:51,077 --> 00:06:52,453

You know why? Why the “Not yet”?

128

00:06:52,453 --> 00:06:56,749

I think the the one of the reasons is

if you, you know, say you

129

00:06:56,749 --> 00:06:58,918

you project out

some number of years and you,

130

00:06:58,918 --> 00:07:02,713

you allow Moore's law, which is beginning

to show asymptotic behavior.

131

00:07:03,005 --> 00:07:03,256

Right?

132

00:07:03,256 --> 00:07:07,885

You if you were not doubling the you know

our our speeds are getting quicker,

133

00:07:07,885 --> 00:07:11,931

but we're running into computation issues

with these kind of things.

134

00:07:11,931 --> 00:07:14,016

It's largely why GPUs have become

135

00:07:14,016 --> 00:07:17,019

the standard architecture

for doing this kind of work.

136

00:07:17,145 --> 00:07:20,398

one, the not yet is an admission

that I don't that we don't know.

137

00:07:20,690 --> 00:07:20,898

Right.

138

00:07:20,898 --> 00:07:24,068

We you know,

can can you create a technical system

139

00:07:24,068 --> 00:07:27,363

that that is a version of intelligence?

140

00:07:27,905 --> 00:07:30,032

That's a bigger question

than I think AI can answer.

141

00:07:30,032 --> 00:07:34,078

That's a, you know, what is what has God

defined intelligence as.

142

00:07:34,078 --> 00:07:35,913

And can you create a system that has that?

143

00:07:37,832 --> 00:07:38,541

you know,

144

00:07:38,541 --> 00:07:41,586

we sort of think of it

in a, in a narrow band in that,

145

00:07:42,003 --> 00:07:45,214

you know, intelligence

is the ability to collect and synthesize

146

00:07:45,214 --> 00:07:48,217

information, to answer a question

or to develop something.

147

00:07:48,342 --> 00:07:50,761

but there's there's other types. Right?

148

00:07:50,761 --> 00:07:52,472

We have a sense of asthetics.

149

00:07:52,472 --> 00:07:55,183

Right? Humans look at what is beautiful.

150

00:07:55,183 --> 00:08:00,271

We have a sense of awe,

that is uniquely human, right?

151

00:08:00,271 --> 00:08:04,650

And you, you know,

if you ask, a large language model,

152

00:08:04,984 --> 00:08:09,238

you know, if a given painting is,

is beautiful or if a piece of artwork

153

00:08:09,238 --> 00:08:13,701

is just abhorrent to you, like,

it's likely does not have an opinion.

154

00:08:14,202 --> 00:08:17,705

And if it does, it has to create

it based on some quantitative

155

00:08:17,705 --> 00:08:18,873

aspect of the artwork.

156

00:08:18,873 --> 00:08:23,127

It has to look at, you know, the okay,

the color contrast is within this range.

157

00:08:23,127 --> 00:08:26,172

And most people think that that means it's

not a very attractive painting.

158

00:08:26,422 --> 00:08:27,340

It has.

159

00:08:27,340 --> 00:08:30,593

That's the way it has to think

because of its computational nature.

160

00:08:31,052 --> 00:08:33,387

we don't think that way. Right?

161

00:08:33,387 --> 00:08:35,139

We we look at it. And what is it?

162

00:08:35,139 --> 00:08:38,351

What does it do to our to our spirit,

to our soul, to our mind?

163

00:08:38,684 --> 00:08:40,144

And we have a response. And so,

164

00:08:41,187 --> 00:08:44,190

you know, I think that the not yet is a,

165

00:08:45,149 --> 00:08:47,193

you know, is a is a fair question.

166

00:08:47,193 --> 00:08:48,736

I think we're, we're also running

167

00:08:48,736 --> 00:08:52,114

into some mathematical challenges

in that the, the number of,

168

00:08:52,907 --> 00:08:56,410

I'll call them synapses or nodes

that we'd have to do to simulate

169

00:08:56,744 --> 00:09:00,915

actual human thought and cognition

is still many orders of magnitude

170

00:09:00,915 --> 00:09:03,918

above what the most powerful systems

can handle right now.

171

00:09:04,126 --> 00:09:07,296

So even if Moore's Law holds,

which is question whether it will,

172

00:09:08,923 --> 00:09:12,426

we have a long time to go

before we exponentially reach that point.

173

00:09:12,426 --> 00:09:15,304

We would need to be and who

and who knows?

174

00:09:15,304 --> 00:09:17,932

On the way, we may encounter a whole

nother obstacle we didn't anticipate.

175

00:09:17,932 --> 00:09:23,145

So, the the challenge is that we

you don't.

176

00:09:24,063 --> 00:09:26,232

When you make a forecast,

you have to acknowledge

177

00:09:26,232 --> 00:09:28,568

what could cause the forecast to fail.

178

00:09:28,568 --> 00:09:29,193

Right?

179

00:09:29,193 --> 00:09:31,362

Things don't always continue as they were.

180

00:09:31,362 --> 00:09:34,782

You look at, you know, population

explosion in the 1970s, right.

181

00:09:34,782 --> 00:09:35,950

There was this huge concern

182

00:09:35,950 --> 00:09:38,953

that it was going to lead to global famine

and billions of deaths.

183

00:09:38,953 --> 00:09:41,163

But that never happened. And so,

184

00:09:42,373 --> 00:09:43,207

so we want to look back

185

00:09:43,207 --> 00:09:46,586

soberly at those examples,

and I think with a bit of wisdom,

186

00:09:47,545 --> 00:09:52,466

and just encourage ourselves that, like,

the Lord has this in hand, right?

187

00:09:52,508 --> 00:09:55,636

He's not he's not going to let AI ruin

188

00:09:55,636 --> 00:09:57,847

his creation.

189

00:09:58,723 --> 00:10:01,726

Okay, so a theological answer there.

190

00:10:02,310 --> 00:10:05,313

Confidence in God.

191

00:10:06,439 --> 00:10:07,273

Yeah.

192

00:10:07,273 --> 00:10:10,943

And the title Ontological Limits kind of

highlights what I wanted to the question.

193

00:10:10,943 --> 00:10:12,945

I kind of want to press here.

194

00:10:12,945 --> 00:10:15,948

if we talk about limits of AI,

195

00:10:16,490 --> 00:10:20,119

maybe the first limit

we think about is technological.

196

00:10:20,328 --> 00:10:21,245

Can we build it?

197

00:10:21,245 --> 00:10:22,163

What can we build?

198

00:10:22,163 --> 00:10:23,956

What kind of

199

00:10:23,956 --> 00:10:28,044

computers can we build,

what kinds of neural networks and so on?

200

00:10:30,254 --> 00:10:33,591

or epistemological

what do we know how to do?

201

00:10:33,591 --> 00:10:35,134

Knowledge.

202

00:10:35,134 --> 00:10:38,512

when we say ontological limits,

we're getting into.

203

00:10:39,347 --> 00:10:42,892

Some ways, the strongest of those terms

because we're trying to say what

204

00:10:43,934 --> 00:10:46,937

what is the limit by the very,

205

00:10:47,146 --> 00:10:50,149

by the very nature of what this thing is,

206

00:10:50,983 --> 00:10:53,986

that produces the limit,

207

00:10:55,404 --> 00:10:57,698

and yeah, I think maybe you've hinted

at some of that,

208

00:10:57,698 --> 00:11:00,701

but.

209

00:11:01,911 --> 00:11:04,664

Yeah, kind of at its core,

210

00:11:04,664 --> 00:11:07,667

what do you see is that biggest core limit

211

00:11:08,334 --> 00:11:11,337

or way of talking about that core limit.

212

00:11:11,420 --> 00:11:15,341

I think probably the the,

the most stark ontological limit

213

00:11:15,341 --> 00:11:19,136

of AI is is just what we've created it out

of, right?

214

00:11:19,136 --> 00:11:22,139

This is a this is a creation of,

215

00:11:23,015 --> 00:11:25,434

Well, it's it's a, sounds funny, a

216

00:11:25,434 --> 00:11:26,602

creation of creation.

217

00:11:26,602 --> 00:11:26,811

Right.

218

00:11:26,811 --> 00:11:32,191

We've our ingenuity has developed this

and it's immensely powerful in some areas.

219

00:11:32,233 --> 00:11:32,358

Right.

220

00:11:32,358 --> 00:11:35,361

I don't want to to minimize

its effectiveness in its aid in some

221

00:11:35,361 --> 00:11:38,364

things, but there

222

00:11:39,407 --> 00:11:39,740

it is

223

00:11:39,740 --> 00:11:43,703

a it does

not involve the supernatural, right.

224

00:11:43,744 --> 00:11:44,912

It does not.

225

00:11:44,912 --> 00:11:48,749

You know, it is a it is a, by definition,

natural development.

226

00:11:48,749 --> 00:11:52,420

It is limited by the laws that the

that God has put in place

227

00:11:52,420 --> 00:11:56,132

in terms of physics and computation

and, and information,

228

00:11:57,299 --> 00:11:58,217

in mathematics. Right.

