Craig Tucker: Why Your AI Feels Cold and How Emotional Intelligence Changes Everything

1

:: 00:02

Good morning, good afternoon, good evening wherever you are in

2

:: 00:05

the world. I'm here with Craig, my good friend. He

3

:: 00:09

is on the bleeding edge of AI. He's doing things

4

:: 00:13

that are going to change the world. Has inspired me

5

:: 00:17

every time we've spoken. Craig is working with AI avatars,

6

:: 00:21

but he's working with emotion. He's working with human language.

7

:: 00:24

He's working with a much deeper understanding. Welcome, Craig. It's

8

:: 00:28

really good to have you on the podcast. Thank you,

9

:: 00:30

Roy. I appreciate it. Always nice to see you. And

10

:: 00:33

I think it's probably worthwhile starting with how this started

11

:: 00:37

for you. Not going that far back, but yeah, this

12

:: 00:41

current journey, it all started for me because my grandpa

13

:: 00:44

died, Grandma was lonely and I thought I would be

14

:: 00:46

the good Grandson. Build an AI companion. Was in a

15

:: 00:50

PhD program. Michigan State University. Thought I'm the right one

16

:: 00:53

for this. Right. And along the way found the real

17

:: 00:56

key was emotions and how we communicate. Did patented on

18

:: 00:59

that were a couple years of validating and here we

19

:: 01:02

are. That story with your grandma resonates with so many

20

:: 01:06

people. People are going to be sitting there saying but

21

:: 01:08

really, AI, it's not good enough. So tell us why

22

:: 01:11

it's different. Sure. Most AI that we've all experienced has

23

:: 01:15

been some kind of a chatbot. Right? The chatbots that

24

:: 01:18

we get are fine and good, but they just only

25

:: 01:20

do a certain thing or they act a certain way.

26

:: 01:22

That's always the same. They. You get the same robotic

27

:: 01:25

type of response back. It doesn't seem personal, it doesn't

28

:: 01:29

connect with you. And you're on other sides of what

29

:: 01:31

we call the uncanny valley. Right. Area in which you

30

:: 01:34

know that you don't really accept a technology because it's

31

:: 01:37

not quite human enough or it's too human and you've

32

:: 01:41

got to find a bridge or a way across that.

33

:: 01:44

We detect the emotions that are communicated between people. It's

34

:: 01:47

part of our language, the structure of communication, both on

35

:: 01:51

the subtextual level and also on the contextual level. It

36

:: 01:54

becomes important to use these clues and signals to inform

37

:: 01:59

any type of AI that has to interact with a

38

:: 02:01

human being because we have feelings and what we interact

39

:: 02:04

with should understand that. Otherwise it just becomes those chatbots

40

:: 02:08

with five options that you yell at. I know I

41

:: 02:10

do when I'm trying to book my flight or get

42

:: 02:13

some help from the utilities. Really that's the why and

43

:: 02:15

what we do and what makes it different. So you

44

:: 02:17

can use that in enterprise for customer service, wayfinding or

45

:: 02:21

in mental health, help people connect or help people try

46

:: 02:25

to get better. Even since I've known you probably six

47

:: 02:29

months, the technology has just moved on so much and

48

:: 02:32

I think it's a good time for us to give

49

:: 02:35

a demonstration of one of your avatars. And Greg's got

50

:: 02:38

a whole bunch of different avatars. He has a set

51

:: 02:40

of humans that you can go and play with and

52

:: 02:42

we'll share some of that. And literally the interactions you

53

:: 02:46

have with them is so different. Let's have a chat.

54

:: 02:49

Let me get everything going here and I'll bring everybody

55

:: 02:52

in. So what we're going to see is a personal

56

:: 02:55

shopper. Her name is Carrie and we created her to

57

:: 02:58

help out with a shopping transaction and we used a

58

:: 03:02

lot of real world experience because we're all online shoppers

59

:: 03:05

as well. Let's go ahead and talk to Carrie and

60

:: 03:07

see what we can do. Hi, I'm Carrie.

61

:: 03:12

Tell me what you're shopping for today. Yeah, I'm looking

62

:: 03:16

for some pants. What do you have in pants today?

