1
00:00:00,000 --> 00:00:04,440
Okay, so this time we're looking at, well, it's a lot.

2
00:00:04,440 --> 00:00:08,080
Tesla robots, AI teachers, nuclear power plants

3
00:00:08,080 --> 00:00:09,800
that are built like, you know, Legos.

4
00:00:09,800 --> 00:00:10,640
Yeah, it's-

5
00:00:10,640 --> 00:00:12,520
I mean, you sent over some articles and announcements

6
00:00:12,520 --> 00:00:13,360
and it seems like-

7
00:00:13,360 --> 00:00:14,180
I think not.

8
00:00:14,180 --> 00:00:15,960
It's gonna be a challenge to like sort through the hype

9
00:00:15,960 --> 00:00:17,240
and figure out what's real.

10
00:00:17,240 --> 00:00:20,600
Yeah, separating the hype from the hope

11
00:00:20,600 --> 00:00:24,120
and the actual progress and what it all means.

12
00:00:24,120 --> 00:00:25,220
Exactly. Yeah.

13
00:00:25,220 --> 00:00:26,840
So let's jump right in.

14
00:00:26,840 --> 00:00:30,000
Let's start with Tesla's WeRobot event.

15
00:00:30,000 --> 00:00:30,840
Okay.

16
00:00:30,840 --> 00:00:32,960
They showed off their new Cybertruck

17
00:00:32,960 --> 00:00:35,200
and CyberVamp prototypes driverless, of course.

18
00:00:35,200 --> 00:00:39,720
But are these vehicles really ready to change transportation

19
00:00:39,720 --> 00:00:41,960
or are we just looking at another round of hype?

20
00:00:41,960 --> 00:00:44,020
Well, you know, Tesla's not exactly known

21
00:00:44,020 --> 00:00:46,960
for underselling their products.

22
00:00:46,960 --> 00:00:49,560
So, you know, you may recall Elon Musk

23
00:00:50,560 --> 00:00:53,400
promised a million robo-taxis by 2020.

24
00:00:53,400 --> 00:00:54,240
Great.

25
00:00:54,240 --> 00:00:56,360
Of course, that didn't quite pan out.

26
00:00:56,360 --> 00:00:57,200
Right.

27
00:00:57,200 --> 00:00:58,160
But it'll be interesting to see

28
00:00:58,160 --> 00:01:00,800
how these autonomous features actually perform

29
00:01:00,800 --> 00:01:02,580
in real world testing.

30
00:01:02,580 --> 00:01:03,420
Right.

31
00:01:03,420 --> 00:01:04,560
And especially since those cyber calves

32
00:01:04,560 --> 00:01:05,960
are supposed to be on the road next year.

33
00:01:05,960 --> 00:01:07,120
Yeah.

34
00:01:07,120 --> 00:01:09,760
And what about those Optimus robot demonstrations?

35
00:01:09,760 --> 00:01:10,600
Yeah.

36
00:01:10,600 --> 00:01:13,800
Cool moves, but were those actual AI advancements?

37
00:01:13,800 --> 00:01:14,640
Right.

38
00:01:14,640 --> 00:01:16,080
Or just really good programming?

39
00:01:16,080 --> 00:01:18,000
Yeah, that's the question.

40
00:01:18,880 --> 00:01:20,200
I mean, from what I saw,

41
00:01:20,200 --> 00:01:22,020
it seemed like the Optimus demonstrations

42
00:01:22,020 --> 00:01:26,840
relied more on preprogrammed routines than on any kind

43
00:01:26,840 --> 00:01:27,880
of groundbreaking AI.

44
00:01:27,880 --> 00:01:28,760
I mean, don't get me wrong,

45
00:01:28,760 --> 00:01:31,480
building a robot that can dance and wave,

46
00:01:31,480 --> 00:01:32,740
it's no small feat.

47
00:01:32,740 --> 00:01:35,500
But we're still a ways away from, you know,

48
00:01:35,500 --> 00:01:37,640
robots that can really think for themselves.

49
00:01:37,640 --> 00:01:39,960
So more spectacle than substance then.

