1
00:00:38,600 --> 00:00:42,310
Hi and welcome back to
the Neil Ashton podcast.

2
00:00:42,750 --> 00:00:47,040
Today I wanted to talk about a topic
that is probably on many people's

3
00:00:47,040 --> 00:00:52,450
minds. This is the topic
of how will AI affect my

4
00:00:52,450 --> 00:00:57,210
career? How much does it affect
the choice that you may be making?

5
00:00:57,350 --> 00:01:03,150
If you're an undergraduate or
postgraduate wondering how is this gonna

6
00:01:03,150 --> 00:01:05,950
affect my job prospects,
what career should I go in?

7
00:01:06,040 --> 00:01:11,250
If you're in a sector that you believe
is potentially gonna change because of

8
00:01:11,250 --> 00:01:13,450
AI, should you retrain?

9
00:01:13,710 --> 00:01:19,180
I believe there is tremendous
positivity and promise with AI,

10
00:01:19,270 --> 00:01:23,670
but I am also sensitive to the
fact that it can cause anxiety and

11
00:01:23,670 --> 00:01:29,330
uncertainty, particularly for younger
people who are early on in their careers.

12
00:01:29,330 --> 00:01:34,280
And I wanted to at least give my
opinion and it's just an opinion

13
00:01:34,280 --> 00:01:39,580
and it's based on limited knowledge that
I have and could be completely wrong.

14
00:01:39,610 --> 00:01:44,990
But I wanted to go through and also
really focus probably a little bit

15
00:01:44,990 --> 00:01:49,940
more on people who are earlier on in
their careers and the things that

16
00:01:49,940 --> 00:01:52,780
I think are good to focus on.

17
00:01:54,190 --> 00:01:59,910
The idea is really: should you still be,
you know, specializing in engineering

18
00:01:59,910 --> 00:02:05,490
or sciences or maths in the era of AI.

19
00:02:05,490 --> 00:02:10,490
So, why ask this now? Well, you
know, it's clear to see that AI

20
00:02:10,490 --> 00:02:14,840
is getting better and better
and by AI in some sense,

21
00:02:14,840 --> 00:02:16,990
I mean, large language models.

22
00:02:17,000 --> 00:02:21,970
So their ability to write code and
analyze data is just tremendous.

23
00:02:22,590 --> 00:02:28,450
I struggle to find anybody
nowadays who isn't using some

24
00:02:28,450 --> 00:02:31,360
sort of coding agent to develop code.

25
00:02:31,670 --> 00:02:36,430
More and more, I speak to friends
of mine who are not at all

26
00:02:36,430 --> 00:02:39,140
in the technology industry.

27
00:02:39,140 --> 00:02:45,060
And they are, you know, really
on board and love the desktop

28
00:02:45,060 --> 00:02:50,600
and mobile applications, you know,
whether it's Gemini or ChatGPT or any

29
00:02:50,600 --> 00:02:55,870
others. And so it really is
now, I think today in August of

30
00:02:56,300 --> 00:03:02,080
2026, gone past any sense
of, this is just hype.

31
00:03:02,080 --> 00:03:08,190
You know, I think people are seeing
genuine usefulness out of AI.

32
00:03:08,370 --> 00:03:13,980
But if you look at those various
reports from like the World Economic

33
00:03:13,980 --> 00:03:19,930
Forum, I picked out some stat here,
around 39% of workers' skills will

34
00:03:19,930 --> 00:03:25,560
change by 2030. And so the
question is really how to

35
00:03:25,560 --> 00:03:28,010
prepare for that change.

36
00:03:28,310 --> 00:03:33,570
Now, when it comes to
engineering, and maths and

37
00:03:33,570 --> 00:03:38,030
sciences, I cannot see
a short or medium term,

38
00:03:38,030 --> 00:03:42,650
and longer term it's science
fiction to think about longer term.

39
00:03:42,650 --> 00:03:47,440
But in the short to medium
term, as in years and decades,

40
00:03:47,890 --> 00:03:52,880
there's only an increased
need for engineers.

41
00:03:53,230 --> 00:03:57,970
If we look across energy,
for nuclear, for wind,

42
00:03:58,060 --> 00:04:04,050
even still for fossil fuels, there
is a huge need to build that

43
00:04:04,110 --> 00:04:07,910
out. Of course, for AI, there's
a massive focus on energy.

44
00:04:07,920 --> 00:04:13,430
There's still a lot of amazing
research going on on nuclear

45
00:04:13,430 --> 00:04:17,280
fusion. And that sector needs people.

46
00:04:17,370 --> 00:04:21,600
Transport, buildings,
electronics, manufacturing.

47
00:04:21,610 --> 00:04:24,580
There's a buzz now in the
semiconductor industry.

48
00:04:24,730 --> 00:04:29,510
There's a huge focus in defence,
whether we like it or not.

