WEBVTT

1
00:00:01.079 --> 00:00:04.640
Alright, welcome to the Neuvottelija channel.

2
00:00:04.640 --> 00:00:09.000
This is a bit of a glimpse into Neuvottelija Sisäpiiri.

3
00:00:09.000 --> 00:00:16.199
Neuvottelija Sisäpiiri is this kind of exclusive, or

4
00:00:16.199 --> 00:00:21.680
extra content, and it has several tiers. You get access to it when you go

5
00:00:21.680 --> 00:00:27.439
over to YouTube and subscribe to the Insider content.

6
00:00:27.439 --> 00:00:31.679
There's typically an extra discussion of about 10 minutes

7
00:00:31.679 --> 00:00:37.559
with the guests, which only you get to see. And on top of that there's

8
00:00:37.559 --> 00:00:42.399
a useful feature, that when you contact me after that, say

9
00:00:42.399 --> 00:00:48.520
on WhatsApp, we check on LinkedIn, say, that you're a real person, then

10
00:00:48.520 --> 00:00:52.719
then you can get into Neuvottelija Sisäpiiri, where there are now

11
00:00:52.719 --> 00:01:00.039
around 170 people and several of my guests. And there we then discuss

12
00:01:00.039 --> 00:01:05.119
these things properly, like adults. I'm one of the moderators there.

13
00:01:05.119 --> 00:01:11.000
There are only real people there. Messages disappear in seven days and no one

14
00:01:11.000 --> 00:01:16.439
represents their organization, unless they specifically want to.

15
00:01:16.439 --> 00:01:21.280
And then there's an extra feature that you can't really find elsewhere in Finland. So

16
00:01:21.280 --> 00:01:25.880
I'm actually a pretty good AI user. Throughout the year I've

17
00:01:25.880 --> 00:01:32.119
specifically been developing an AI called Samantha. It's OpenClaw-based, running on its own

18
00:01:32.119 --> 00:01:36.920
computer, an AI agent. It has its own kind of Hermes research

19
00:01:36.920 --> 00:01:42.640
agent, a Hermes agent system. It has the best language model available, and

20
00:01:42.640 --> 00:01:48.000
for example these helper models then for this Hermes agent

21
00:01:48.000 --> 00:01:53.159
through. It has a very well-developed memory, three-tiered

22
00:01:53.159 --> 00:01:59.119
and then you can ask it whatever you want. And in my opinion it answers very well

23
00:01:59.119 --> 00:02:05.200
for example about investing, politics, in general

24
00:02:05.200 --> 00:02:09.039
science, armchair economists,

25
00:02:09.039 --> 00:02:13.520
about basically anything, and people have fun there, and I'm trying to train it

26
00:02:13.520 --> 00:02:19.560
like an actual employee. Uh, I'm going to demo it now. I was

27
00:02:19.560 --> 00:02:24.599
in Tampere for the second time within a week. The first time, of course, I went to celebrate, or

28
00:02:24.599 --> 00:02:30.280
to watch Tappara win another championship. I bought this kind of

29
00:02:30.280 --> 00:02:37.680
nice gold pin here for my jacket lapel, but now I went again

30
00:02:37.680 --> 00:02:43.720
to the Tampere investment fair. Tampere-talo was completely full of people who

31
00:02:43.720 --> 00:02:49.000
wanted to discuss investing and look at these Finnish listed companies

32
00:02:49.000 --> 00:02:53.400
and I was there as an invited panelist for the investment club panel

33
00:02:53.400 --> 00:02:58.879
on using AI in investing. It was moderated by Aleksi Kopponen.