50
00:01:39,960 --> 00:01:42,480
So the real question is, does it even matter

51
00:01:42,480 --> 00:01:44,680
if the technology isn't really there yet?

52
00:01:44,680 --> 00:01:46,560
I mean, is Tesla's enthusiasm,

53
00:01:46,560 --> 00:01:48,640
is it doing more harm than good?

54
00:01:48,640 --> 00:01:50,360
Yeah, that's the question, isn't it?

55
00:01:50,360 --> 00:01:51,400
Yeah.

56
00:01:51,400 --> 00:01:52,720
I mean, on the one hand,

57
00:01:52,720 --> 00:01:55,120
that enthusiasm does drive innovation.

58
00:01:55,120 --> 00:01:58,040
It pushes the limits of what's possible.

59
00:01:58,040 --> 00:02:01,160
But on the other hand, if those promises don't come true,

60
00:02:01,160 --> 00:02:04,200
if they fall short consistently,

61
00:02:04,200 --> 00:02:07,260
then it could affect how the public views

62
00:02:07,260 --> 00:02:08,800
self-driving tech as a whole.

63
00:02:08,800 --> 00:02:11,360
Yeah, and trust is a big deal when we're talking about,

64
00:02:11,360 --> 00:02:14,000
you know, letting a machine drive us around.

65
00:02:14,000 --> 00:02:14,840
Absolutely.

66
00:02:14,840 --> 00:02:17,800
So let's switch gears from self-driving cars

67
00:02:17,800 --> 00:02:21,120
to something a little closer to home, to the classroom.

68
00:02:21,120 --> 00:02:22,640
We got a case study here in Nevada

69
00:02:22,640 --> 00:02:24,940
where they started using an AI system

70
00:02:24,940 --> 00:02:26,960
to figure out which students were at risk.

71
00:02:26,960 --> 00:02:27,800
Right.

72
00:02:27,800 --> 00:02:30,000
And the idea was to make sure resources

73
00:02:30,000 --> 00:02:31,840
were being used well.

74
00:02:31,840 --> 00:02:33,880
But the result was a little more complicated than that.

75
00:02:33,880 --> 00:02:36,280
Yeah, it is interesting how the system was perceived.

76
00:02:36,280 --> 00:02:39,280
So on paper, this big drop in the number of students

77
00:02:39,280 --> 00:02:42,220
flagged as at risk, it might seem like a success,

78
00:02:42,220 --> 00:02:43,920
like great, less need.

79
00:02:43,920 --> 00:02:44,760
Right.

80
00:02:44,760 --> 00:02:47,360
But it really meant big cuts for these support programs

81
00:02:47,360 --> 00:02:49,720
that a lot of schools and students really needed.

82
00:02:49,720 --> 00:02:51,400
And then suddenly, we're not just

83
00:02:51,400 --> 00:02:52,840
talking about algorithms and data.

84
00:02:52,840 --> 00:02:56,240
We're talking about real students and educators

85
00:02:56,240 --> 00:02:58,520
just trying to work in a system that's not

86
00:02:58,520 --> 00:02:59,640
working for them anymore.

87
00:02:59,640 --> 00:03:01,860
Right, and that's where it gets really interesting

88
00:03:01,860 --> 00:03:04,600
because it's not just about the technology anymore.

89
00:03:04,600 --> 00:03:07,640
It's about how we're using it to make decisions that have

90
00:03:07,640 --> 00:03:09,600
very real consequences for people.

91
00:03:09,600 --> 00:03:11,120
It really shows how important it is

92
00:03:11,120 --> 00:03:14,440
to think about not just what we want these systems to do,

93
00:03:14,440 --> 00:03:16,920
but also what might happen that we don't expect.

94
00:03:16,920 --> 00:03:17,720
Absolutely.

95
00:03:17,720 --> 00:03:18,320
Yeah.

96
00:03:18,320 --> 00:03:22,280
But speaking of technology that could really shake things up,

97
00:03:22,280 --> 00:03:25,520
let's move from the classroom to, well, a really

98
00:03:25,520 --> 00:03:26,520
big construction site.