49
00:04:29,510 --> 00:04:33,420
And there's a shortage of engineers.

50
00:04:33,630 --> 00:04:37,170
Anybody who speaks to a company will
know that it's difficult to hire,

51
00:04:37,170 --> 00:04:39,080
it's difficult to find good people.

52
00:04:39,210 --> 00:04:43,140
Now that may sound odd to graduates
who are struggling to find a job,

53
00:04:43,190 --> 00:04:47,890
but there is still a need for it.

54
00:04:47,890 --> 00:04:53,920
And it is hard to predict what tasks will

55
00:04:54,620 --> 00:05:00,230
change, but I would say
routine coding, you know,

56
00:05:00,230 --> 00:05:04,230
simulation setup, writing

57
00:05:04,230 --> 00:05:09,700
documentation, that sort of thing is

58
00:05:09,700 --> 00:05:15,480
going to become much more of an AI-driven

59
00:05:15,480 --> 00:05:20,590
thing. That doesn't mean that
we're hiring less developers,

60
00:05:20,650 --> 00:05:25,730
but it does mean that the skills
that you need are changing.

61
00:05:25,730 --> 00:05:28,960
And in some ways I see
this as a positive thing.

62
00:05:28,960 --> 00:05:33,730
And I think I've told this story
before, but Like many other people

63
00:05:33,730 --> 00:05:39,600
who did a more classical engineering

64
00:05:39,600 --> 00:05:42,540
degree, I was not a hardcore programmer
where when I did it, 20 years

65
00:05:42,540 --> 00:05:47,060
ago whenever it was, we didn't
do as much on programming,

66
00:05:47,070 --> 00:05:51,980
we didn't do as much on that,
it was more about engineering.

67
00:05:51,980 --> 00:05:56,300
And probably 10 years ago or five years ago,

68
00:05:56,370 --> 00:06:01,960
I did sometimes feel that that
held me back a little bit when I

69
00:06:01,960 --> 00:06:05,810
wanted to, you know, test
an idea out, write a code,

70
00:06:05,990 --> 00:06:09,860
do some more sophisticated analysis.

71
00:06:09,860 --> 00:06:12,110
I kind of realized I couldn't do that.

72
00:06:12,110 --> 00:06:14,680
And I actually felt that was a limitation.

73
00:06:14,680 --> 00:06:17,940
And I kept saying to myself,
I should really go and,

74
00:06:17,940 --> 00:06:21,060
you know, do some immersive
course in Python or,

75
00:06:21,060 --> 00:06:27,000
you know, do something to really
become amazing at programming as if

76
00:06:27,000 --> 00:06:32,320
that would help me. Or, for things
that I've mentioned before like,

77
00:06:32,320 --> 00:06:35,400
computer-aided design (CAD), I always said,

78
00:06:35,400 --> 00:06:38,040
if someone today said,
Neil, design a, you know,

79
00:06:38,040 --> 00:06:41,830
front wing of a Formula One
car, I would conceptually,

80
00:06:41,830 --> 00:06:46,050
well, more than conceptually, I
would know what I need to do,

81
00:06:46,050 --> 00:06:51,850
but I just don't have the
ability to do it in CATIA or NX

82
00:06:51,850 --> 00:06:54,590
or whatever CAD package.

83
00:06:54,590 --> 00:06:57,810
And so again, I would
probably therefore think,

84
00:06:57,810 --> 00:07:01,310
well, I can't do it because
I don't have those skills.

85
00:07:01,310 --> 00:07:07,260
So now with AI, I know
that I can code pretty

86
00:07:07,260 --> 00:07:12,140
much anything up. Now I'm not
talking about building an

87
00:07:12,140 --> 00:07:16,970
enterprise-quality thing and we
can get to that in a minute,

88
00:07:16,970 --> 00:07:22,550
which still requires people with the
skills of how to create a good program,

89
00:07:22,610 --> 00:07:27,570
how to organize, how to have a
team of people working on it.

90
00:07:27,570 --> 00:07:33,320
But just in terms of developing
prototypes or just automating

91
00:07:33,320 --> 00:07:37,010
things in a typical engineering company.

92
00:07:38,300 --> 00:07:42,810
When I was in Formula One, I was
doing a lot of the methodology.

93
00:07:43,630 --> 00:07:47,240
I think those teams exist
basically in every company,

94
00:07:47,460 --> 00:07:51,790
which is, you know, you have a bunch of
designers or aerodynamicists and this

95
00:07:51,790 --> 00:07:53,950
is across all of CAE now.

96
00:07:53,990 --> 00:07:58,230
But you had a core group of people who
would actually decide what methods

97
00:07:58,230 --> 00:07:59,970
would be used. How would you script it?