34
00:02:58.879 --> 00:03:04.879
Also there were Iida Hyynnilä and Timo Heikkilä and me. And now I'll show

35
00:03:04.879 --> 00:03:09.959
here the demo that I showed there too, in the hall. So here

36
00:03:09.959 --> 00:03:15.360
Samantha, my main agent, uh, gets a challenge, that first

37
00:03:15.360 --> 00:03:20.120
I talk with Ville Skogberg, the founder of Nebula, about whether we

38
00:03:20.120 --> 00:03:25.120
have both held Alphabet for years and whether it's been a good investment. So

39
00:03:25.120 --> 00:03:29.640
then first we discuss that, well, the results were really good, so is this

40
00:03:29.640 --> 00:03:34.239
still a good investment, and then we concluded that it's not worth selling at this point.

41
00:03:34.239 --> 00:03:40.120
especially since you can't even own Anthropic or OpenAI through the stock exchange. But

42
00:03:40.120 --> 00:03:43.000
then we start thinking about the fact that there's a bit of

43
00:03:43.000 --> 00:03:47.640
bubble-like characteristics to this AI boom. The Americans are putting in hundreds

44
00:03:47.640 --> 00:03:51.879
of billions of dollars into this AI race, so what, which are the

45
00:03:51.879 --> 00:03:57.560
stocks that could benefit from it. And Samantha then discusses with other

46
00:03:57.560 --> 00:04:03.040
AI working-group bots about possible sectors.

47
00:04:03.040 --> 00:04:07.120
The task is that this investment could be more like

48
00:04:07.120 --> 00:04:11.120
a value investment rather than a growth investment. That is, buying at a reasonable price

49
00:04:11.120 --> 00:04:14.439
probably somewhat weaker growth. We go through the use of AI

50
00:04:14.439 --> 00:04:19.600
in investing, and on stage come Sami Miettinen, Iida Hyynnilä and Timo

51
00:04:19.600 --> 00:04:23.960
Heikkilä. Welcome. Judging by the names, Aleksi Kopponen

52
00:04:23.960 --> 00:04:27.560
was the one who actually came on stage first, but without further ado

53
00:04:27.560 --> 00:04:32.680
let's bring Iida Hynnilä on stage first. Welcome.

54
00:04:33.320 --> 00:04:38.880
Then, uh, established Tappara man Sami Miettinen, and then investor

55
00:04:38.880 --> 00:04:43.120
Timo Heikkilä. Since we've already gone through Gemini here too, we can probably state

56
00:04:43.120 --> 00:04:46.759
roughly what they are. So they're very similar services

57
00:04:46.759 --> 00:04:51.960
to each other, like with ChatGPT, but is there really much

58
00:04:51.960 --> 00:04:55.960
difference between them from an ordinary user's point of view? Would Sami like to comment on

59
00:04:55.960 --> 00:04:58.639
this? Yeah, I have all of them in use, for the

60
00:04:58.639 --> 00:05:02.600
reason that, well, there's this kind of game going on, where Anthropic has been

61
00:05:02.600 --> 00:05:06.560
so popular and so good specifically in office use, because of its Cowork

62
00:05:06.560 --> 00:05:11.320
feature, that a huge number of users have come to it, and they've started

63
00:05:11.320 --> 00:05:16.039
solving this problem with capitalism. That is, they've started removing the queue

64
00:05:16.039 --> 00:05:20.840
model, where everyone got to use it almost unlimited for the monthly fee, and

65
00:05:20.840 --> 00:05:24.560
moved to that kind of API pricing. So, so you have to pay for the extra usage

66
00:05:24.560 --> 00:05:28.039
like real hard euros, which can be tens or hundreds on top of

67
00:05:28.039 --> 00:05:32.720
the monthly fee. This is a thing, then, OpenAI and Gemini are

68
00:05:32.720 --> 00:05:35.880
still a bit cheaper, which maybe shows then that they weren't as

69
00:05:35.880 --> 00:05:39.120
popular, but all of them are currently almost equally

70
00:05:39.120 --> 00:05:42.840
good in terms of capability. I actually use OpenAI's GPT-5 the most. It's

71
00:05:42.840 --> 00:05:46.560
the most cost-effective. So there's no big difference between them, but I would