63

:: 03:25

For pants we currently have a few options. The Tiller

64

:: 03:28

Jean Essential Ashland Pant Vintage fit Double front canvas pant

65

:: 03:33

and the Ashland Bib. Let me know if any of

66

:: 03:35

these catch your eye and we can dive into the

67

:: 03:37

details. Yeah, Can I see the Essential Ashland, please? Here

68

:: 03:45

are. One second, I'll check. Would you like some insight

69

:: 03:50

into how others have felt about wearing the Essential Ashland

70

:: 03:52

pants? I can show you some customer reviews that highlight

71

:: 03:56

the fit and durability. If you're interested from the reviews.

72

:: 03:59

People have been praising the Essential Ashland pants for their

73

:: 04:02

comfort and versatility. Many have noted they hold up well

74

:: 04:06

in both casual and and more rugged environments. Let me

75

:: 04:09

know if you're leaning towards a particular color or size.

76

:: 04:13

Yeah, I like the black one. The

77

:: 04:19

black Essential Ashland Pant is a solid choice. Just let

78

:: 04:22

me know your size preference. Yeah. What sizes do they

79

:: 04:25

come in? The Essential Ashland Pant

80

:: 04:30

is available in sizes from extra small to double extra

81

:: 04:33

large. Just let me know which size fits you best.

82

:: 04:36

Excellent. I'll do a large. Please add that to cart.

83

:: 04:45

Great. I've added the black Essential Ashland Pant. If there's

84

:: 04:48

anything else you need or if you're ready to check

85

:: 04:50

out, just let me know. Yeah, Carrie, let me see

86

:: 04:54

some long sleeve shirts that you'd recommend that would go

87

:: 04:56

with those pants. I

88

:: 05:02

can see we have the Dearborn Loose LS logo T

89

:: 05:05

shirt and Durham knit overshirt among others that could match

90

:: 05:08

those black pants beautifully. Would you like more details on

91

:: 05:11

either of those? Yeah, the Dearborn Loose. Let me see

92

:: 05:13

that please. Hold

93

:: 05:19

on a second. Here's

94

:: 05:26

the Dearborn Loose long sleeve logo T shirt. It has

95

:: 05:29

a relaxed fit and is available in a few colors

96

:: 05:31

that could pair well with your black pants. Like the

97

:: 05:33

smoke teal green or the classic black. Yes. Let's do

98

:: 05:37

a large teal and add that to cart, please. I've

99

:: 05:45

added the Dearborn Loose long sleeve logo T shirt in

100

:: 05:48

smoked, teal, green and large size to your cart. Is

101

:: 05:51

there anything else you'd like to explore or another item

102

:: 05:54

you're considering? We'll mute her right now. Toss it back

103

:: 05:57

over to you, Roy. She's a big shopping assistant. Match

104

:: 06:00

and pair things together, build carts and then be able

105

:: 06:04

to answer customer questions. That usually would prevent somebody from

106

:: 06:07

adding stuff to cart or completing a transaction. So helps

107

:: 06:11

to speed that up. The friction's gone and does it

108

:: 06:14

in a very personal way. A character that people can

109

:: 06:16

relate to. Obviously she's emotionally intelligent too. So I was

110

:: 06:20

to have a meltdown about one of the sizes or

111

:: 06:22

anything else. She could respond to me in real time

112

:: 06:25

and get me over. Let's try that. Let's be a

113

:: 06:27

little bit irritated about something and see what. Sure. Excellent.

114

:: 06:30

Let me go ahead and see that now. Carrie, I'm

115

:: 06:32

really frustrated and frankly, I don't know what the hell

116

:: 06:34

you're doing. I did not want to see any of

117

:: 06:37

those T shirts. I really wanted to see some pants.

118

:: 06:45

Sorry for any misunderstanding. Let's get back to the pants.

119

:: 06:48

Is there a specific type of pants or style you're

120

:: 06:51

interested in seeing again or something I can clarify for

121

:: 06:53

you? Yeah, I just want to see the pants, please.