99
00:03:26,520 --> 00:03:27,380
OK.

100
00:03:27,380 --> 00:03:29,800
You sent over some info on Blue Energy

101
00:03:29,800 --> 00:03:31,760
and how they're building nuclear power plants.

102
00:03:31,760 --> 00:03:34,240
Yeah, what's really fascinating about Blue Energy

103
00:03:34,240 --> 00:03:36,140
is that they're tackling the biggest problem

104
00:03:36,140 --> 00:03:37,780
with nuclear power head on.

105
00:03:37,780 --> 00:03:39,640
And that problem is cost.

106
00:03:39,640 --> 00:03:40,320
OK.

107
00:03:40,320 --> 00:03:42,780
So instead of building these plants on site,

108
00:03:42,780 --> 00:03:46,280
which everyone knows takes forever and costs a fortune,

109
00:03:46,280 --> 00:03:48,560
they're using a modular approach.

110
00:03:48,560 --> 00:03:49,640
OK.

111
00:03:49,640 --> 00:03:51,240
So picture this.

112
00:03:51,240 --> 00:03:54,020
Giant prefab pieces shipped to the site

113
00:03:54,020 --> 00:03:56,960
and then put together a really advanced LEGO set.

114
00:03:56,960 --> 00:03:58,760
So they're kind of using the same playbook

115
00:03:58,760 --> 00:03:59,640
as shipbuilders.

116
00:03:59,640 --> 00:04:01,180
That's actually really smart.

117
00:04:01,180 --> 00:04:03,880
But how much cheaper could this make nuclear power?

118
00:04:03,880 --> 00:04:06,680
And what would that mean for clean energy in general?

119
00:04:06,680 --> 00:04:08,200
Well, they say they can build plants

120
00:04:08,200 --> 00:04:11,120
for a fraction of the cost of traditional nuclear

121
00:04:11,120 --> 00:04:14,440
construction, even competitive with fossil fuel plants.

122
00:04:14,440 --> 00:04:14,920
Wow.

123
00:04:14,920 --> 00:04:16,300
If they can actually pull that off,

124
00:04:16,300 --> 00:04:18,400
it'd be a total game changer.

125
00:04:18,400 --> 00:04:21,800
I mean, imagine clean and reliable energy

126
00:04:21,800 --> 00:04:24,200
that doesn't cost more than fossil fuels.

127
00:04:24,200 --> 00:04:27,040
That's a huge incentive to go green and get away

128
00:04:27,040 --> 00:04:28,680
from carbon-based energy.

129
00:04:28,680 --> 00:04:29,880
Yeah, definitely.

130
00:04:29,880 --> 00:04:32,320
It's a big idea for sure.

131
00:04:32,320 --> 00:04:35,600
Turning the economics of nuclear power on its head.

132
00:04:35,600 --> 00:04:39,200
But while Blue Energy is out there tackling the energy grid,

133
00:04:39,200 --> 00:04:41,960
let's talk about something a little more, well, personal.

134
00:04:41,960 --> 00:04:44,480
Let's talk about the AI-powered shopping experience.

135
00:04:44,480 --> 00:04:45,000
OK.

136
00:04:45,000 --> 00:04:46,120
What does that even mean?

137
00:04:46,120 --> 00:04:50,640
So imagine you walk into a store and you're greeted by name.

138
00:04:50,640 --> 00:04:53,280
And not in a creepy way, but because the AI system knows

139
00:04:53,280 --> 00:04:55,360
you, it pulls up your shopping history.

140
00:04:55,360 --> 00:04:59,160
No more generic coupons, just personalized recommendations

141
00:04:59,160 --> 00:05:01,120
based on stuff you actually buy.

142
00:05:01,120 --> 00:05:04,220
OK, that is a little creepy, but also kind of amazing.

143
00:05:04,220 --> 00:05:08,120
My grocery store can barely keep cilantro and stock.

144
00:05:08,120 --> 00:05:10,040
And they're going to be able to predict

145
00:05:10,040 --> 00:05:11,800
every craving I have.