98
00:07:59,970 --> 00:08:03,610
How would you automate it? How would
you make it the most efficient?

99
00:08:03,610 --> 00:08:05,880
And that's what I was working on.

100
00:08:06,060 --> 00:08:11,390
And a lot of what we were doing was
scripting and automating things.

101
00:08:11,590 --> 00:08:15,420
Now, I would imagine if you
go into a Formula One team,

102
00:08:15,500 --> 00:08:21,460
it would be totally transformative
compared to when I was there

103
00:08:21,460 --> 00:08:27,010
in terms of there's no barrier to
the sophistication because you have

104
00:08:27,420 --> 00:08:32,490
probably the world's best coding
agent in some of these AI tools now.

105
00:08:32,670 --> 00:08:36,210
So why am I saying this?

106
00:08:36,710 --> 00:08:42,560
I'm saying that in some ways AI is
a great enabler and democratizes

107
00:08:42,560 --> 00:08:45,150
knowledge. It is a level setter.

108
00:08:45,220 --> 00:08:50,920
It actually in some ways makes
more people able to do the jobs

109
00:08:50,920 --> 00:08:53,970
that they couldn't do before.

110
00:08:54,270 --> 00:08:57,710
I strongly believe that.
I think AI can help you;

111
00:08:57,750 --> 00:09:02,310
you could apply for a job and
AI could help you to do that

112
00:09:02,310 --> 00:09:07,950
job. So in that sense, rather
than it being a negative thing,

113
00:09:08,020 --> 00:09:12,890
I would say I would now feel more
comfortable going into a job that

114
00:09:12,890 --> 00:09:17,650
I wouldn't probably have applied for
before because I know AI is gonna be

115
00:09:17,650 --> 00:09:22,550
there to help me. So that is
a positive thing in my mind.

116
00:09:22,740 --> 00:09:28,140
Now, if you think about it,
what really matters are

117
00:09:28,140 --> 00:09:30,580
still the fundamentals.

118
00:09:30,950 --> 00:09:36,700
So another example I give is
if someone said to me now,

119
00:09:36,790 --> 00:09:42,320
go and work in a bio company
looking at drug discovery.

120
00:09:42,500 --> 00:09:44,840
I haven't got a biology degree.

121
00:09:44,990 --> 00:09:48,020
I have no idea how drug discovery works.

122
00:09:48,190 --> 00:09:51,220
I have no chemistry background,
no biology background,

123
00:09:51,220 --> 00:09:54,790
no medical background. For
sure, I could, you know,

124
00:09:54,790 --> 00:09:56,620
get a really good AI model.

125
00:09:56,620 --> 00:10:01,630
And I'm sure it could do many of
the things that someone would ask

126
00:10:01,630 --> 00:10:05,260
me to do. But I would have
no idea if it was right or

127
00:10:05,260 --> 00:10:09,540
not. I could be smart and I could
maybe get a couple of different

128
00:10:09,540 --> 00:10:12,190
models and get the model
to check the other model.

129
00:10:12,310 --> 00:10:15,210
But I really would be just,
you know, a dumb user of it.

130
00:10:15,360 --> 00:10:17,840
I really wouldn't have any sense.

131
00:10:17,950 --> 00:10:20,910
I have no background in that.

132
00:10:21,230 --> 00:10:27,380
So in that way, until AI
models become so unbelievably

133
00:10:27,460 --> 00:10:33,480
good. And that, I think in a broad
sense, will still take a long

134
00:10:33,710 --> 00:10:39,040
time. There will always need
to be the people there to

135
00:10:39,040 --> 00:10:42,200
orchestrate and guide and check and sort of,

136
00:10:42,200 --> 00:10:45,000
yeah, orchestrate is the word I like to use.

137
00:10:45,430 --> 00:10:49,140
So in that sense, does the fact
that AI is great at drug discovery

138
00:10:49,140 --> 00:10:53,530
mean that you shouldn't study biology
and chemistry and go into that?

139
00:10:53,530 --> 00:10:56,240
No, of course you need to, you
need people with that background.

140
00:10:56,380 --> 00:11:01,400
If we go to engineering, the
reason that I find AI so useful to

141
00:11:01,400 --> 00:11:07,420
me is I have an engineering degree
and a good knowledge of the area.

142
00:11:07,670 --> 00:11:10,790
So if I ask the AI to help me to do a task,

143
00:11:10,820 --> 00:11:15,150
I'm able to steer it to point
it and tell it and validate it

144
00:11:15,150 --> 00:11:18,250
and check it. I'm on all those levels.

145
00:11:18,380 --> 00:11:22,660
That's where the fundamentals still matter.

146
00:11:22,910 --> 00:11:26,760
You still need to sort of
understand physical principles.

147
00:11:26,980 --> 00:11:30,100
I'm talking about engineering now.
You still need the domain knowledge.