122

:: 07:01

Let me see. I've got the pants collection up for

123

:: 07:06

you. Take a look and see which styles might fit

124

:: 07:08

your needs better this time. Thank you for your patience.

125

:: 07:12

As you can see, she handles the frustration and the

126

:: 07:15

anger that I'm showing her and being able to not

127

:: 07:19

comply. So does her best. She apologizes. Basically mirror the

128

:: 07:23

user so that they can understand that they understand what's

129

:: 07:27

going on and that they're trying to please them. I

130

:: 07:30

think this is something that's missing currently from technology today

131

:: 07:33

and is essential for a really good full range shopping

132

:: 07:37

experience. You don't want your chatbot to be an anger

133

:: 07:40

machine. So somebody's emotions actually help guide the conversation so

134

:: 07:44

the customer doesn't spend too much time being angry or

135

:: 07:48

fearful. Let's go back to chatting. Sure. So

136

:: 07:54

sitting where you're sitting at the moment, Craig, you're seeing

137

:: 07:59

so much opportunity in so many different places. You're early

138

:: 08:03

to this game. Have you seen the change in how

139

:: 08:06

people are viewing AI in the last six months? Yeah,

140

:: 08:09

absolutely, Roy. I think it changes week to week. It's

141

:: 08:12

comes something being coined as the speed of AI and

142

:: 08:16

we're obviously early adopters. We started doing this stuff 2015,

143

:: 08:20

2016, right. So we've been around a while doing this

144

:: 08:24

stuff and seeing the hockey stick of progression is just

145

:: 08:27

through the chart. It's exponential now because AI really enables

146

:: 08:31

that. Right? Enables productivity to compound exponentially. Knowledge

147

:: 08:36

and information to compound exponentially. And I think that it

148

:: 08:40

really enables everybody to have a little bit of a

149

:: 08:43

superhuman power. And from what we're seeing in our perspective

150

:: 08:47

is it is disruptive. It can be disruptive, but in

151

:: 08:52

a good way. It solves some of the biggest problems

152

:: 08:55

that humanity has right now. Jobs we don't want to

153

:: 08:58

do that, tax us that are mentally draining, that leave

154

:: 09:02

us with feeling terrible and as a psychological burden on

155

:: 09:06

us. There are a lot of jobs that are like

156

:: 09:08

that. Customer service support gives a really good example. You

157

:: 09:12

don't call a customer service or support because you're having

158

:: 09:15

a good time, right. Usually it's a problem and you're

159

:: 09:18

not happy about it, right. So you're getting a mental

160

:: 09:22

dump of all these people's problems on any csr. So

161

:: 09:26

burnout becomes a problem. Churning of employees becomes a problem

162

:: 09:30

that Verm can help solve by putting AI there to

163

:: 09:33

do that first level stuff to connect with a human

164

:: 09:36

being to make them feel like they're hurt. And we

165

:: 09:38

always tell everybody, Roy, that we want to be like

166

:: 09:42

Bill Barr, right? Or Bob Barr, rather. In Incredibles, in

167

:: 09:45

the first scene when the old lady trying to get

168

:: 09:47

help and the insurance company can't. And Bob basically tells

169

:: 09:50

her. Exactly. The guy on the inside tells her how

170

:: 09:53

to cut all the red tape and achieve her goals.

171

:: 09:55

And she's extraordinarily happy and satisfied. That's what we want

172

:: 09:59

Vern to be for the user is to act like

173

:: 10:03

they're the person on the inside that they can trust,

174

:: 10:05

get them the information they need to accomplish their goals,

175

:: 10:09

and then good things come from that. And it's such

176

:: 10:12

a risky journey for companies, isn't it? And this is

177

:: 10:16

what you're struggling with because you've seen bigger companies laying

178

:: 10:20

off 5,000 people because AI is going to do it,

179

:: 10:23

and then realizing that, not bringing the emotion to it,

180

:: 10:26

and the clients are going away. Exactly. And another thing

181

:: 10:30

too is there's a term called AI washing, which I

182

:: 10:32

think is really true. A lot of these companies that

183

:: 10:34

are shedding people added them gratuitously during the pandemic. If

184

:: 10:39

you look at all the hiring that meta Microsoft, you

185

:: 10:42

name any one of these companies that were trying to

186

:: 10:44

expand and grow, seeing work from home being an opportunity.