146
00:05:11,800 --> 00:05:13,160
How does that even work?

147
00:05:13,160 --> 00:05:15,120
What's all about the data?

148
00:05:15,120 --> 00:05:19,280
AI algorithms look at these huge data sets, your purchase

149
00:05:19,280 --> 00:05:22,920
history, what you look at online, even stuff on social media

150
00:05:22,920 --> 00:05:25,400
to make this whole profile of what you like.

151
00:05:25,400 --> 00:05:27,320
So it's not even just what I bought last week.

152
00:05:27,320 --> 00:05:31,800
It's like predicting what I might want

153
00:05:31,800 --> 00:05:34,280
based on everything I do.

154
00:05:34,280 --> 00:05:37,600
That's powerful, but isn't too powerful.

155
00:05:37,600 --> 00:05:40,400
Where's the line between personalized and digital

156
00:05:40,400 --> 00:05:41,440
stalker?

157
00:05:41,440 --> 00:05:42,680
That's the question, isn't it?

158
00:05:42,680 --> 00:05:44,880
See, the tech itself isn't good or bad.

159
00:05:44,880 --> 00:05:46,840
It's all about how people use it.

160
00:05:46,840 --> 00:05:48,880
Transparency is really important here.

161
00:05:48,880 --> 00:05:51,880
People need to know what data is being collected,

162
00:05:51,880 --> 00:05:54,040
how it's being used, and we need rules

163
00:05:54,040 --> 00:05:56,920
to make sure this technology doesn't get too nosy or even

164
00:05:56,920 --> 00:05:57,960
worse discriminatory.

165
00:05:57,960 --> 00:06:01,080
Because imagine if the AI starts deciding

166
00:06:01,080 --> 00:06:04,000
what you should buy based on your age, your income, even

167
00:06:04,000 --> 00:06:05,760
your race.

168
00:06:05,760 --> 00:06:07,240
That's a recipe for disaster.

169
00:06:07,240 --> 00:06:08,000
Exactly.

170
00:06:08,000 --> 00:06:09,480
It's a lot to think about.

171
00:06:09,480 --> 00:06:11,880
But before we go too far down the rabbit hole

172
00:06:11,880 --> 00:06:14,680
of targeted ads, let's switch gears one last time.

173
00:06:14,680 --> 00:06:15,880
OK.

174
00:06:15,880 --> 00:06:17,800
You've flied some really interesting research

175
00:06:17,800 --> 00:06:21,600
about the reasoning abilities of large language models, LLMs,

176
00:06:21,600 --> 00:06:24,360
for those of us who don't speak AI.

177
00:06:24,360 --> 00:06:25,840
What are we even talking about here?

178
00:06:25,840 --> 00:06:31,080
Think of LLMs as the brains behind those chatbots,

179
00:06:31,080 --> 00:06:34,600
the ones that can write poems, music, even

180
00:06:34,600 --> 00:06:39,640
have conversations that sound like a real person.

181
00:06:39,640 --> 00:06:40,560
Right.

182
00:06:40,560 --> 00:06:43,320
They're incredibly complicated, trained

183
00:06:43,320 --> 00:06:48,400
on tons of data, which lets them create human quality text based

184
00:06:48,400 --> 00:06:49,600
on all sorts of prompts.

185
00:06:49,600 --> 00:06:52,280
So they're basically just really fancy parrots repeating

186
00:06:52,280 --> 00:06:53,280
what they've been trained on.

187
00:06:53,280 --> 00:06:53,960
In a way, yeah.

188
00:06:53,960 --> 00:06:55,760
I mean, that's impressive, but it's not really thinking,

189
00:06:55,760 --> 00:06:56,240
is it?

190
00:06:56,240 --> 00:06:57,560
That's the big question.

191
00:06:57,560 --> 00:06:59,680
While LLMs are great at sounding human,

192
00:06:59,680 --> 00:07:01,320
this new research suggests that they

193
00:07:01,320 --> 00:07:04,000
have a hard time with even simple reasoning tasks.