148
00:11:30,170 --> 00:11:33,420
You still need to know what
codes should I be using.

149
00:11:33,510 --> 00:11:37,640
You still need to have a sense of economics,

150
00:11:38,120 --> 00:11:44,000
of team working. So in that sense,
if you want to go into industry,

151
00:11:44,120 --> 00:11:49,490
I still think an engineering, computer
science, maths or physics degree is the right

152
00:11:49,490 --> 00:11:55,420
choice, even though AI

153
00:11:55,420 --> 00:11:58,890
is great at programming
and statistics and data,

154
00:11:58,930 --> 00:12:00,760
you still need to know it.

155
00:12:01,030 --> 00:12:06,200
At this point, I would say
the biggest opportunities

156
00:12:06,200 --> 00:12:09,960
available now are for people who know AI.

157
00:12:10,230 --> 00:12:14,600
So if you look at the job market,
the best-paying jobs today,

158
00:12:14,720 --> 00:12:20,750
the greatest need is for people

159
00:12:20,750 --> 00:12:23,580
who can help the company
to use AI in the best way.

160
00:12:23,580 --> 00:12:26,530
Because AI is not cheap. And so
if you're gonna be spending,

161
00:12:26,530 --> 00:12:29,370
you know, huge amount of money
on tokens and subscriptions,

162
00:12:29,670 --> 00:12:31,640
you wanna extract the most out of it.

163
00:12:31,750 --> 00:12:36,510
So actually at the moment, you need
people who know how to use AI.

164
00:12:36,620 --> 00:12:38,530
That's probably the skill.

165
00:12:38,630 --> 00:12:43,730
And to be fair, this is
actually where younger

166
00:12:43,730 --> 00:12:46,940
graduates are in a better position.

167
00:12:47,140 --> 00:12:52,260
I've spoken to several people and they
have genuinely agreed that I think there

168
00:12:52,260 --> 00:12:57,360
was three scenarios. And this is probably
gonna be a bit negative for older

169
00:12:57,360 --> 00:12:59,710
people, but I'll say it anyway.

170
00:12:59,830 --> 00:13:03,780
So if you were gonna hire an engineer today,

171
00:13:03,830 --> 00:13:09,970
would you hire a 50-year-old
with 30 years of experience,

172
00:13:10,020 --> 00:13:15,770
but who has no AI experience and
actually has some AI negativity?

173
00:13:16,000 --> 00:13:20,250
And I do see that a lot, unfortunately.

174
00:13:20,350 --> 00:13:26,190
Would I hire an AI person that
has no background at all in

175
00:13:26,830 --> 00:13:31,800
engineering? Or would I find
somebody who's sort of early career,

176
00:13:31,860 --> 00:13:37,850
but extremely proficient with
AI and has the right mentality?

177
00:13:38,110 --> 00:13:41,150
It used to be the case that a
lot of companies would want the

178
00:13:41,150 --> 00:13:46,620
extreme expert. And it wouldn't matter if
they know AI because that's somebody what

179
00:13:46,620 --> 00:13:49,210
IT will do or something.

180
00:13:49,550 --> 00:13:53,250
When I spoke to most people, they would
actually prefer option three to one

181
00:13:53,250 --> 00:13:57,580
or two. It's becoming

182
00:13:58,020 --> 00:14:02,210
essential to have those AI skills.

183
00:14:02,550 --> 00:14:07,480
To some, they would rather sacrifice a
little bit of that domain knowledge because

184
00:14:07,710 --> 00:14:13,430
right now what matters more is someone
who can really take advantage of AI.

185
00:14:13,670 --> 00:14:15,980
And I don't mean just take it
and use it and blame it but

186
00:14:15,980 --> 00:14:17,940
be the person who wants to improve it.

187
00:14:18,000 --> 00:14:22,860
In a larger company that's maybe helping
to negotiate which AI tool they should

188
00:14:22,860 --> 00:14:26,210
use, whether they should be building
things internally, understanding all the

189
00:14:26,210 --> 00:14:28,900
different players where
it's good, where it's bad.

190
00:14:29,070 --> 00:14:33,610
So in that sense, if you're
not an expert in AI right now,

191
00:14:33,640 --> 00:14:38,960
at least in today's job market, I would
say you're at a massive disadvantage

192
00:14:38,960 --> 00:14:40,860
and you should learn very quickly.

193
00:14:41,070 --> 00:14:44,960
And by being an expert in AI, I
don't mean just using the chat

194
00:14:44,960 --> 00:14:48,100
interface of ChatGPT. I don't
mean just copy and pasting stuff.

195
00:14:48,310 --> 00:14:52,960
I mean actually getting in and
understanding how it works,

196
00:14:53,030 --> 00:14:58,490
understanding how you could build out some

197
00:14:58,490 --> 00:15:01,960
agents. You can still be using
some of these commercial tools.