187

:: 10:48

So they added a lot of headcount towards risks and

188

:: 10:51

bets that they were making. Some of them didn't come

189

:: 10:54

true. There's a gravitational pullback into community and offices and

190

:: 10:58

in person. So you're seeing some of that is a

191

:: 11:02

direct result of they just had so much overhead. Some

192

:: 11:05

of that is you're also seeing too from insiders is

193

:: 11:08

we're getting rid of these people because we think AI

194

:: 11:10

can fill the gap. We don't know and it's your

195

:: 11:12

job to figure it out. That's a recipe for disaster.

196

:: 11:16

It absolutely is. And it will be doomed to fail.

197

:: 11:18

The incentives aren't aligned. And what we see where AI

198

:: 11:22

can do the most good is some of the most

199

:: 11:25

sensitive places, like within mental health. Let's talk about that

200

:: 11:30

a little bit, because that's something. People are going to

201

:: 11:33

be sitting here, not having had the experience with your

202

:: 11:36

avatars. There's got to be so many guardrails. How do

203

:: 11:39

you take somebody that's nearly suicidal and put an AI

204

:: 11:44

model in front of them? What's happening at the moment,

205

:: 11:46

which we start seeing in the news and it starts

206

:: 11:48

getting blown up, is people are using ChatGPT as a

207

:: 11:51

service because that's all they've got, right? Yeah, it's not

208

:: 11:54

responsible either. And the big problems is with large language

209

:: 11:58

models, there's no consensus on what emotions are in psychology,

210

:: 12:02

so they don't know which one to use at any

211

:: 12:04

given moment and usually mix them up and give you

212

:: 12:06

a mixture of theories, frameworks, you're basically getting slop.

213

:: 12:11

And people are taking this as good psychotherapy because it

214

:: 12:15

sounds good. They don't know any better and they shouldn't

215

:: 12:18

have to know better. But the difference is in what

216

:: 12:20

Vern allows is we don't tell or we don't have

217

:: 12:24

the AI. The large language models guess what the emotions

218

:: 12:28

are. They know, we annotate them. We're a measurement tool.

219

:: 12:32

You'll get the same answer, same input every single time.

220

:: 12:35

If not, it's not a measurement tool, you're guessing. So

221

:: 12:38

being able to measure everything gives the AI a very

222

:: 12:42

good idea of the motionality within the scenario. And so

223

:: 12:46

what we do is we use the emotion to guide

224

:: 12:49

the conversation. Right. High anger, repetitively, certain

225

:: 12:55

types of emotional expression leads to a triage logic

226

:: 13:01

which would then help to put everything else aside and

227

:: 13:05

address the person's issues. This is something we pioneered back

228

:: 13:09

in 23. We won the bot a thon at core

229

:: 13:11

AI with this module that could detect when somebody was

230

:: 13:14

angry and wanted to speak to a manager. It would

231

:: 13:16

calm them down, figure out what the problem is, buy

232

:: 13:19

some time, and then put in a transfer to a

233

:: 13:21

live agent. That type of concept of triage is something

234

:: 13:25

you can do with Vern and it's why we are

235

:: 13:28

trusted in deployments that are very sensitive. You've had some

236

:: 13:32

really interesting deployments and was there an autistic deployment that

237

:: 13:36

you told me about? Yes. Yeah. Yeah. Hello, My Friend

238

:: 13:38

is an app for kids with autism or neurodiversity. Have

239

:: 13:42

issues with understanding their own emotions. With Vern, being able

240

:: 13:45

to identify, label emotions for them allows the avatars to

241

:: 13:49

practice with the kids, to review things in their life

242

:: 13:53

and their world that they're confused about because they may

243

:: 13:55

not understand the emotions. The AI avatars can be age

244

:: 13:59

appropriate, gender appropriate. So anything from young kids to teens,

245

:: 14:03

young adults can connect with them and look like them

246

:: 14:08

and be able to discuss whatever that they want within

247

:: 14:12

the respective frameworks using the appropriate psychological therapy.