229

00:11:58,217 --> 00:12:01,971

We look at, you know, questions

like there's a famous question called

230

00:12:01,971 --> 00:12:06,434

the halting problem in Computer science

that basically is a

231

00:12:06,767 --> 00:12:10,271

we now know is a unsolvable question

for a computer system.

232

00:12:10,730 --> 00:12:13,733

And so, you know, it's not without

233

00:12:13,899 --> 00:12:17,361

computing is not a finished business,

right?

234

00:12:17,361 --> 00:12:19,029

That we, you know,

we know how to do everything

235

00:12:19,029 --> 00:12:20,114

as long as we have enough power.

236

00:12:20,114 --> 00:12:23,325

Enough, enough systems, enough

electricity.

237

00:12:23,909 --> 00:12:27,538

we don't we

and we can prove that we don't.

238

00:12:27,538 --> 00:12:29,623

Not only that, but never will.

239

00:12:29,623 --> 00:12:29,832

Right.

240

00:12:29,832 --> 00:12:32,042

There's you know,

mathematicians have worked on that.

241

00:12:32,042 --> 00:12:36,130

That there are some fundamental,

fundamentally

242

00:12:36,130 --> 00:12:40,134

unknowable things

in computation in the technical fields.

243

00:12:41,844 --> 00:12:44,305

and so I

think those are you're going to encounter

244

00:12:44,305 --> 00:12:48,768

some of those questions in the growth of

AI based on its ontology, where you can't

245

00:12:49,435 --> 00:12:53,063

there is not a, it is bound by the laws

246

00:12:53,063 --> 00:12:56,358

that that all physical,

247

00:12:56,817 --> 00:13:00,321

digital systems are bound by.

248

00:13:00,696 --> 00:13:00,863

Right?

249

00:13:00,863 --> 00:13:03,866

It can only go so fast to go

and get so hot before it breaks down.

250

00:13:04,200 --> 00:13:06,243

It's limited by the laws of physics.

251

00:13:06,243 --> 00:13:10,331

It it is also limited by the ingenuity

in which humans can insert into it.

252

00:13:10,414 --> 00:13:10,664

Right?

253

00:13:10,664 --> 00:13:13,793

We create it's steps, but

254

00:13:15,377 --> 00:13:16,712

we are limited and finite.

255

00:13:16,712 --> 00:13:19,715

So I, I have a difficult time

256

00:13:20,090 --> 00:13:22,927

imagining how a system is going to,

257

00:13:22,927 --> 00:13:26,597

you know, by its own volition, exceed

that based on what it's created on.

258

00:13:28,015 --> 00:13:31,018

Well, is part of the the argument there.

259

00:13:31,352 --> 00:13:32,353

Especially for the.

260

00:13:32,353 --> 00:13:35,981

I don't know, either

AI optimist or AI pessimist.

261

00:13:35,981 --> 00:13:42,446

Whichever one it is that, you know,

has these very high expectations for AI.

262

00:13:42,446 --> 00:13:44,448

I suppose that it's more a question

263

00:13:44,448 --> 00:13:48,327

of your outlook, whether that makes you

an AI optimist or pessimist.

264

00:13:49,328 --> 00:13:50,371

but it's part of the

265

00:13:50,371 --> 00:13:53,332

argument that, well,

these are neural networks.

266

00:13:53,332 --> 00:13:56,335

They're functioning the same way

as the brain.

267

00:13:56,710 --> 00:13:59,713

We don't have to give it a precise

algorithm because

268

00:14:01,257 --> 00:14:04,468

it has more capacities

for self-learning and self adjustment.

269

00:14:04,468 --> 00:14:07,471

And is the

270

00:14:08,013 --> 00:14:10,975

is the expectation there

that something about

271

00:14:11,767 --> 00:14:14,770

that architecture

is going to let it get past

272

00:14:16,021 --> 00:14:19,650

what normally applies

to other computer systems or...

273

00:14:21,485 --> 00:14:22,486

It's a good question.

274

00:14:22,486 --> 00:14:25,906

There's a, I remember there was a

there was a French institute

275

00:14:25,906 --> 00:14:31,453

some years ago that had begun to try

and create neural networks with the scale

276

00:14:31,453 --> 00:14:34,874

that would attempt to approximate

some of the cognition of the human brain.

277

00:14:35,541 --> 00:14:38,544

And, you know, they had the the resources

of the French government.

278

00:14:38,544 --> 00:14:42,965

They had an immense amount

of computational power behind them.

279

00:14:43,382 --> 00:14:46,844

And even with all that,

they were able to to roughly get to,

280

00:14:46,886 --> 00:14:48,137

if I remember the number correctly.

281

00:14:48,137 --> 00:14:50,681

So don't hold me on the on the citation.

282

00:14:50,681 --> 00:14:53,684

Something like 2 to 3% of brain function.

283

00:14:54,476 --> 00:14:57,021

You know, this is immensely complex

284

00:14:57,021 --> 00:15:00,232

because network complexity is

does not grow linearly.

285

00:15:00,232 --> 00:15:01,400

It grows exponentially. Right?

286

00:15:01,400 --> 00:15:06,280

To gain, to grow a system

that can get strong enough,

287

00:15:08,407 --> 00:15:09,783

to be a big enough neural network,

288

00:15:09,783 --> 00:15:12,786

even even that is,

289

00:15:13,537 --> 00:15:16,332

is not going to give you human cognition,

290

00:15:16,332 --> 00:15:20,044

because we think about the the use cases

that we use neural networks for.

291

00:15:20,628 --> 00:15:23,881

There's, there's really there's two

major ones in machine learning.

292

00:15:23,881 --> 00:15:25,090

One is classification.

293

00:15:25,090 --> 00:15:28,636

Basically telling,

you know, you look at if you take as input

294

00:15:28,636 --> 00:15:31,931

something and your brain tells you

what the thing is, right?

295

00:15:31,931 --> 00:15:35,476

Our brains are that naturally

you have image recognition software

296

00:15:35,476 --> 00:15:37,019

or other neural networks

that do the same thing.

297

00:15:37,019 --> 00:15:40,022

They take in inputs

and they identify something.

298

00:15:40,481 --> 00:15:41,649

the other one is regression.

299

00:15:41,649 --> 00:15:44,902

It's making a, a relationship

between two quantities.

300

00:15:44,902 --> 00:15:49,323

So, and allowing you to do a sort of

what we call an in-sample prediction.

301

00:15:49,323 --> 00:15:49,573

Right.

302

00:15:49,573 --> 00:15:54,036

If you're if, you know,

you know, house size A

303

00:15:54,036 --> 00:15:56,121

and house size B

and you see something in the middle,

304

00:15:56,121 --> 00:15:58,457

what should the cost of the home be,

right?

305

00:15:58,457 --> 00:16:02,294

That's a that's a regressive,

I'm sorry, it's an example of regression.

306

00:16:02,294 --> 00:16:04,088

It's not regressive, but the,

307

00:16:05,089 --> 00:16:07,716

and those are the two main neural network

applications.

308

00:16:07,716 --> 00:16:10,803

And there's now,

there's variations on them

309

00:16:10,803 --> 00:16:13,847

now, those the base layers of those are

what's been built up

310

00:16:13,847 --> 00:16:16,725

to create these transformers

that LMS are built on.

311

00:16:16,725 --> 00:16:19,728

So, they exist and they've been extended.

312

00:16:19,937 --> 00:16:22,481

But you

313

00:16:22,481 --> 00:16:25,484

I would say we are

we are too far away from,

314

00:16:26,819 --> 00:16:28,612

actual

315

00:16:28,612 --> 00:16:32,950

human brain function to predict

that it will it will continue along

316

00:16:32,950 --> 00:16:37,746

that road, even even on any,

uninterrupted path.

317

00:16:37,746 --> 00:16:40,541

And finally get to

how the human brain works,

318

00:16:40,541 --> 00:16:44,795

there's going to be breaks

or discontinuities in the progress

319

00:16:45,045 --> 00:16:48,048

and that and who knows,

some of those may scupper the whole thing.

320

00:16:48,090 --> 00:16:48,382

Right?

321

00:16:48,382 --> 00:16:52,094

You may only be able to get so far

with the applications of AI.

322

00:16:52,094 --> 00:16:53,929

You just can't go further.

323

00:16:53,929 --> 00:16:57,307

They estimate that the cost to train

training is what's most expensive

324

00:16:57,307 --> 00:17:01,270

right now

for these systems, to go beyond GPT four.

325

00:17:01,478 --> 00:17:04,398

so for little 040,

326

00:17:04,398 --> 00:17:09,528

to GPT five, GPT 5,6,7

is trillions of dollars per iteration

327

00:17:10,154 --> 00:17:13,741

to accumulate all the data

to do the training, the electricity cost

328

00:17:13,741 --> 00:17:15,242

to run the models.

329

00:17:15,242 --> 00:17:18,245

at some point we're going

to we're going to run out of money.

330

00:17:18,787 --> 00:17:21,665

We just, you know,

humanity will either have to decide,

331

00:17:21,665 --> 00:17:23,375

okay, we're invested in this or we're not.