198
00:15:02,200 --> 00:15:06,710
I don't mean write it all from
scratch, but somebody who knows how to

199
00:15:06,710 --> 00:15:12,450
do it, knows how MCP servers work, how
to connect different tools together,

200
00:15:12,710 --> 00:15:17,430
has an inquisitive mind, is
looking at where are things going.

201
00:15:17,650 --> 00:15:21,570
That is so valuable. If I
interview somebody now for a job,

202
00:15:21,590 --> 00:15:25,620
I want someone who is passionate about it.

203
00:15:25,910 --> 00:15:28,710
Because like it or not, that
is the technology of today.

204
00:15:28,970 --> 00:15:32,350
It's a bit like hiring
somebody who is, well,

205
00:15:32,430 --> 00:15:35,940
I don't really like to use
computers, 40 years ago.

206
00:15:36,110 --> 00:15:38,880
Yeah, I prefer to do everything by hand.

207
00:15:39,180 --> 00:15:42,170
At that time, maybe
computers were seen as like,

208
00:15:42,170 --> 00:15:45,370
we don't really need it. But for
sure, the company that did have it,

209
00:15:45,370 --> 00:15:48,760
or the same with the internet. If
you hired somebody who was a whiz

210
00:15:48,760 --> 00:15:52,980
with building websites during
the time of the dot com,

211
00:15:53,220 --> 00:15:56,080
that was really valuable for your company.

212
00:15:56,300 --> 00:16:00,950
Having that mindset might have helped you
shift to an online company instead of

213
00:16:00,950 --> 00:16:04,580
just physical stores,
which as many people know,

214
00:16:04,580 --> 00:16:09,320
there's many examples of companies that
failed because they didn't anticipate where

215
00:16:09,320 --> 00:16:14,820
things were going. So that's a
long-winded way of saying that

216
00:16:15,070 --> 00:16:19,930
today I would say it really
matters to know AI itself,

217
00:16:20,140 --> 00:16:23,910
not just background in
engineering or science,

218
00:16:23,950 --> 00:16:28,240
which is still super
important, but AI itself.

219
00:16:28,590 --> 00:16:31,170
And the great thing is now there's so
many YouTube courses, there's so much

220
00:16:31,170 --> 00:16:35,660
free content, many companies are desperate
to give you free training so that you

221
00:16:35,660 --> 00:16:38,050
use their products essentially.

222
00:16:38,150 --> 00:16:40,940
And this is true if you're
working in the software side,

223
00:16:41,390 --> 00:16:46,850
in the area I'm in, in more CAE
and EDA, absolutely massive

224
00:16:46,850 --> 00:16:52,830
opportunities there. So let's look
at what practical skills should

225
00:16:52,830 --> 00:16:56,660
you develop? And maybe we can
look across undergraduate,

226
00:16:56,830 --> 00:17:01,710
postgraduate, and then early
and late stage careers.

227
00:17:01,910 --> 00:17:06,890
If you're an undergraduate, I would
not be tempted to use AI to do

228
00:17:06,890 --> 00:17:08,710
all of the coursework for you.

229
00:17:08,710 --> 00:17:13,570
Even though it's very tempting to do it,
you're sort of ultimately gonna cheat

230
00:17:13,570 --> 00:17:17,650
yourself out of learning some of these
fundamentals that do matter later.

231
00:17:17,710 --> 00:17:21,840
And even I, there's certain things
now I kind of wish that I'd got

232
00:17:21,840 --> 00:17:25,790
even better grounding on
when I was at university.

233
00:17:26,030 --> 00:17:30,880
You know, some of these like fundamental
understanding of mathematical theory or

234
00:17:30,880 --> 00:17:36,900
engineering principles, they
really do give you that intuition

235
00:17:37,240 --> 00:17:41,820
and physical understanding
and logic that helps.

236
00:17:44,390 --> 00:17:48,720
But at the same time, I would
absolutely be using AI as a tutor.

237
00:17:48,810 --> 00:17:52,770
I think I've said this before, I think
AI's greatest role at the moment

238
00:17:52,770 --> 00:17:55,930
is to help you to study. So if
you get a piece of coursework,

239
00:17:55,960 --> 00:17:58,860
instead of sort of copying
and pasting it in and saying,

240
00:17:58,950 --> 00:18:03,330
you know, give me the answer,
essentially, I would do it to like create

241
00:18:03,330 --> 00:18:05,920
me a podcast that will teach me this area.

242
00:18:06,100 --> 00:18:11,130
Or I quite often have it where
I will ask it to create me

243
00:18:11,130 --> 00:18:14,990
a presentation that will teach me the area.