72
00:05:46.560 --> 00:05:52.400
say that Anthropic is the one that gets you furthest in office work,

73
00:05:52.400 --> 00:05:55.919
because they've built a really good stack for how those Office

74
00:05:55.919 --> 00:06:00.000
Microsoft Office files, for example, get analyzed, or Excel files like that. That

75
00:06:00.000 --> 00:06:03.880
my multiples analysis was this kind of Claude Excel analysis, which by the way is

76
00:06:03.880 --> 00:06:07.639
in investing. I mean, since many of us are this kind of gut-feeling investor, where

77
00:06:07.639 --> 00:06:10.880
you just throw money at it because it sounded like a good company, because it had good business,

78
00:06:10.880 --> 00:06:15.280
so already this kind of finding relative valuation multiples kind of

79
00:06:15.280 --> 00:06:18.479
anchors the investment in reality, whether this really is

80
00:06:18.479 --> 00:06:22.280
nine times EBITDA for this value, so that's like kind of hard data

81
00:06:22.280 --> 00:06:26.639
and Claude is better at that than these others. Let's take a look at how the bots

82
00:06:26.639 --> 00:06:31.599
perform on this. So here they're kind of considering this initial thesis, whether

83
00:06:31.599 --> 00:06:36.720
Alphabet is still a good investment. There are grounds for this, because you find

84
00:06:36.720 --> 00:06:42.520
this kind of processor, TPUs, cloud, also this Google Search and other

85
00:06:42.520 --> 00:06:47.199
services. And then like a really solid infrastructure. Then from here

86
00:06:47.199 --> 00:06:50.520
you find these surrounding things, data center power and

87
00:06:50.520 --> 00:06:56.440
cooling equipment, power grid, data center operators.

88
00:06:56.440 --> 00:07:02.240
So these bots find these, then, network and data, traffic, infra,

89
00:07:02.240 --> 00:07:07.120
chips and memory layers. And then they start digging into, at the same time,

90
00:07:07.120 --> 00:07:10.800
thinking about what might be found here. And it throws out some, like, actual

91
00:07:10.800 --> 00:07:16.479
listed companies. Vertiv, Eaton, Schneider, GE Vernova, Siemens Energy,

92
00:07:16.479 --> 00:07:22.319
Quanta Services, Prysmian, Nexans, Arista, Broadcom, and now let's bring in a second AI.

93
00:07:22.319 --> 00:07:26.360
This is now a manual run from me, so I type into Claude's Excel these

94
00:07:26.360 --> 00:07:30.160
company names and then ask it to do this kind of investment banker

95
00:07:30.160 --> 00:07:34.000
relative multiples analysis of all these companies. Now it starts

96
00:07:34.000 --> 00:07:39.680
gathering this data via Claude, market cap, forecast,

97
00:07:39.680 --> 00:07:43.440
profitability, revenue, debt ratio

98
00:07:43.440 --> 00:07:48.319
and then we build this relative value matrix from all of these. So, that is

99
00:07:48.319 --> 00:07:52.479
like, say, this kind of enterprise value relative to

100
00:07:52.479 --> 00:07:57.280
EBITDA. And here now we find this Nexans

101
00:07:57.280 --> 00:08:01.080
seems to be this kind of value-play candidate, so

102
00:08:01.080 --> 00:08:05.560
smaller than the others, this kind of 8-billion valuation, and yet these multiples are

103
00:08:05.560 --> 00:08:11.039
more moderate, and of course the growth is a bit weaker too. Well then I throw

104
00:08:11.039 --> 00:08:16.560
this at the other bots now, three of them here. There are several of this Excel and

105
00:08:16.560 --> 00:08:21.159
then these bots read Excel fluently, so they analyze this, Samantha's

106
00:08:21.159 --> 00:08:25.000
work, or sorry, the one I made, this Excel

107
00:08:25.000 --> 00:08:29.879
in the Claude in Excel application, and then give recommendations based on