332

00:17:23,375 --> 00:17:26,587

We just can't keep spending that

333

00:17:26,587 --> 00:17:29,590

those amount of resources

to develop a system like this.

334

00:17:29,631 --> 00:17:32,342

so I don't, you know, fortunately,

I don't have to make.

335

00:17:32,342 --> 00:17:33,552

I'm glad I'm not in charge of an

336

00:17:33,552 --> 00:17:36,555

AI company

that I'd have to make that choice, but,

337

00:17:37,598 --> 00:17:38,849

there's,

338

00:17:38,849 --> 00:17:41,060

you know, whether

it's the actual technology, whether it's

339

00:17:41,060 --> 00:17:44,188

the mathematics behind it

or the funding to generate these things,

340

00:17:44,521 --> 00:17:47,900

any, any combination of those

can cause this thing to fail.

341

00:17:48,484 --> 00:17:51,445

So we have to

almost have the perfect storm to go up the

342

00:17:51,445 --> 00:17:54,531

the the graph of progress towards human

343

00:17:54,531 --> 00:17:57,534

cognition, at least in my view.

344

00:17:57,576 --> 00:17:59,119

Yeah. Those are helpful.

345

00:17:59,119 --> 00:18:01,872

One the point you made there at the end,

those

346

00:18:01,872 --> 00:18:03,749

literal physical constraints.

347

00:18:03,749 --> 00:18:06,752

I mean,

we have all kinds of energy available, but

348

00:18:07,044 --> 00:18:10,089

that as you try to exponentially scale

349

00:18:10,089 --> 00:18:13,092

up the amount of energy

you're literally running into.

350

00:18:13,884 --> 00:18:17,221

I don't know what the scale is,

but it's registering on

351

00:18:17,888 --> 00:18:21,141

things like grid capacity

and electricity generation capacity

352

00:18:21,141 --> 00:18:22,392

and that kind of thing. At some point.

353

00:18:23,519 --> 00:18:25,938

Not a small thing.

354

00:18:25,938 --> 00:18:29,274

no. The other thing

that was really helpful, for me

355

00:18:29,274 --> 00:18:33,862

and what you said there was

AI is really only doing at the base

356

00:18:33,862 --> 00:18:36,865

the two operations,

357

00:18:37,866 --> 00:18:40,911

classification, classifying objects.

358

00:18:41,120 --> 00:18:42,955

And it's gotten

359

00:18:42,955 --> 00:18:46,166

sophisticated at that by by training

360

00:18:47,751 --> 00:18:49,920

humans, coding examples.

361

00:18:49,920 --> 00:18:53,090

and then it going off of those examples

362

00:18:54,341 --> 00:18:57,803

and regression problems, which I'm not

363

00:18:59,471 --> 00:19:02,474

not familiar

with, the mathematics of those, but

364

00:19:03,642 --> 00:19:05,686

again, those are fairly simple operations.

365

00:19:05,686 --> 00:19:08,689

I mean, takes a lot of power

to carry them out.

366

00:19:10,607 --> 00:19:13,610

fairly simple compared to

367

00:19:13,694 --> 00:19:16,697

the human mind and living a human life.

368

00:19:19,324 --> 00:19:20,284

so, yeah,

369

00:19:20,284 --> 00:19:23,287

I think just breaking it down

and saying it does these two things

370

00:19:24,163 --> 00:19:28,000

and builds on them and

puts them to good use, to me really helps

371

00:19:28,000 --> 00:19:32,087

to see the a little bit of see the limits

or demystify things a little bit.

372

00:19:33,297 --> 00:19:34,548

Yeah.

373

00:19:34,548 --> 00:19:36,800

And I, one thing I would add in to

374

00:19:36,800 --> 00:19:39,761

is that we you see on,

375

00:19:40,012 --> 00:19:42,014

maybe on social media at points people

376

00:19:42,014 --> 00:19:46,101

some very like AI negative people who like

AI is not any good at anything.

377

00:19:46,393 --> 00:19:49,313

And they'll bring up an example

like I saw one on LinkedIn the other day

378

00:19:49,313 --> 00:19:52,566

that there was a

there was like a stone pillar in a field.

379

00:19:52,858 --> 00:19:56,236

And on one side of it

had like the rear half of a cow.

380

00:19:56,862 --> 00:19:59,031

And then the stone

pillar was maybe ten feet wide.

381

00:19:59,031 --> 00:20:02,117

And on the other end of the stone pillar

was like the head of a cow poking out.

382

00:20:02,659 --> 00:20:05,871

And now we would say like,

okay, there's two cows in the picture,

383

00:20:06,330 --> 00:20:09,750

but that like an AI system,

might put a box around it and say,

384

00:20:09,791 --> 00:20:12,169

here's one cow and its length is 15ft,

385

00:20:13,420 --> 00:20:14,171

right?

386

00:20:14,171 --> 00:20:15,714

And so

they would look at an example like that.

387

00:20:15,714 --> 00:20:18,217

And we know, humans know the error.

388

00:20:18,217 --> 00:20:20,636

We we see it and we're like, oh, okay.

389

00:20:20,636 --> 00:20:22,638

a computer system doesn't.

390

00:20:22,638 --> 00:20:26,266

And so I think those, those examples

get brought up to say,

391

00:20:26,516 --> 00:20:30,187

like AI is of

no, like it's never going to be a threat.

392

00:20:30,187 --> 00:20:34,066

It's, but I think what it points to

is there are,

393

00:20:34,733 --> 00:20:38,111

there are there are limits

and easy mistakes that AI makes,

394

00:20:38,111 --> 00:20:41,323

especially early ChatGPT days

made tons of mistakes.

395

00:20:42,032 --> 00:20:44,618

And so, but humans

396

00:20:44,618 --> 00:20:48,038

will will continue to try and sort of plug

the holes in the dam.

397

00:20:48,038 --> 00:20:48,247

Right.

398

00:20:48,247 --> 00:20:50,207

They'll we'll say, okay,

here's a, here's an error

399

00:20:50,207 --> 00:20:53,377

that we're getting ridiculed

for this thing that the system can't do.

400

00:20:53,585 --> 00:20:55,045

We'll fix the thing.

401

00:20:55,045 --> 00:20:58,840

and I'm, I'm led to believe we're going

to keep finding more of those mistakes

402

00:20:58,840 --> 00:21:00,133

that humans have to fix.

403

00:21:00,133 --> 00:21:03,679

And it'll it may get better

and better and better, but it, again,

404

00:21:03,720 --> 00:21:06,098

you're you're going to keep finding,

405

00:21:07,057 --> 00:21:10,060

reasons, reasons to make fun of it

ultimately.

406

00:21:10,352 --> 00:21:11,770

Now, that doesn't mean it's not powerful.

407

00:21:11,770 --> 00:21:14,856

It's not something to think about,

but it's, you know, those

408

00:21:15,107 --> 00:21:21,196

I think the, the, the satirical view of AI

is becoming more and more prevalent.

409

00:21:21,196 --> 00:21:23,240

And I don't know what that's going to do

to people's view of it.

410

00:21:23,240 --> 00:21:24,950

probably very little.

411

00:21:24,950 --> 00:21:29,079

But, we want I just want to

I want to try and think clearly about

412

00:21:29,121 --> 00:21:33,125

is this a like

what is the truth here about AI? Yes.

413

00:21:33,125 --> 00:21:34,835

That's a silly mistake that it made.

414

00:21:34,835 --> 00:21:37,713

But does that invalidate

the whole project?

415

00:21:37,713 --> 00:21:38,880

Well, of course not. Right.

416

00:21:38,880 --> 00:21:41,967

That's you know,

amongst the billions of things it can do.

417

00:21:42,009 --> 00:21:44,886

Here's one that you thought was

silly. All right. So move on.

418

00:21:46,763 --> 00:21:48,890

Yeah, exactly.

419

00:21:48,890 --> 00:21:51,893

I think just to pursue that a little bit,

though,

420

00:21:53,103 --> 00:21:56,648

and. Yeah,

I like the case you've made, kind of,

421

00:21:57,899 --> 00:21:59,818

limits and it's limited.

422

00:21:59,818 --> 00:22:02,821

but would you agree that we can get

423

00:22:03,780 --> 00:22:07,075

we can get a lot better

at using some of the tools even without

424

00:22:08,410 --> 00:22:11,413

even without really fundamental changes

in capacity.

425

00:22:11,663 --> 00:22:14,374

like, for example, I was,

426

00:22:14,374 --> 00:22:17,252

listening to an accountant

427

00:22:17,252 --> 00:22:19,713

and his statement was, there's

428

00:22:19,713 --> 00:22:22,716

going to come a time sooner or later where

429

00:22:23,175 --> 00:22:26,261

you're basic bookkeeping,

categorizing transactions, and so on.