244
00:18:15,150 --> 00:18:19,490
And you can ask it to help
you to do things as long as

245
00:18:20,170 --> 00:18:23,400
it's legally allowed in
whatever study you're doing.

246
00:18:23,460 --> 00:18:26,250
But I would then be like, explain
to me in a high level how

247
00:18:26,250 --> 00:18:28,470
this works and then dive deeper.

248
00:18:28,470 --> 00:18:31,460
It's like having the ultimate tutor.

249
00:18:31,710 --> 00:18:34,980
So in that sense at the undergraduate
level I would use AI as a

250
00:18:34,980 --> 00:18:39,940
teacher not as something that does it
for you or you just you're ultimately

251
00:18:39,940 --> 00:18:43,790
just cheating yourself and you're
not going to learn the things.

252
00:18:43,790 --> 00:18:49,000
In the PhD, wow, if you do
do a PhD or research thing,

253
00:18:49,040 --> 00:18:52,700
if I had AI when I was doing my PhD, my God,

254
00:18:52,750 --> 00:18:57,640
that would have been so
useful, unbelievably useful.

255
00:18:57,830 --> 00:19:00,750
I mean, just incredible productivity.

256
00:19:00,990 --> 00:19:04,430
I just think how much I could have
done, how much more studies I

257
00:19:04,430 --> 00:19:09,750
could have investigated, and the way that
it would have helped me to understand

258
00:19:09,750 --> 00:19:12,840
new areas, just mind blowing.

259
00:19:13,110 --> 00:19:16,930
So I would fully embrace it
as the ultimate companion.

260
00:19:17,100 --> 00:19:21,510
At a sort of research level,
whether it's masters or PhD,

261
00:19:21,590 --> 00:19:26,970
I think the AI should be the bit
where you really hone those skills,

262
00:19:27,300 --> 00:19:31,150
help it to be your own
PhD student a little bit,

263
00:19:31,860 --> 00:19:35,710
to help you to do things. And that's
the mentality that I think most

264
00:19:35,710 --> 00:19:40,470
companies want. They want you
to use AI as your helper,

265
00:19:40,540 --> 00:19:44,710
not just a lazy way that you
just sort of do this for me,

266
00:19:45,100 --> 00:19:49,230
but rather something where you
almost have your own team.

267
00:19:49,430 --> 00:19:52,270
And that's why a lot of companies
are now talking about a sort of

268
00:19:52,270 --> 00:19:56,610
agentic engineer or an AI engineer, and
they don't really mean it to replace

269
00:19:56,610 --> 00:20:00,320
you as an engineer, but it's
almost to give you extra help.

270
00:20:00,520 --> 00:20:04,650
And so you're sat there farming
off things to go and do.

271
00:20:04,870 --> 00:20:07,180
So that's how I would do it.

272
00:20:07,390 --> 00:20:12,550
Now, of course, there's a
debate of what would you

273
00:20:12,550 --> 00:20:18,150
actually study. In that sense, it's very

274
00:20:18,950 --> 00:20:21,910
dependent on your area of study.

275
00:20:21,910 --> 00:20:25,590
But at least if you were doing
it now, then I would still focus

276
00:20:25,590 --> 00:20:31,590
very much on, let's say in engineering,
how can AI work for engineering in

277
00:20:31,590 --> 00:20:35,420
a more holistic way?
That's clearly one topic.

278
00:20:35,510 --> 00:20:38,560
But it doesn't have to be just on AI itself.

279
00:20:38,710 --> 00:20:43,280
If you're coming up with new theories,
or studying fluid mechanics,

280
00:20:43,280 --> 00:20:46,510
AI can just help you to do the
post-processing, it can help you to

281
00:20:46,510 --> 00:20:49,280
understand what's going on.

282
00:20:50,050 --> 00:20:53,030
You don't have to be
studying AI as the topic,

283
00:20:53,230 --> 00:20:57,830
it's just AI can help you essentially.

284
00:20:58,070 --> 00:21:01,020
And I know probably as I'm saying
this, because I've said the word AI

285
00:21:01,020 --> 00:21:04,360
so many times, some people
may be skeptical and think,

286
00:21:04,360 --> 00:21:10,300
but yeah, it would just be unwise
for you not to be looking into

287
00:21:10,300 --> 00:21:14,440
this. The good thing is if you
set that sort of strong basis,

288
00:21:14,480 --> 00:21:19,180
a good strong maths background,
engineering background,

289
00:21:19,190 --> 00:21:22,650
computational science, I would
understand how chips work,

290
00:21:22,680 --> 00:21:26,420
semiconductors work. That's a big area now.