194
00:07:04,000 --> 00:07:06,360
So you're saying all of that impressive stuff they do

195
00:07:06,360 --> 00:07:06,960
is just a trick?

196
00:07:06,960 --> 00:07:08,120
Kind of, yeah.

197
00:07:08,120 --> 00:07:12,800
How can something that seems so smart be fooled so easily?

198
00:07:12,800 --> 00:07:13,880
It's like this.

199
00:07:13,880 --> 00:07:17,600
Imagine you ask an LLM to write a story about a cat.

200
00:07:17,600 --> 00:07:18,960
It does a fantastic job.

201
00:07:18,960 --> 00:07:21,280
It describes the cat's soft fur, how playful it is,

202
00:07:21,280 --> 00:07:23,120
how much it loves to nap.

203
00:07:23,120 --> 00:07:25,760
But then you add a little twist to the story.

204
00:07:25,760 --> 00:07:27,280
You mention a dog.

205
00:07:27,280 --> 00:07:29,200
Suddenly the story makes no sense.

206
00:07:29,200 --> 00:07:31,200
The cat starts barking, chasing cars,

207
00:07:31,200 --> 00:07:33,160
burying bones in the yard.

208
00:07:33,160 --> 00:07:35,480
The LLM faced with information that

209
00:07:35,480 --> 00:07:38,920
doesn't fit this narrow understanding of cats and dogs.

210
00:07:38,920 --> 00:07:39,420
Yeah.

211
00:07:39,420 --> 00:07:41,000
It just falls apart.

212
00:07:41,000 --> 00:07:41,840
So much for that.

213
00:07:41,840 --> 00:07:43,080
Right.

214
00:07:43,080 --> 00:07:46,640
So are we overestimating AI?

215
00:07:46,640 --> 00:07:49,080
Should we be worried about a future where

216
00:07:49,080 --> 00:07:51,320
machines are smarter than us?

217
00:07:51,320 --> 00:07:52,520
Not exactly.

218
00:07:52,520 --> 00:07:55,400
This research doesn't mean AI can't reason,

219
00:07:55,400 --> 00:07:58,120
but it does highlight the limits of how we're approaching it.

220
00:07:58,120 --> 00:08:00,240
Now, we're still in the early stages of figuring out

221
00:08:00,240 --> 00:08:02,880
how to build AI that can truly think for itself.

222
00:08:02,880 --> 00:08:05,120
That's actually kind of a humbling thought.

223
00:08:05,120 --> 00:08:07,240
We tend to look at these machines

224
00:08:07,240 --> 00:08:10,440
and assume that because they can process information so fast,

225
00:08:10,440 --> 00:08:13,160
like we do, that they must be thinking like us.

226
00:08:13,160 --> 00:08:17,040
It's like we're trying to teach a fish to climb a tree.

227
00:08:17,040 --> 00:08:19,840
We're so busy focusing on what these LLMs can do

228
00:08:19,840 --> 00:08:22,760
that we forget to even ask, what should they be doing?

229
00:08:22,760 --> 00:08:23,960
And that's exactly it.

230
00:08:23,960 --> 00:08:25,560
As we create these tools and they're

231
00:08:25,560 --> 00:08:26,960
getting more and more complex, we

232
00:08:26,960 --> 00:08:30,160
have to make sure we're using them ethically, thoughtfully.

233
00:08:30,160 --> 00:08:31,680
Yeah, no robot overlords.

234
00:08:31,680 --> 00:08:32,180
Yeah.

235
00:08:32,180 --> 00:08:32,680
Got it.

236
00:08:32,680 --> 00:08:36,200
But seriously, how do we even make sure these technologies

237
00:08:36,200 --> 00:08:37,560
are used for good?

238
00:08:37,560 --> 00:08:39,360
It seems like we're always playing catch up,

239
00:08:39,360 --> 00:08:41,720
trying to fix the problems after they've already happened.

240
00:08:41,720 --> 00:08:43,240
It's tough, for sure.

241
00:08:43,240 --> 00:08:46,040
But I think it starts with just talking to each other,

242
00:08:46,040 --> 00:08:47,280
working together.