430

00:22:27,554 --> 00:22:30,349

Somebody is going to come up with a tool

that does that well enough

431

00:22:30,349 --> 00:22:34,895

that it's barely worth your time

to have a human bookkeeper go through

432

00:22:36,688 --> 00:22:38,815

in detail, because it's going to get it

433

00:22:38,815 --> 00:22:41,151

close enough, especially for the purposes

of small business.

434

00:22:41,151 --> 00:22:42,527

It's going to get it close enough that

435

00:22:43,862 --> 00:22:46,365

it's not going to matter.

436

00:22:46,365 --> 00:22:48,575

And I don't know, that kind of strikes

me as plausible.

437

00:22:48,575 --> 00:22:50,327

It also doesn't strike me

as requiring anything

438

00:22:50,327 --> 00:22:54,081

radically new from AI,

maybe just some human ingenuity

439

00:22:54,081 --> 00:22:57,209

and how to how to apply it

right to the process.

440

00:22:58,752 --> 00:22:59,086

I don't know.

441

00:22:59,086 --> 00:23:02,089

Does that sounds like a fair estimate.

442

00:23:02,672 --> 00:23:03,298

Oh, I think so.

443

00:23:03,298 --> 00:23:06,551

I think there's, like, accounting

as an application of AI.

444

00:23:06,551 --> 00:23:07,719

Sure, you can do the.

445

00:23:07,719 --> 00:23:10,180

The mechanics are not tremendously

complicated.

446

00:23:10,180 --> 00:23:13,308

You know, they take care

and some background knowledge, but the,

447

00:23:13,392 --> 00:23:16,520

nothing about the mathematics

is incredibly difficult.

448

00:23:17,020 --> 00:23:21,942

But the, I think, like with that example,

one of the constraints we're going to

449

00:23:21,942 --> 00:23:26,738

encounter is at the end of it, you,

you make a legally binding declaration.

450

00:23:26,947 --> 00:23:29,366

Right. These are the

these are the truth of the accounts.

451

00:23:29,366 --> 00:23:33,954

And so who who is going to

to put their legal weight behind that.

452

00:23:34,496 --> 00:23:36,873

Is it going to be the,

you know, is it going to be the individual

453

00:23:36,873 --> 00:23:39,876

who's used the system like I certify

this is correct,

454

00:23:40,001 --> 00:23:43,004

or are we going to try and pass

that off to the AI system?

455

00:23:43,588 --> 00:23:45,048

So like, well no, the AI did it.

456

00:23:45,048 --> 00:23:46,883

So if there's a mistake it isn't my fault.

457

00:23:46,883 --> 00:23:48,135

It's the AI’s fault.

458

00:23:48,135 --> 00:23:52,848

And if we do that which AI companies

are going to shoulder the legal burden

459

00:23:52,848 --> 00:23:56,685

for that of you know, that's

I think it's, it's going to sort of

460

00:23:56,685 --> 00:23:59,688

be the, the,

the fingers pointed at one another issue

461

00:23:59,855 --> 00:24:03,483

with why hasn't this become, why hasn't

this use case become widespread?

462

00:24:03,567 --> 00:24:06,736

I think it's

because no one is confident enough yet.

463

00:24:07,362 --> 00:24:12,242

That and I'll say yet because I think it

it may come a day that people will be.

464

00:24:12,242 --> 00:24:14,411

And this will the dam will break loose.

465

00:24:14,411 --> 00:24:17,080

But I don't think anyone yet

is ready to sign up

466

00:24:17,080 --> 00:24:20,542

and put their legal business life

behind this there.

467

00:24:20,625 --> 00:24:24,463

I think people are still waiting for

for more and more evidence that it's okay.

468

00:24:25,547 --> 00:24:25,797

Yeah.

469

00:24:25,797 --> 00:24:27,841

And that could apply to.

470

00:24:27,841 --> 00:24:30,635

To a lot of pieces

where you have software doing a lot of it.

471

00:24:30,635 --> 00:24:34,598

I mean, it's thinking,

you know, friends that build roof trusses

472

00:24:35,849 --> 00:24:38,685

and this isn't necessarily AI,

the software is doing basically

473

00:24:38,685 --> 00:24:42,689

the engineering and saying, yes, this,

this truss will meet specs or it won't.

474

00:24:43,899 --> 00:24:47,736

but if they're actually doing a job

that requires an engineer's stamp,

475

00:24:48,737 --> 00:24:51,364

well, it's still got to be an engineer

that does it, which is an extra step.

476

00:24:51,364 --> 00:24:54,367

That liability part.

477

00:24:54,618 --> 00:24:55,535

They're so good.

478

00:24:55,535 --> 00:24:57,579

Yeah.

479

00:24:57,579 --> 00:24:58,079

Okay.

480

00:24:58,079 --> 00:25:01,082

So I'm guessing given the arguments

you've made, that,

481

00:25:02,209 --> 00:25:05,212

you know, artificial general intelligence,

482

00:25:05,295 --> 00:25:08,298

actual purpose, purposeful

483

00:25:08,298 --> 00:25:12,177

behavior by AI systems and so on.

484

00:25:12,302 --> 00:25:15,514

I'm taking it

you're very skeptical of that.

485

00:25:17,432 --> 00:25:20,519

Or very skeptical

that being in the all things.

486

00:25:22,229 --> 00:25:23,688

Yeah, I'd say that's true.

487

00:25:23,688 --> 00:25:26,983

I have a good degree of skepticism

about that. The

488

00:25:28,276 --> 00:25:30,070

I think that we have humans

489

00:25:30,070 --> 00:25:33,073

have struggled to define

what AGI actually is.

490

00:25:33,198 --> 00:25:35,534

Right.

How do you test it. How do you verify it.

491

00:25:35,534 --> 00:25:36,326

How do you.

492

00:25:36,326 --> 00:25:40,664

And so I think that that particular

question of what is AGI is going to,

493

00:25:40,997 --> 00:25:43,959

remain in the academic circles

for a while.

494

00:25:43,959 --> 00:25:46,461

It's going to, you know, we're going to

there's going to be arguments for

495

00:25:46,461 --> 00:25:50,632

and against of various kinds

that, may or may not prove fruitful.

496

00:25:50,840 --> 00:25:51,049

Right.

497

00:25:51,049 --> 00:25:54,052

I don't know that they're going

to come to any conclusion.

498

00:25:54,970 --> 00:25:57,055

and I think in the background,

AI companies

499

00:25:57,055 --> 00:26:00,308

are going to continue to build systems

that more and more closely approximate,

500

00:26:01,059 --> 00:26:04,771

you know, human cognition,

even though they we might say

501

00:26:04,771 --> 00:26:07,482

you're still light years away,

but we're making progress.

502

00:26:07,482 --> 00:26:09,025

And that's probably true.

503

00:26:09,025 --> 00:26:11,820

And so, so I don't

504

00:26:13,196 --> 00:26:15,448

the yeah, I'm

505

00:26:15,448 --> 00:26:18,410

not tremendously worried

about the development of AGI.

506

00:26:18,618 --> 00:26:21,580

you know, sometimes the cases of,

507

00:26:21,871 --> 00:26:25,000

chess or go these board games

that have had a long and storied history

508

00:26:25,166 --> 00:26:29,546

in computation,

I like, companies like DeepMind.

509

00:26:29,546 --> 00:26:33,091

So, which is now owned by Google,

but that's out of London.

510

00:26:33,675 --> 00:26:36,970

have done some really neat work,

like developing superhuman

511

00:26:37,137 --> 00:26:40,974

chess engines and go engines that play

the game at a, at a level that surpasses

512

00:26:41,600 --> 00:26:42,809

human understanding.

513

00:26:42,809 --> 00:26:45,812

which is pretty cool, the way

what they had to develop to do that,

514

00:26:46,104 --> 00:26:49,107

but those are still games with a,

515

00:26:49,482 --> 00:26:53,737

a defined feature space

and defined rules and defined options.

516

00:26:53,737 --> 00:26:53,987

Right.

517

00:26:53,987 --> 00:26:58,074

You like when you can do that,

when you can do that, when you can box

518

00:26:58,074 --> 00:27:01,077

in the system, the universe,

you can make something really neat.

519

00:27:01,494 --> 00:27:04,998

But, boy,

humanity resists being boxed in like that.

520

00:27:05,081 --> 00:27:06,041

We just don't.

521

00:27:06,041 --> 00:27:07,042

That's not our nature.

522

00:27:07,042 --> 00:27:11,630

And and so I think part of

that is the root of my skepticism is,

523

00:27:12,130 --> 00:27:14,841

like if you look into a field

called reinforcement

524

00:27:14,841 --> 00:27:17,844

learning,

it's a it's a subpart of machine learning.

525

00:27:18,053 --> 00:27:19,804

But one of the to do it.

526

00:27:19,804 --> 00:27:24,184

Well, this is how the, the,

the chess engine from DeepMind was built.

527

00:27:26,144 --> 00:27:26,770

One of the things

528

00:27:26,770 --> 00:27:30,440

they have to do is write down, here's

the rules of the game, and here's what,

529

00:27:31,024 --> 00:27:33,985

doing a certain

good thing pays you, right?