108
00:08:29.879 --> 00:08:35.000
that. And then here, this Ville's Donna bot

109
00:08:35.000 --> 00:08:38.279
also thinks that, as a value investor, this Nexans could be

110
00:08:38.279 --> 00:08:42.640
useful. Then I still give the final word on this to Samantha, that,

111
00:08:42.640 --> 00:08:46.920
now go through this Excel again and read the other bots' analyses and then give me

112
00:08:46.920 --> 00:08:52.120
a final investment proposal, which I could then present here

113
00:08:52.120 --> 00:08:58.279
at the Tampere investment fair, and Samantha then starts ticking away

114
00:08:58.279 --> 00:09:02.959
and also starts leaning toward this Nexans. It's a good infra thesis and

115
00:09:02.959 --> 00:09:09.880
a cheap price. This could be, well, a trap, meaning it could well be that [unclear] doesn't then see it

116
00:09:09.880 --> 00:09:14.320
in the orders, but there's also this kind of traditional cable business here. There's

117
00:09:14.320 --> 00:09:18.440
a bit of history to it. This is, well, the cable company left out of

118
00:09:18.440 --> 00:09:21.680
Nokia's Alcatel-Lucent acquisition. And this is not

119
00:09:21.680 --> 00:09:26.480
investment advice. I'm putting out to the market this

120
00:09:26.480 --> 00:09:30.399
buy order. I'm shading the price down a bit, in case it could be a bit

121
00:09:30.399 --> 00:09:34.320
cheaper, and let's see if it gets filled at that.

122
00:09:34.320 --> 00:09:39.640
If it goes through, then I'm actually putting in €4,900 into this investment, and that

123
00:09:39.640 --> 00:09:43.399
goes through, and this really isn't investment advice, but now there are 30

124
00:09:43.399 --> 00:09:48.480
shares of this stock, Nexans, a French company.

125
00:09:48.480 --> 00:09:54.000
Well, this is a demonstration of how I, and many of my friends there in Sisäpiiri, do

126
00:09:54.000 --> 00:09:58.399
analysis. Whether it's buying a stock, it's not enough that you just prompt it.

127
00:09:58.399 --> 00:10:03.959
It's not enough either that you just run this kind of Excel pass on it, there's this kind of

128
00:10:03.959 --> 00:10:07.600
second opinion, another opinion, or a third and a fourth, from other bots and

129
00:10:07.600 --> 00:10:12.600
their owners, so, so that then really tests the thesis. It doesn't, though,

130
00:10:12.600 --> 00:10:16.760
guarantee anything. This could even be a bad investment, and I did promise that I'll offer

131
00:10:16.760 --> 00:10:20.279
free tokens and beers if this works out, but otherwise it's entirely

132
00:10:20.279 --> 00:10:26.560
my own screw-up if this, if this doesn't fly. But you can do this over there in Neuvottelija's

133
00:10:26.560 --> 00:10:31.680
Sisäpiiri. I'd recommend it, try it, join there, get in touch with me

134
00:10:31.680 --> 00:10:36.120
and then apply for access to this Neuvottelija Sisäpiiri, and there then

135
00:10:36.120 --> 00:10:42.320
you can meet this Samantha, that is, my AI worker, and then use

136
00:10:42.320 --> 00:10:49.519
her know-how for your questions. Thanks, and no pressure. If it's not your

137
00:10:49.519 --> 00:10:53.440
thing, then I'd appreciate it if you liked the channel and hit

138
00:10:53.440 --> 00:10:59.600
subscribe. And, uh, I really do enjoy going out there to talk about AI and

139
00:10:59.600 --> 00:11:03.519
show it, and go to things like that. It's always nice to visit Tampere, not

140
00:11:03.519 --> 00:11:08.959
just for watching Tappara, but that kind of investment fair over there

141
00:11:08.959 --> 00:11:12.680
at Tampere-talo is just a perfect example of an event where I'm

142
00:11:12.680 --> 00:11:18.160
very happy to be a speaker. Thanks.