291
00:21:26,620 --> 00:21:30,640
If you think about it, if AI does
become the revolutionary technology,

292
00:21:30,640 --> 00:21:34,440
you need to understand the
methods, how AI works,

293
00:21:34,460 --> 00:21:40,430
how it's trained, how inference works,
and all the stuff around fine-tuning and

294
00:21:40,430 --> 00:21:43,670
agents, etc.—harnesses
and scaffolding, blah,

295
00:21:43,710 --> 00:21:49,620
blah, blah. But arguably what
powers it and what powers

296
00:21:49,620 --> 00:21:54,540
it are chips. So I would argue
that actually learning more about

297
00:21:54,540 --> 00:21:58,380
semiconductors and EDA is
probably a very useful thing.

298
00:21:58,550 --> 00:22:01,320
Learning about, you know, the engineering,

299
00:22:01,340 --> 00:22:06,710
the power, the economics, that whole
stack will be very useful to understand.

300
00:22:07,140 --> 00:22:10,980
But I think you're in a great position
if you have that engineering background,

301
00:22:11,440 --> 00:22:15,710
science background, and you're very
proficient with AI, I think you could turn

302
00:22:18,090 --> 00:22:23,990
to many jobs. But I guess the elephant

303
00:22:23,990 --> 00:22:27,260
in the room is how do you stand out?

304
00:22:27,460 --> 00:22:32,970
If in some ways AI is the great
enabler, how do you make sure

305
00:22:32,970 --> 00:22:35,020
you get the job versus somebody else?

306
00:22:35,090 --> 00:22:38,400
It used to be that you would
just be the best specialist,

307
00:22:38,580 --> 00:22:41,140
the best university, the best degree.

308
00:22:41,500 --> 00:22:47,340
And that is to some extent
still true, but I would argue

309
00:22:47,340 --> 00:22:52,770
that if some of the technical ability is now

310
00:22:52,770 --> 00:22:58,310
taken by AI, then the
soft skills matter even

311
00:22:58,310 --> 00:23:01,460
more. Maybe I, you know,
lean too much into this,

312
00:23:01,460 --> 00:23:06,040
but I cannot tell you how
much I value the soft skills.

313
00:23:06,380 --> 00:23:11,030
I want to see someone who
can work well in a team,

314
00:23:11,030 --> 00:23:16,490
who has the right attitude, who has
the right organisational ability,

315
00:23:17,330 --> 00:23:22,240
who understands emotional
intelligence and sensitivity,

316
00:23:22,670 --> 00:23:28,600
who maybe has skills in languages and
understands different cultures—just

317
00:23:28,600 --> 00:23:30,250
someone who is, you know, good with people.

318
00:23:30,250 --> 00:23:36,190
More and more, anybody who goes
into leadership roles knows this:

319
00:23:36,190 --> 00:23:39,640
how well a company does is really
down to the people and the vibe

320
00:23:39,740 --> 00:23:43,090
and that is a lot about how you
recruit people and the people you

321
00:23:43,090 --> 00:23:46,090
have. The technical ability
matters a huge amount,

322
00:23:46,180 --> 00:23:50,940
but if we argue that AI is gonna
solve some of that technical ability,

323
00:23:52,990 --> 00:23:55,640
the dynamic of the team and the
people is gonna matter more.

324
00:23:55,640 --> 00:24:01,110
So I don't believe you can be
the sort of super brainy but a

325
00:24:01,110 --> 00:24:05,310
bit awkward person that
won't stand out as well.

326
00:24:05,500 --> 00:24:10,640
So I think the sort of soft skills
and I don't mean everyone needs

327
00:24:10,640 --> 00:24:14,470
to be an extrovert, but someone
who can do well in the team is

328
00:24:14,470 --> 00:24:18,180
probably gonna matter even more.

329
00:24:18,520 --> 00:24:23,080
Showing demonstrable proof is important.

330
00:24:23,390 --> 00:24:26,370
I always find that if
someone just sends a CV,

331
00:24:26,400 --> 00:24:32,450
if someone actually shows me a
presentation of something they've built,

332
00:24:32,450 --> 00:24:35,460
something they've done,
something—that's important.

333
00:24:35,460 --> 00:24:37,880
Internships are super important.

334
00:24:38,030 --> 00:24:41,040
That is ultimately the way that
a lot of companies hire now.

335
00:24:41,040 --> 00:24:45,990
They get somebody as an intern, it's
a bit of an extended job interview,

336
00:24:46,050 --> 00:24:48,710
you know, and then they hire.

337
00:24:48,710 --> 00:24:53,090
So actually, because there's so much
open-source stuff out there now,

338
00:24:53,090 --> 00:24:56,690
because AI is so great, you
should be able to build things.

339
00:24:56,690 --> 00:25:00,960
You should be able to demonstrate
your ability much easier than before.