530

00:27:33,985 --> 00:27:35,820

They call it a payoff function.

531

00:27:35,820 --> 00:27:38,823

And so if you, you know, if you take the,

the opponent's pawn little piece,

532

00:27:39,157 --> 00:27:39,824

here's the payoff.

533

00:27:39,824 --> 00:27:42,160

Or if you take their queen,

it's a bigger payoff.

534

00:27:42,160 --> 00:27:46,498

And so, if you can define that in chess,

you can actually define that.

535

00:27:47,374 --> 00:27:50,168

You can make a system

that that does very well.

536

00:27:50,168 --> 00:27:54,339

But think about a dating relationship

or a marriage like.

537

00:27:54,756 --> 00:27:57,842

And we pretty quickly

realize the our attempts

538

00:27:57,842 --> 00:28:00,845

to quantify and define value.

539

00:28:01,680 --> 00:28:04,474

It's we're in a different layer

of abstraction.

540

00:28:04,474 --> 00:28:07,477

We you can't the these don't go together.

541

00:28:07,644 --> 00:28:10,355

And so you know those are

542

00:28:10,355 --> 00:28:13,358

some of the, I guess musings to me of why

543

00:28:13,775 --> 00:28:18,029

I think AGI is still some ways off,

if we even can define it.

544

00:28:20,615 --> 00:28:22,701

Yeah.

545

00:28:22,701 --> 00:28:25,995

So along the lines of, you know, AGI.

546

00:28:25,995 --> 00:28:29,290

And again,

these questions about intelligence,

547

00:28:29,332 --> 00:28:32,335

the Turing test.

548

00:28:32,627 --> 00:28:36,423

I used to use a version of this

with students, long before,

549

00:28:37,132 --> 00:28:40,760

you know, generative

AI was on the on the horizons

550

00:28:41,886 --> 00:28:45,432

came from my,

my days in college, philosophy 101.

551

00:28:45,432 --> 00:28:49,436

And, you know, basically we're asked to

I don't have a technical definition

552

00:28:49,436 --> 00:28:52,522

in front of me,

but we're asked to consider the question,

553

00:28:54,107 --> 00:28:56,901

you know,

if a computer can give the same answers

554

00:28:56,901 --> 00:29:00,864

so that if you're communicating with it

through text or whatever, you can't tell

555

00:29:00,864 --> 00:29:03,867

if there's a computer

or a person on the other end.

556

00:29:04,993 --> 00:29:07,245

well, then should we ascribe it

557

00:29:07,245 --> 00:29:11,207

the same mental life and intelligence

that we ascribe to a person?

558

00:29:12,625 --> 00:29:13,877

yeah. How do you think about that?

559

00:29:13,877 --> 00:29:17,672

And maybe you have a more precise,

computer science definition of the test.

560

00:29:18,965 --> 00:29:20,049

No, the.

561

00:29:20,049 --> 00:29:22,510

The Turing

test was all the way back in the 1950s.

562

00:29:22,510 --> 00:29:26,264

Alan Turing, when he was working

in the early theory of computation.

563

00:29:26,681 --> 00:29:29,142

no, you

you had a good working definition.

564

00:29:29,142 --> 00:29:32,020

You know, if you're if, you know,

there's two rooms behind you

565

00:29:32,020 --> 00:29:34,981

and you're getting text

inputs from questions you ask,

566

00:29:35,106 --> 00:29:38,109

how do you tell if one is a human

and one's a computer, and if it's,

567

00:29:38,568 --> 00:29:42,155

you know, a system

that is statistically indistinguishable

568

00:29:42,155 --> 00:29:45,158

from a human, we would say, okay, passes

the Turing test.

569

00:29:45,241 --> 00:29:48,620

I think we are actually already there

with that, with the Turing test,

570

00:29:48,620 --> 00:29:52,123

I think we have systems that you can,

you can write

571

00:29:52,123 --> 00:29:56,711

fairly complex questions to,

and this is always an interesting, like,

572

00:29:56,711 --> 00:30:01,299

thought experiment of like, what question

would you ask that a, artificial

573

00:30:01,299 --> 00:30:04,427

intelligence system that you know is

an AI system would fail to answer well.

574

00:30:04,427 --> 00:30:05,637

Right.

575

00:30:05,637 --> 00:30:09,724

that's kind of an interesting

sort of a subfield here, but I think we're

576

00:30:10,099 --> 00:30:12,685

I think we already have systems

that pass the Turing test.

577

00:30:12,685 --> 00:30:16,105

The, but I think this it's

578

00:30:17,482 --> 00:30:20,276

the current application

of the Turing test, I think is a moving,

579

00:30:20,276 --> 00:30:23,947

a bit of a moving standard in that you,

580

00:30:24,531 --> 00:30:26,699

you have a system

that seems to pass the Turing test,

581

00:30:26,699 --> 00:30:29,702

and then humans get used to

how it communicates.

582

00:30:29,828 --> 00:30:30,078

Right?

583

00:30:30,078 --> 00:30:33,081

You begin to pick up the flavor

or the nuance of, like,

584

00:30:33,248 --> 00:30:35,542

this sounds like it was

AI generated, right?

585

00:30:35,542 --> 00:30:36,459

I think we all,

586

00:30:36,459 --> 00:30:39,462

if we've read AI generated text,

we sort of know what that feels like.

587

00:30:39,587 --> 00:30:42,298

It's very it's very linear, very clear.

588

00:30:42,298 --> 00:30:44,509

There's no halting pausing.

589

00:30:44,509 --> 00:30:48,680

There's it's

like it's mechanical in a sense.

590

00:30:48,680 --> 00:30:50,723

And so,

591

00:30:50,723 --> 00:30:52,684

now what an AI company might say is,

592

00:30:52,684 --> 00:30:55,687

okay,

I'll adjust my, my generation algorithm

593

00:30:55,770 --> 00:30:58,982

to make it a little more clunky,

a little more human like.

594

00:30:59,399 --> 00:30:59,649

Right.

595

00:30:59,649 --> 00:31:00,817

And then it it

596

00:31:00,817 --> 00:31:03,987

passes the new standard of the Turing test

that people can't distinguish.

597

00:31:04,404 --> 00:31:09,409

but I whenever you have a test that

relies on human perception of something

598

00:31:09,909 --> 00:31:13,997

that, like, we grow,

we learn, right, that that the benchmark

599

00:31:13,997 --> 00:31:15,540

for that test

is going to change over time.

600

00:31:15,540 --> 00:31:18,918

So, you know, I think we we have systems

601

00:31:18,918 --> 00:31:21,921

that your early GPT

is probably passed the Turing test.

602

00:31:22,380 --> 00:31:23,965

Now the standards even higher.

603

00:31:23,965 --> 00:31:24,173

Right.

604

00:31:24,173 --> 00:31:26,926

For a system

to be indistinguishable from a human.

605

00:31:26,926 --> 00:31:29,929

and it's not

because the systems have changed

606

00:31:30,138 --> 00:31:32,098

so much as the

we have changed and learned.

607

00:31:32,098 --> 00:31:34,058

And so, We've learned how to.

608

00:31:34,058 --> 00:31:35,393

How to deal with it.

609

00:31:35,393 --> 00:31:37,687

Yeah, that's a good perspective.

610

00:31:37,687 --> 00:31:38,021

And just.

611

00:31:38,021 --> 00:31:39,981

Yeah, as we're talking about this

612

00:31:39,981 --> 00:31:42,984

and it occurs to me

that it'd be very interesting to read what

613

00:31:43,151 --> 00:31:47,655

philosophers wrote about the Turing test

and how that has changed as the

614

00:31:48,865 --> 00:31:49,574

computer

615

00:31:49,574 --> 00:31:52,327

computer capacities have gone up,

because that was the angle

616

00:31:52,327 --> 00:31:55,121

from which I came,

which which I would have encountered.

617

00:31:55,121 --> 00:31:58,124

It was kind of philosophy of mind. And,

618

00:31:58,958 --> 00:32:01,210

what is it about the human mind,

619

00:32:01,210 --> 00:32:04,213

what makes mind and so on.

620

00:32:04,255 --> 00:32:07,133

And it does strike me

that those kinds of questions are really

621

00:32:07,133 --> 00:32:10,136

kind of relevant

how we're thinking about AI here, because

622

00:32:11,512 --> 00:32:14,766

if you're coming

into the whole discussion,

623

00:32:16,309 --> 00:32:19,312

with a philosophy that says.

624

00:32:21,314 --> 00:32:22,398

You know, we can explain this

625

00:32:22,398 --> 00:32:26,319

all physically or we can explain this

by the functions of physical things.

626

00:32:26,319 --> 00:32:28,655

The human mind is just

627

00:32:28,655 --> 00:32:31,950

just is the human brain,

or it's the functions

628

00:32:31,950 --> 00:32:34,953

and processes that the human brain runs.

629

00:32:35,036 --> 00:32:36,329

Then it seems to be a fair question.

630

00:32:36,329 --> 00:32:38,164

Well,

can we duplicate it with something else?