248

:: 14:17

I think it's fascinating. It's one of the biggest things

249

:: 14:21

that's missing in the game at the moment. Without a

250

:: 14:24

shadow of a doubt. I talk so much about keeping

251

:: 14:27

the human and raising the human to the top. If

252

:: 14:31

you've got a staff of 10 people and those 10

253

:: 14:34

people are doing 70% of their work and as repetitive,

254

:: 14:37

boring tasks and 30% of their work that they like

255

:: 14:41

doing, can you ship that 70% off to the AI

256

:: 14:45

and give them the chance to spend more time connecting,

257

:: 14:48

communicating, engaging with clients? And that I think is going

258

:: 14:52

to be the telling company in 2026 and beyond. It's

259

:: 14:57

not the company that just automates everything and lets it

260

:: 15:00

go. It's a company that lets human come to the

261

:: 15:03

surface and understands how important that connection is with people.

262

:: 15:07

And I think this goes a long way to addressing

263

:: 15:10

and honoring that humanity. We couldn't be here without the

264

:: 15:14

humanity. And AI wouldn't be worthy to us if they

265

:: 15:17

weren't serving human ends. I think that's critically important that

266

:: 15:20

we keep the person. And I hear all the time,

267

:: 15:23

what about Terminator? But what people tend to Forget is

268

:: 15:26

in T2 he was reprogrammed with empathy, set back in

269

:: 15:29

time to save humanity. Right. So if we don't teach

270

:: 15:33

AI how to understand feelings and what they mean to

271

:: 15:36

human beings, how they're communicated, then we're doomed to the

272

:: 15:39

same fate. Right? Hopefully they'll see the beauty in humanity

273

:: 15:43

and how we communicate and express ourselves as worthy as

274

:: 15:47

a 0 and a 1. And so having that translation

275

:: 15:50

layer, what VERN allows, allows for the AI to understand

276

:: 15:55

humans, and we've found for humans to understand themselves better.

277

:: 15:59

I think one of the biggest lessons I've had in

278

:: 16:01

this whole journey is with the current large language models,

279

:: 16:05

it allowed me to communicate with my people better because

280

:: 16:10

as I'm putting the prompts in and I'm getting rubbish

281

:: 16:14

out, I'm thinking what's not clear? And then you analyze

282

:: 16:17

yourself and you say, actually let me read what I

283

:: 16:19

just typed in there or talked in there. Ah, this

284

:: 16:22

is how it could have been translated. I need to

285

:: 16:25

be much more clear. And yeah, in my business relationships,

286

:: 16:28

in my personal relationships, in all of them, I feel

287

:: 16:31

that's given me an uplift in being more clear. 101

288

:: 16:37

of the things that we've really found is obviously working

289

:: 16:40

with these emotional tools is that we detect emotions a

290

:: 16:45

lot better. So our personal relationships have gotten better. But

291

:: 16:49

I will say there is such thing as an emotional

292

:: 16:51

container contagion, long theorized, never proven. We've got some paperwork

293

:: 16:54

on it, so we'll show it. But yeah, you have

294

:: 16:58

to also be careful about surrounding yourself with one emotion

295

:: 17:02

or another. Because when we were doing study and building

296

:: 17:05

out these models, the team members that were working on

297

:: 17:08

say anger started to be more angry and sad. Similar

298

:: 17:14

experience, Love and Joy. I don't know how to and

299

:: 17:17

if they were a little bit more spunky or not,

300

:: 17:20

but it definitely seemed when anybody working on Love and

301

:: 17:22

Joy was a lot happier and we had less issues.