340
00:25:01,150 --> 00:25:06,010
So if you wanna go into a job,
and it's about doing data centre

341
00:25:06,010 --> 00:25:10,520
design, you could probably pick
up some research yourself,

342
00:25:10,910 --> 00:25:14,050
you could build some simulations yourself,

343
00:25:14,220 --> 00:25:19,060
you could do so much just to
prove that you want and are able

344
00:25:19,110 --> 00:25:23,160
to do it, all with relatively
open-source and free tooling.

345
00:25:23,270 --> 00:25:26,790
Now of course the actual
token cost is another thing,

346
00:25:26,790 --> 00:25:29,890
and that's where there's a
lot of student programs,

347
00:25:29,910 --> 00:25:33,280
but frankly, it's probably worth it
for you to pay, and I think most

348
00:25:33,280 --> 00:25:37,360
people do, for a decent
subscription to an AI thing,

349
00:25:38,100 --> 00:25:40,880
because it's gonna help you out.

350
00:25:41,080 --> 00:25:46,050
So, it's hard for me to think,
but if I was starting again,

351
00:25:46,070 --> 00:25:49,920
I think engineering would
still be the right choice.

352
00:25:50,070 --> 00:25:54,640
I would still do it. And I think
it's even more relevant today as

353
00:25:54,640 --> 00:25:57,550
a good degree, regardless
of what you go into,

354
00:25:57,670 --> 00:26:00,690
just because every industry
has been transformed by AI.

355
00:26:00,750 --> 00:26:05,270
And I think engineering gives you
that grounding in those subjects.

356
00:26:05,420 --> 00:26:08,470
Maybe maths, you could
argue, it's very specific.

357
00:26:08,470 --> 00:26:14,050
An engineering degree, by
design, has elements of maths,

358
00:26:14,350 --> 00:26:16,650
elements of physics, elements
of computer science.

359
00:26:16,670 --> 00:26:19,670
I actually think it's a great
all-round degree if I'm being honest.

360
00:26:19,670 --> 00:26:24,640
I mean, I'm biased obviously, but I
think today it's even more relevant for

361
00:26:24,640 --> 00:26:28,980
you. So I've probably not

362
00:26:28,980 --> 00:26:34,550
addressed all of the points, but I

363
00:26:34,550 --> 00:26:39,970
really don't think that this
is gonna be a matter of a mass

364
00:26:39,970 --> 00:26:45,050
wipeout of jobs. Okay, long term if you

365
00:26:45,050 --> 00:26:50,520
extrapolate things and you say robots can
do everything and AI could do everything.

366
00:26:50,520 --> 00:26:53,880
Okay, maybe there's a very
uncertain long term future,

367
00:26:53,980 --> 00:26:57,510
but frankly, none of us know and
it is almost like a sci-fi thing

368
00:26:57,510 --> 00:27:00,430
to look at. I'm not even sure
it's worth looking at that.

369
00:27:00,460 --> 00:27:05,560
I can look at the short to medium
term, years or decades and

370
00:27:06,560 --> 00:27:10,810
I don't see any mass wipeout
of jobs in engineering.

371
00:27:10,870 --> 00:27:15,780
I see just it being a different skill
set, a different sort of productivity.

372
00:27:15,990 --> 00:27:21,770
And that is not something
where people are just gonna

373
00:27:21,770 --> 00:27:23,670
suddenly hire like 50% less.

374
00:27:25,510 --> 00:27:27,960
I just don't see that. I think
you just need to be flexible in

375
00:27:27,960 --> 00:27:31,360
what you will be doing for that company.

376
00:27:31,430 --> 00:27:37,210
At least for engineers, I can't
really comment about all the

377
00:27:37,210 --> 00:27:41,690
other jobs in the job
market, but as engineers,

378
00:27:41,720 --> 00:27:47,320
I can't see a point where we're
gonna suddenly need 50% less

379
00:27:47,480 --> 00:27:50,710
engineers. I think they're
in such short supply.

380
00:27:50,710 --> 00:27:53,180
So I think if you are listening to
this and you are doing engineering

381
00:27:53,180 --> 00:27:58,220
or want to do engineering, I think that's
a very sensible and future-proofed idea.

382
00:27:58,320 --> 00:28:03,110
And, yeah, I probably waffled on
a bit, but I just wanted to give

383
00:28:03,110 --> 00:28:09,050
you some of my perspectives
on this and if you have

384
00:28:09,050 --> 00:28:14,400
questions for me or comments
or thoughts that you have then

385
00:28:15,050 --> 00:28:17,350
please put them in the chat.

386
00:28:17,550 --> 00:28:20,640
I guess YouTube is probably
the easiest place to do it.

387
00:28:21,060 --> 00:28:23,220
All right, I hope you enjoyed that.

388
00:28:23,220 --> 00:28:28,190
The next few episodes will be
back to speaking with a guest,

389
00:28:28,190 --> 00:28:30,250
I think. So, I hope you enjoyed it.

390
00:28:30,510 --> 00:28:31,410
Thanks