631

00:32:39,999 --> 00:32:41,167

yeah.

632

00:32:41,167 --> 00:32:42,418

If we're coming in as Christians,

633

00:32:42,418 --> 00:32:45,421

we still may have a large variety

of philosophy of mind.

634

00:32:46,047 --> 00:32:49,050

We're going to believe

that the brain is important.

635

00:32:50,885 --> 00:32:52,929

but generally,

636

00:32:52,929 --> 00:32:54,097

you know, God is a spirit.

637

00:32:54,097 --> 00:32:57,100

God has a mind without having a brain.

638

00:32:57,308 --> 00:33:00,311

generally we believe that our mind is.

639

00:33:00,979 --> 00:33:03,481

In a certain sense,

independent of our brain,

640

00:33:03,481 --> 00:33:06,484

although obviously it functions

through our brain.

641

00:33:06,693 --> 00:33:08,653

yeah.

642

00:33:08,653 --> 00:33:12,657

I just have to wonder how much that plays

into the whole discussion.

643

00:33:14,534 --> 00:33:17,245

if you simply are a materialist,

644

00:33:17,245 --> 00:33:20,039

then it seems like

so kind of naturally lead to this

645

00:33:20,039 --> 00:33:23,042

more AI optimistic view of things.

646

00:33:24,043 --> 00:33:26,754

I don't know how that maps

into the academic landscape

647

00:33:26,754 --> 00:33:29,757

either, but.

648

00:33:30,174 --> 00:33:30,466

There's.

649

00:33:30,466 --> 00:33:31,509

There's certainly

650

00:33:31,509 --> 00:33:35,263

in the academic world, there's

certainly a sense of techno optimism.

651

00:33:36,222 --> 00:33:37,223

And it's,

652

00:33:37,223 --> 00:33:40,727

And, you know, they treat it again

if you if you come at it

653

00:33:40,727 --> 00:33:45,064

from a largely materialist worldview,

you end up with just,

654

00:33:46,774 --> 00:33:48,234

you know, an understanding that

655

00:33:48,234 --> 00:33:52,697

that all I'm, all I'm creating technically

is a smaller approximation to me.

656

00:33:52,697 --> 00:33:55,491

And it's going to get closer

and closer and closer over time.

657

00:33:55,491 --> 00:34:00,079

And so there's no there isn't really

a connection between soul, spirit

658

00:34:00,079 --> 00:34:03,374

and mind in the, in the secular worldview

659

00:34:03,958 --> 00:34:06,586

because we're just,

660

00:34:06,586 --> 00:34:06,794

you know,

661

00:34:06,794 --> 00:34:10,757

we're just a random collection of atoms

that, that have no lasting eternal value.

662

00:34:10,757 --> 00:34:13,760

I think that's the conclusion,

663

00:34:14,385 --> 00:34:16,554

whereas I think like

that is from a Christian perspective.

664

00:34:16,554 --> 00:34:21,350

That's where we come at AI

with a different view of like, you know.

665

00:34:21,559 --> 00:34:21,893

Yeah.

666

00:34:21,893 --> 00:34:26,105

So you can create a neural network that is

that has the trillions of connections

667

00:34:26,105 --> 00:34:27,523

that our brain does.

668

00:34:27,523 --> 00:34:30,526

you're still missing a piece

that you can't simulate.

669

00:34:30,693 --> 00:34:34,530

There's a spirit,

there's a soul, there's a, humans are,

670

00:34:35,031 --> 00:34:38,076

you know, by a mathematical definition,

irrational creatures.

671

00:34:39,243 --> 00:34:40,286

But in

672

00:34:40,286 --> 00:34:43,915

if we because we are we, you know,

we look at we do things that are silly,

673

00:34:43,956 --> 00:34:48,419

not in our best interest,

but in from a theological perspective,

674

00:34:48,419 --> 00:34:51,798

that makes sense

because we, we identify this conflict

675

00:34:51,798 --> 00:34:55,009

of our of our sin nature

and our redeemed nature in Christ.

676

00:34:55,009 --> 00:35:00,515

And so, if you don't have theology, this

gets leads to a natural techno optimism.

677

00:35:00,890 --> 00:35:04,602

But I think only in the,

in the sense that it is self beneficial.

678

00:35:05,186 --> 00:35:05,394

Right.

679

00:35:05,394 --> 00:35:09,482

Like I'm optimistic technically,

if it benefits me, right?

680

00:35:09,482 --> 00:35:13,444

if it doesn't, if it's a threat to me,

then I'm not as optimistic about it.

681

00:35:13,444 --> 00:35:15,822

And I have no moral qualms either

way. Right.

682

00:35:15,822 --> 00:35:17,490

That's the materialist view.

683

00:35:17,490 --> 00:35:20,368

It's not a moral question.

It's it's pragmatic.

684

00:35:21,869 --> 00:35:22,328

Yeah.

685

00:35:22,328 --> 00:35:25,581

So actually, while

I was getting ready for this episode,

686

00:35:25,623 --> 00:35:28,793

I was sitting in my little,

687

00:35:30,211 --> 00:35:33,673

office at the school I help with,

and I looked out the window

688

00:35:33,923 --> 00:35:37,176

and I noticed plants, flowers.

689

00:35:37,260 --> 00:35:39,095

It's a beautiful sunny day.

690

00:35:39,095 --> 00:35:42,014

And butterflies flitting around

691

00:35:42,014 --> 00:35:43,891

on top of there.

692

00:35:43,891 --> 00:35:46,978

And. Yeah, so I'm drawn into the beauty.

693

00:35:47,353 --> 00:35:49,230

Something a lot of people are.

694

00:35:49,230 --> 00:35:53,109

And so just got me thinking, you know,

those are the things we paint.

695

00:35:53,985 --> 00:35:56,696

we draw pictures of,

696

00:35:56,696 --> 00:36:00,658

But not only that, like, we really have,

we really have dug into that.

697

00:36:00,658 --> 00:36:03,035

What is the life cycle of the flower?

698

00:36:03,035 --> 00:36:06,956

How can we breed the flower

to maximize blooms?

699

00:36:07,957 --> 00:36:10,960

You know, we've traced the life cycle

of those butterflies

700

00:36:11,419 --> 00:36:14,422

and learned how they function.

701

00:36:14,422 --> 00:36:18,176

learned the biology of how they function,

the life cycle, all of that.

702

00:36:19,135 --> 00:36:20,136

and, you know, I even had

703

00:36:20,136 --> 00:36:23,431

to think about a book we're using in our,

in our science program.

704

00:36:24,056 --> 00:36:26,601

It's called The Girl Who Drew Butterflies,

705

00:36:26,601 --> 00:36:29,812

and it tells the story of a girl

who was a painter.

706

00:36:30,730 --> 00:36:32,857

Her art, but just was really

707

00:36:32,857 --> 00:36:35,818

was really closely observing,

708

00:36:35,943 --> 00:36:38,237

you know, insects, butterflies

came to understand

709

00:36:38,237 --> 00:36:41,240

these stages of metamorphosis and so on.

710

00:36:41,240 --> 00:36:45,328

when a lot of people, at a time when

a lot of people didn't understand those,

711

00:36:46,412 --> 00:36:49,081

and so just kind of really hit me like,

okay,

712

00:36:49,081 --> 00:36:52,501

this is what humans do with things

that we see and pay attention to.

713

00:36:53,419 --> 00:36:58,549

is there any reason to think that AI

714

00:37:00,426 --> 00:37:03,012

models or agents,

715

00:37:03,012 --> 00:37:06,390

would do any of the same thing

with have any of that same,

716

00:37:07,350 --> 00:37:07,808

really It's a

717

00:37:07,808 --> 00:37:10,770

relationship, ongoing relationship

with reality.

718

00:37:11,354 --> 00:37:15,024

and does that get at any

of the difference, this kind of difference

719

00:37:15,024 --> 00:37:19,320

we're talking about between a human mind

and an AI model?

720

00:37:20,613 --> 00:37:20,863

yeah.

721

00:37:20,863 --> 00:37:23,866

I'd love to hear

some of your thoughts on that.

722

00:37:24,825 --> 00:37:27,828

I love the question. I,

723

00:37:28,955 --> 00:37:29,914

I mean, my my thoughts.

724

00:37:29,914 --> 00:37:33,668

First, go to the the the the genesis

of what we would call curiosity.

725

00:37:34,293 --> 00:37:36,212

Right? The the.

726

00:37:36,212 --> 00:37:38,631

You know, I've, I have children at home.

727

00:37:38,631 --> 00:37:40,216

Some of them are very curious.

728

00:37:40,216 --> 00:37:42,260

And what is it that.

And we love curiosity.

729

00:37:42,260 --> 00:37:46,222

It's how we we develop an ever

expanding knowledge of the world.

730

00:37:47,556 --> 00:37:49,976

you know, so on the one hand, I think AI

731

00:37:49,976 --> 00:37:52,979

could be constrained

and that it has to like

732

00:37:54,021 --> 00:37:57,316

it doesn't really have the ability

to explore aside from the digital space

733

00:37:57,316 --> 00:37:57,900

it inhabits.