302

:: 17:25

The emotional signals you surround yourself with do have a

303

:: 17:29

profound impact on you. We're going to prove that out

304

:: 17:32

in the next few years. So where do you see

305

:: 17:35

this all going? What do you see? The journey over

306

:: 17:36

26. Yeah. For us, this year really is about enabling

307

:: 17:42

folks to create interface with AI that's responsible, that's

308

:: 17:48

governed so that they can trust. Right. It's all about

309

:: 17:51

building trust with AI. It's not going away. It's definitely

310

:: 17:54

a tool. Using it and deploying it is very important,

311

:: 17:58

obviously all the way up in the board level. But

312

:: 18:00

having the ability to put a layer in between the

313

:: 18:03

human and the AI to ensure that both sides play

314

:: 18:05

nice. Right. I think what we're doing and for us

315

:: 18:08

this year now down the road, I see as what

316

:: 18:12

everybody wants to have is the bat computer or Jarvis,

317

:: 18:15

is the AI that is remarkably human, but still computerized

318

:: 18:20

and digital and obviously has Immense amount of resources.

319

:: 18:25

That's basically the whole world at your fingertips concept and

320

:: 18:30

that assistant that's there to help you achieve your goals.

321

:: 18:33

So I think that's the path everybody's on and we're

322

:: 18:37

really happy to be able to enable a big part

323

:: 18:39

of that. Yeah, and it's really this piece for me,

324

:: 18:42

this doom and gloom beforehand, you know, now we're moving

325

:: 18:46

into a change environment, I think, where people are accepting

326

:: 18:51

and they're saying, okay, how can I use this? Hopefully

327

:: 18:53

we will get to a stage where everything is just

328

:: 18:56

done so much more efficiently, that freedom actually becomes a

329

:: 19:00

reality for so many people. Right. Food's going to be

330

:: 19:03

created, the farming is going to be better, medical is

331

:: 19:06

going to be better. The access of information to all

332

:: 19:08

the people in the world that can't get it is

333

:: 19:10

going to get be better. We understand and be able

334

:: 19:13

to communicate much more. And there's going to be a

335

:: 19:16

change with that time in the near future where more

336

:: 19:20

people are going on holidays, more people are spending time

337

:: 19:22

with the families, more people are exploring the world and

338

:: 19:25

seeing the beauty out there. And that'll because of AI.

339

:: 19:28

Right. And it's the same glide path we've been on

340

:: 19:31

since the Enlightenment. Right. That steady increase of productivity that

341

:: 19:36

leads to more leisure time, more pursuits of things that

342

:: 19:40

interest us and that have uncalculable value. Priceless.

343

:: 19:46

I saw a couple of things and then we'll end

344

:: 19:48

after that. I saw a couple of things on. Yeah,

345

:: 19:50

I think it was aluminum like you guys. Aluminium, as

346

:: 19:54

we said, where the MIT scientists put the AI on

347

:: 19:58

the task of finding an alloy that was stronger. And

348

:: 20:01

in one afternoon they were able to test 300,000 different

349

:: 20:05

possible molecules on it. And they came out with something

350

:: 20:07

that was way better. And if you just go and

351

:: 20:10

apply that to all the other problems humanity has, we

352

:: 20:13

can start solving them and start fixing them. We're never

353

:: 20:16

going to run out of problems. There's always going to

354

:: 20:19

be use for people. Yeah, for sure. It's just going

355

:: 20:22

to change. And every major shift in the world, every

356

:: 20:25

printing press and every bit of electricity and all these

357

:: 20:27

things have all taken us further. There's been a change.

358

:: 20:31

We've adapted. We found out what to do with it

359

:: 20:33

and then we've moved on. And I just. Yeah, I'm

360

:: 20:35

waiting for the day where your day becomes going out

361

:: 20:38

and connecting with people and having deep and meaningful conversations

362

:: 20:42

and understand emotions better. And I just think that's where

363

:: 20:45

we're going to go. Somewhere, Socrates, Plato, Aristotle are sitting

364

:: 20:48

around laughing because that's what they saw as the world

365

:: 20:52

being in the future is one day when man can

366

:: 20:55

step aside from the toils of everyday labor and really

367

:: 20:59

look at existence, science, philosophy and why we're here and

368

:: 21:04

where we're going. I love it man. Awesome. Thank you

369

:: 21:06

for joining me. It's been inspiring talking to you. I'll

370

:: 21:09

put your contact details so people can get hold of

371

:: 21:10

you and look forward to chatting in a few months

372

:: 21:13

time and see where you've got to. Sounds good Roy.

373

:: 21:17

Thanks for having me.