243
00:08:47,280 --> 00:08:50,360
We need to get experts from different areas, ethicists,

244
00:08:50,360 --> 00:08:53,000
people who make policy, the tech people themselves,

245
00:08:53,000 --> 00:08:55,640
all in the same room to figure out some ground rules,

246
00:08:55,640 --> 00:08:58,760
some regulations to make sure AI is developed responsibly.

247
00:08:58,760 --> 00:09:01,920
Some more brainstorming sessions and less of that move fast

248
00:09:01,920 --> 00:09:04,240
and break things mentality.

249
00:09:04,240 --> 00:09:04,960
Exactly.

250
00:09:04,960 --> 00:09:07,440
It's all about finding that sweet spot

251
00:09:07,440 --> 00:09:11,080
between encouraging new ideas and protecting what matters.

252
00:09:11,080 --> 00:09:15,400
Privacy, fairness, making sure people are held accountable.

253
00:09:15,400 --> 00:09:16,960
Because really, these technologies,

254
00:09:16,960 --> 00:09:17,960
they're just tools.

255
00:09:17,960 --> 00:09:18,440
Right.

256
00:09:18,440 --> 00:09:20,200
And their impact depends on us.

257
00:09:20,200 --> 00:09:21,200
Absolutely.

258
00:09:21,200 --> 00:09:23,620
We have the power to shape how AI develops.

259
00:09:23,620 --> 00:09:27,360
And it's on us to make sure that future benefits everyone,

260
00:09:27,360 --> 00:09:29,200
not just a select few.

261
00:09:29,200 --> 00:09:30,920
Well said.

262
00:09:30,920 --> 00:09:34,040
So as we wrap up our deep dive into the world of cutting edge

263
00:09:34,040 --> 00:09:36,320
tech, one thing's clear.

264
00:09:36,320 --> 00:09:38,240
Nobody knows what the future holds,

265
00:09:38,240 --> 00:09:41,120
whether it's Tesla and their robots or AI in our kids'

266
00:09:41,120 --> 00:09:44,040
classrooms or even those chat bots that seem a little too

267
00:09:44,040 --> 00:09:45,720
smart for their own good.

268
00:09:45,720 --> 00:09:48,280
All of these advancements come with exciting possibilities

269
00:09:48,280 --> 00:09:49,680
and complicated challenges.

270
00:09:49,680 --> 00:09:50,180
Yeah.

271
00:09:50,180 --> 00:09:51,560
And the biggest takeaway, don't be

272
00:09:51,560 --> 00:09:53,040
afraid to ask the tough questions.

273
00:09:53,040 --> 00:09:54,040
Challenge the assumptions.

274
00:09:54,040 --> 00:09:55,200
Demand transparency.

275
00:09:55,200 --> 00:09:56,480
And stay engaged.

276
00:09:56,480 --> 00:09:58,920
Because it's not enough to just adapt to the future

277
00:09:58,920 --> 00:10:00,240
that technology is creating.

278
00:10:00,240 --> 00:10:02,080
We need to actively shape it.

279
00:10:02,080 --> 00:10:03,240
Couldn't agree more.

280
00:10:03,240 --> 00:10:06,240
So next time you see a headline about some mind blowing new

281
00:10:06,240 --> 00:10:08,320
tech, don't just believe it.

282
00:10:08,320 --> 00:10:09,200
Dig a little deeper.

283
00:10:09,200 --> 00:10:11,400
Ask yourself how it might change your life.

284
00:10:11,400 --> 00:10:13,760
And don't be afraid to speak up.

285
00:10:13,760 --> 00:10:15,360
The future of tech isn't set in stone.

286
00:10:15,360 --> 00:10:16,800
It's being created right now.

287
00:10:16,800 --> 00:10:19,960
And we all have a part to play in how that story unfolds.

288
00:10:19,960 --> 00:10:21,640
Thanks for joining us for this deep dive

289
00:10:21,640 --> 00:10:22,600
into the world of tech.

290
00:10:22,600 --> 00:10:30,760
Until next time, stay curious.