734

00:37:57,900 --> 00:38:02,822

And so, you know, it could but, you know,

that could be that could be overcome

735

00:38:03,155 --> 00:38:06,826

if a, you know, a human, we we give it

inputs from a world beyond its own

736

00:38:06,826 --> 00:38:10,955

or we give it pictures and soil samples

and scientific reports, and we give it

737

00:38:10,955 --> 00:38:15,084

enough information to, discern connections

and patterns and systems.

738

00:38:16,627 --> 00:38:16,919

but I

739

00:38:16,919 --> 00:38:20,798

think at the, at the very bottom of a

of a contrived system

740

00:38:20,798 --> 00:38:24,176

like that, we have like human volition,

we've done it.

741

00:38:24,176 --> 00:38:28,764

We've,

we we have to push AI into the world

742

00:38:28,764 --> 00:38:30,766

to begin to collect this information.

743

00:38:30,766 --> 00:38:33,769

so I don't know that,

744

00:38:34,270 --> 00:38:35,187

you know what?

745

00:38:35,187 --> 00:38:37,857

I'll say this in the, in the world

AI inhabits,

746

00:38:37,857 --> 00:38:42,153

I think it already is acting

as a curious agent within its world.

747

00:38:42,737 --> 00:38:42,945

Right.

748

00:38:42,945 --> 00:38:45,364

It's I think by its incentive structure.

Right.

749

00:38:45,364 --> 00:38:47,867

It says like,

I want to be the best AI system.

750

00:38:47,867 --> 00:38:52,580

So I'm going to collect everything

that I can, and, and look for patterns

751

00:38:52,580 --> 00:38:57,585

and, and begin to create more

and more of an approximation to reality.

752

00:38:58,711 --> 00:39:01,297

there's so much more to the world

than just the digital space, right?

753

00:39:01,297 --> 00:39:05,301

So we have to we have to somehow connect

AI with that space.

754

00:39:05,301 --> 00:39:07,845

That's a more challenging question.

755

00:39:07,845 --> 00:39:09,180

so I don't know that I would,

756

00:39:11,265 --> 00:39:12,224

I wouldn't

757

00:39:12,224 --> 00:39:15,853

I don't think that

AI is going to have agency in that way.

758

00:39:16,187 --> 00:39:19,315

I think one of the problems where, like,

I think we're already seeing an inflection

759

00:39:19,315 --> 00:39:22,318

point with that kind of pattern in AI,

760

00:39:22,818 --> 00:39:24,779

because they estimate the amount

of the amount of data

761

00:39:24,779 --> 00:39:27,990

on the planet doubles every six months,

something like that.

762

00:39:27,990 --> 00:39:31,369

So just the amount of data is exploding

based on what we're creating

763

00:39:31,369 --> 00:39:34,372

movies, films, text, all these things.

764

00:39:35,247 --> 00:39:36,290

now what we're realizing

765

00:39:36,290 --> 00:39:40,211

is that the some of the data

that's being fed back into these models

766

00:39:40,211 --> 00:39:43,339

to learn and remain current

is actually AI generated.

767

00:39:43,798 --> 00:39:43,964

Right?

768

00:39:43,964 --> 00:39:47,301

So AI models are consuming

AI generated things

769

00:39:47,802 --> 00:39:51,472

and taking it as as new truth sources.

770

00:39:51,472 --> 00:39:51,722

Right?

771

00:39:51,722 --> 00:39:55,559

So they're the they're building up on what

they, what they've already created.

772

00:39:56,060 --> 00:39:59,647

And so any, you know,

if there's any error or bias

773

00:39:59,647 --> 00:40:02,650

or anything that's built

into those generated sets of data

774

00:40:02,900 --> 00:40:07,947

that's now been replicated

and virally expanded across the AI space.

775

00:40:07,947 --> 00:40:13,244

And so, so I think, you know,

I my one one thought I have

776

00:40:13,244 --> 00:40:16,330

is there's going to be a bit of a pendulum

swing when you realize AI

777

00:40:16,330 --> 00:40:19,333

is spitting out nonsense,

because that's all it knows.

778

00:40:19,834 --> 00:40:21,919

and the pendulum swing will say, well,

don't

779

00:40:21,919 --> 00:40:25,089

don't let AI have access to anything

new yet.

780

00:40:25,089 --> 00:40:25,464

Right?

781

00:40:25,464 --> 00:40:29,593

We need to curate what it's learning from,

in order to make this truthful.

782

00:40:29,718 --> 00:40:35,099

And so, you know, we have this curious

humans have this beautiful

783

00:40:35,099 --> 00:40:39,770

curiosity to know, you know what,

why do things work the way that they do?

784

00:40:40,062 --> 00:40:40,271

Right.

785

00:40:40,271 --> 00:40:44,358

It's led to advances in cosmology

into in biology, in mathematics.

786

00:40:44,817 --> 00:40:47,027

You know, we do this naturally.

787

00:40:48,028 --> 00:40:48,654

you know, I

788

00:40:48,654 --> 00:40:52,116

loved I loved seeing recently

there was a paper that came out of,

789

00:40:52,700 --> 00:40:53,534

I think it might have been Google

790

00:40:53,534 --> 00:40:56,537

DeepMind as well,

but they actually had developed a

791

00:40:56,537 --> 00:40:59,832

AI generated mathematical theorem

proving system.

792

00:41:00,416 --> 00:41:01,709

They could give it

a mathematical question.

793

00:41:01,709 --> 00:41:03,961

It could prove something mathematically.

794

00:41:03,961 --> 00:41:08,048

I said, that's fascinating because that's

that's an area that I had thought

795

00:41:08,048 --> 00:41:12,344

that humans like that was human

like creativity and intuition.

796

00:41:12,344 --> 00:41:14,513

That was not an accessible region.

797

00:41:14,513 --> 00:41:17,433

but I had but the more I am

798

00:41:17,433 --> 00:41:20,436

reading about it,

the more I'm thinking, okay,

799

00:41:20,686 --> 00:41:24,356

they've not done human things,

but they've they've made progress, right?

800

00:41:24,398 --> 00:41:26,650

They've done some things that I thought

they couldn't do.

801

00:41:26,650 --> 00:41:30,488

And so, so there but I think that's

how progress oftentimes goes in

802

00:41:30,488 --> 00:41:33,532

AI is they, they,

it appears they can't do anything.

803

00:41:33,532 --> 00:41:37,453

And then there's this discontinuity

and jump into now what they can do.

804

00:41:37,453 --> 00:41:40,748

So I think we've seen those

now do I think that

805

00:41:41,332 --> 00:41:44,335

that AI is ever going to have the,

the beautiful,

806

00:41:44,710 --> 00:41:47,796

observation of the butterflies,

like the girl who drew butterflies

807

00:41:47,796 --> 00:41:50,883

to to understand

the biological cycles of these things.

808

00:41:53,177 --> 00:41:55,804

again, I probably

hold a healthy skepticism, just like AGI.

809

00:41:55,804 --> 00:41:56,388

I'm not.

810

00:41:56,388 --> 00:41:59,558

I don't think that AI has, because

we haven't given it the reason to yet.

811

00:41:59,767 --> 00:42:02,269

We've not incentivized it.

812

00:42:02,269 --> 00:42:04,897

you know, we

we learn because it's beautiful to learn

813

00:42:04,897 --> 00:42:07,900

and AI learns because either we tell it to

814

00:42:08,526 --> 00:42:10,444

or we threaten its extinction

if it doesn't.

815

00:42:10,444 --> 00:42:13,739

So there's a, you know, I don't think

816

00:42:13,739 --> 00:42:16,116

we've

it doesn't have the natural curiosity.

817

00:42:16,116 --> 00:42:19,119

We do. And I think that makes us human.

818

00:42:19,870 --> 00:42:21,789

Yeah. Thanks. Yeah.

819

00:42:21,789 --> 00:42:24,375

So I keep hearing you

come back to the, “Not yets” a little bit.

820

00:42:24,375 --> 00:42:27,586

But if there is one theme in your limits

that you talk about,

821

00:42:28,379 --> 00:42:31,382

kind of ontological limits of AI is

822

00:42:33,008 --> 00:42:36,428

is it always being downstream of humanity

in some way or another?

823

00:42:37,179 --> 00:42:39,223

Yeah. We don't understand

exactly what it's doing.

824

00:42:39,223 --> 00:42:43,561

We we set it up and we don't

always understand exactly how it works.

825

00:42:43,561 --> 00:42:45,646

The whole black box effect.

826

00:42:45,646 --> 00:42:47,982

but it's downstream from you talking about

827

00:42:47,982 --> 00:42:51,360

the incentives we give it, the way

we program it, the parameters,

828

00:42:52,403 --> 00:42:54,697

what we want it to do,

829

00:42:54,697 --> 00:42:57,324

the inputs we curate for it.

830

00:42:57,324 --> 00:43:00,869

so yeah, that is helpful.

831

00:43:01,579 --> 00:43:03,372

yeah.

832

00:43:03,372 --> 00:43:06,375

Any particular concluding thoughts

you'd like with,

833

00:43:06,875 --> 00:43:09,878

like to, leave on this topic?

834

00:43:10,254 --> 00:43:11,338

I think just the.

835

00:43:11,338 --> 00:43:14,091

You know,

you're you're right point to highlight.

836

00:43:14,091 --> 00:43:16,719

Like. Yeah. So there is a,

there is a sense of not yet.

837

00:43:16,719 --> 00:43:21,015

And again part of me

I go there because I in candor,

838

00:43:21,015 --> 00:43:24,018

I don't actually know,

but I make a forecast that could have a,

839

00:43:24,268 --> 00:43:26,145

you know,

there's a margin of error in the forecast

840

00:43:26,145 --> 00:43:29,356

and there's things that could go wrong

or right that make it wrong.

841

00:43:29,732 --> 00:43:31,442

And so,

842

00:43:31,442 --> 00:43:33,986

I only forecast out as sort of as far

843

00:43:33,986 --> 00:43:38,449

as I could see, the, but you think about

something like quantum computing.

844

00:43:38,449 --> 00:43:41,577

So the new technology that there's

engineering

845

00:43:41,577 --> 00:43:44,830

challenges to solve, but say we solve them

and we have quantum computers

846

00:43:45,247 --> 00:43:48,250

that can do incredibly quick calculation

847

00:43:48,792 --> 00:43:51,795

in, in some previously inaccessible

problems.

848

00:43:52,379 --> 00:43:55,758

all of these developments,

you know, AI developments, growth

849

00:43:55,758 --> 00:43:59,720

in neural networks or training or,

you know, GPT eight say.

850

00:44:00,012 --> 00:44:01,764

Or quantum computing, right.

851

00:44:01,764 --> 00:44:06,226

They all solve sort of little niche things

and they do it very well.

852

00:44:06,727 --> 00:44:10,856

But I guess the analogy

I go back to in my own mind is that you

853

00:44:11,398 --> 00:44:14,610

without it, without a unified system

or understanding of reality,

854

00:44:14,985 --> 00:44:17,905

all of these things will remain

niche solutions, right?

855

00:44:17,905 --> 00:44:18,697

You don't.

856

00:44:18,697 --> 00:44:20,115

It'd be a little bit like, you know, you

857

00:44:20,115 --> 00:44:22,868

you try and fix something wrong

with your body and you, you know, you,

858

00:44:22,868 --> 00:44:25,663

you sprain your finger

and so you splint it.

859

00:44:25,663 --> 00:44:27,915

Right. You have a solution to the problem.

860

00:44:27,915 --> 00:44:29,750

You've not cured cancer, right?

861

00:44:29,750 --> 00:44:32,002

You've you've solved a problem.

862

00:44:32,002 --> 00:44:32,211

Right.

863

00:44:32,211 --> 00:44:34,963

And you've made progress

towards the solution to the whole.

864

00:44:34,963 --> 00:44:37,758

But you

there is not a systemic understanding yet.

865

00:44:37,758 --> 00:44:41,261

And so all of these developments,

866

00:44:42,471 --> 00:44:44,264

you know, I'll say not yet.

867

00:44:44,264 --> 00:44:46,433

My my prediction is it's a long way off.

868

00:44:46,433 --> 00:44:51,146

But, I think for all of these things

to work in concert to develop anything

869

00:44:51,146 --> 00:44:56,068

that might be, an ontological break

for artificial intelligence,

870

00:44:56,902 --> 00:44:59,238

just this to me, there's still too

many unanswered questions.

871

00:44:59,238 --> 00:45:03,117

Is they don't none of these things

see reality in the same way.

872

00:45:03,117 --> 00:45:06,412

Like quantum and classical computing

also don't see reality in the same way.

873

00:45:06,412 --> 00:45:10,499

So like you gotta find a way to bridge

that before these are effective together.

874

00:45:10,499 --> 00:45:11,709

And it's

875

00:45:11,709 --> 00:45:13,460

I think that there's

876

00:45:13,460 --> 00:45:16,338

we're finding more questions

than we have answers to these days.

877

00:45:16,338 --> 00:45:19,258

And at least in my view.

878

00:45:19,258 --> 00:45:20,592

That's an interesting way

maybe to put your.

879

00:45:20,592 --> 00:45:21,969

Not yet.

880

00:45:21,969 --> 00:45:24,304

You said we won't get to an on. You know,

881

00:45:25,347 --> 00:45:26,056

you don't think we'll get

882

00:45:26,056 --> 00:45:29,059

to an ontological breakthrough.

883

00:45:29,143 --> 00:45:30,227

and so in some ways,

884

00:45:30,227 --> 00:45:33,230

what you're saying is, look,

the ontology of what we have now is,

885

00:45:34,440 --> 00:45:37,443

is, in reality, nothing like artificial

general intelligence.

886

00:45:39,153 --> 00:45:42,156

You can't conclusively rule out

that at some point,

887

00:45:43,323 --> 00:45:45,242

we would be able to develop systems

888

00:45:45,242 --> 00:45:48,537

that are really fundamentally different

at how they work.

889

00:45:49,580 --> 00:45:51,498

And that's kind of the not yet.

890

00:45:51,498 --> 00:45:55,252

And as part of your emphasis is like

it wouldn't just be a development,

891

00:45:55,252 --> 00:45:58,255

it would be something

would be a very fundamental breakthrough.

892

00:45:59,256 --> 00:46:01,300

Yeah, I'd say that's a good

that's a good assessment of it.

893

00:46:01,300 --> 00:46:04,094

It's it's very hard to prove a negative.

894

00:46:04,094 --> 00:46:06,138

Right. So that's one of the challenges

you run into in logic.

895

00:46:06,138 --> 00:46:11,477

And so and it,

so yeah, I think we have there's

896

00:46:12,603 --> 00:46:13,312

and it always

897

00:46:13,312 --> 00:46:16,607

seems, it always seems impossible

898

00:46:16,607 --> 00:46:19,651

to solve some of these problems

until someone solves it.

899

00:46:19,651 --> 00:46:21,320

Right.

There's that's true of mathematics, right?

900

00:46:21,320 --> 00:46:24,323

We have all these unsolved problems

until someone,

901

00:46:24,615 --> 00:46:27,075

you know, goes out of mathematics

for a long enough

902

00:46:27,075 --> 00:46:28,202

and they come up with a solution.

903

00:46:28,202 --> 00:46:33,582

So, I think, you know, my

my skepticism lies in the, in the,

904

00:46:33,582 --> 00:46:37,169

the nature of the questions being asked in

that, you know, we're saying,

905

00:46:37,795 --> 00:46:41,131

you know, in order to create things

that make us uniquely human, we need to,

906

00:46:41,423 --> 00:46:46,178

you know, more and more silicon chips

doesn't get us there, right?

907

00:46:46,386 --> 00:46:47,846

It needs to be something else. Right?

908

00:46:47,846 --> 00:46:51,683

Some other way of understanding reality,

processing information,

909

00:46:53,143 --> 00:46:54,019

that we just don't have.

910

00:46:54,019 --> 00:46:57,147

And I, and again, I don't think too long

911

00:46:57,147 --> 00:47:00,359

about what

it would take to do that, but the,

912

00:47:01,568 --> 00:47:02,486

When we reach a point in

913

00:47:02,486 --> 00:47:05,864

innovation that every time we innovate,

we get more questions.

914

00:47:06,782 --> 00:47:08,909

I'm, I'm waiting for the

915

00:47:08,909 --> 00:47:12,704

I'm not really waiting, but in a sense,

if the tide began to turn and the

916

00:47:12,871 --> 00:47:15,707

we began to answer, a lot of these things

or major breakthroughs are covering

917

00:47:15,707 --> 00:47:19,294

lots of them that might tilt the needle

of optimism versus pessimism.

918

00:47:19,294 --> 00:47:22,923

For me, I might say, okay, I'm

maybe I was in error before,

919

00:47:23,298 --> 00:47:28,428

but at the moment we just keep

getting more questions and it's yeah,

920

00:47:28,428 --> 00:47:31,473

that's that's sort of the root of my

the root of my not yet answer.

921

00:47:34,142 --> 00:47:34,476

Yeah.

922

00:47:34,476 --> 00:47:38,230

Well,

thank you, Ben, for, diving into this.

923

00:47:38,230 --> 00:47:40,148

I've enjoyed this and. Yeah.

924

00:47:40,148 --> 00:47:42,067

Thanks for sharing that.

925

00:47:42,067 --> 00:47:44,069

with our audience here.

926

00:47:44,069 --> 00:47:47,072

Thank you to the audience

927

00:47:47,072 --> 00:47:50,075

for tuning in here

to Anabaptist Perspectives.

928

00:47:50,576 --> 00:47:54,288

And if you've enjoyed this,

you can subscribe to the channel

929

00:47:54,288 --> 00:47:56,248

or share this episode.

930

00:47:56,248 --> 00:47:58,041

And we'll catch you in the next episode.