Latest / In the Loop AI / Why Anthropic Is Giving Away 20% of Its Equity (And Why Everyone's Mad)

Transcript

Why Anthropic Is Giving Away 20% of Its Equity (And Why Everyone's Mad) · In the Loop AI

0:00
Anthropic just gave away 5% of their company to you, and everyone seems to be mad about it.
0:06
Now, at their rumored $2.2 trillion IPO valuation, that's around $100 billion,
0:12
which is the largest philanthropic raise of capital ever.
0:15
That's more than the Gates Foundation ever raised. That is more than the entire
0:19
GDP of Costa Rica, and is enough money to run NASA for four years straight.
0:24
So you might be wondering, why on earth is everyone mad about this?
0:28
Well, there's a few reasons. Number one, the shares don't actually go to you or the public directly.
0:33
It goes to a mixture of NGOs, charities, and in some cases, and I quote,
0:38
specific populations of people.
0:40
Who they are is up to anthropics. So that's a little bit murky.
0:43
But there's also a more unsettling reason, which is something Dario Modi has
0:47
spoken about at length over the last couple of years, which is AI might wipe
0:52
out a huge number of jobs.
0:54
His quote was 50% of white-collar jobs over the next one to five years.
0:59
In the event of that happening, it is uncertain what the purpose of humanity
1:04
or individual humans will be, or more importantly, how they'll make a living to sustain themselves.
1:11
In the event of that, we would need to introduce something along the forms of
1:14
universal basic income, which would be
1:17
a dividend or a payout that comes in the form of a monthly salary or stipend
1:22
to individuals so that they'll be able to pay for things like groceries and keep themselves going.
1:27
Now, this concept might seem dystopian to some, it certainly feels dystopian to me.
1:31
We're going to unpack all of this, what exactly Anthropic is doing,
1:35
and why this is actually part of a broader trend that we're seeing amongst all
1:39
the leading AI labs, which may signal that this isn't something that we can necessarily stop.
1:43
So let's get into it on today's episode brought to you by Qualcomm.
1:51
Firstly, I think it's important to define what universal basic income even is.
1:55
Think of it as like a regular cash payment to every individual on earth.
1:59
So the universal part of universal basic income literally means to anyone and everyone.
2:04
Now, the basic refers to just enough money for people or that individual to
2:09
maintain their basic level of living.
2:11
So paying for things like groceries, rent, stuff like that. And then income
2:15
indicates that it's going to be a regular form of payments. So something that
2:19
you might receive monthly, like a salary or biweekly, like a salary.
2:23
And this isn't a novel concept. This has been spoken about for decades and decades.
2:27
But Sam Altman back in 2021, actually raised this for the first time when he
2:32
introduced something known as the American Equity Fund.
2:35
And the concept behind this was, it would be a tax, a 2.5% tax,
2:40
I believe, on the biggest companies that benefit from AI in the future.
2:45
And the idea here is, this tax or this fee would be paid directly to individuals
2:51
within that economy, within that nation, or all over the world.
2:55
And at the time, that meant that $13,500 would go into the pockets of every
3:00
single American every single year.
3:02
But he also tested this concept over the following three years up until 2024,
3:07
where he gave 1,000 low-income Americans...
3:10
$1,000 each every single month. And the results of that show things that you
3:15
may not have expected, where people just ended up becoming a little more lazier,
3:19
or they didn't really strive to create anything net new.
3:23
Now, fast forward to 2025 of last year, and Dario Modi introduces his infamous
3:29
piece on AI basically wiping out 50% of white-collar jobs over the next one to five years.
3:36
And his reasoning then was to introduce something known as a 3% token tax,
3:41
which would function similarly to what Sam Altman proposed four years earlier.
3:45
And then we had earlier this year, OpenAI converting their structure into some
3:49
form of a non-profit, which roughly holds enough equity that I think is attributable to $135 billion.
3:57
And the idea of this is sort of similar to what Anthropic has just proposed,
4:01
a public benefit corporation that will feed shares or the value that this company
4:06
eventually ends up generating to the public good.
4:10
But then the question becomes, what's different about Anthropic's proposal?
4:13
So it starts off with saying that Anthropic has created a wholly separate entity
4:17
for the public benefit, and that they are committing, subject to board approval, 5%
4:23
of their entire company, so in equity, to this company.
4:27
Now, this company is structured as an LLC, not a public benefit corporation.
4:32
And the reason cited for this is flexibility.
4:35
It would give Anthropic the ability to give in many different ways.
4:40
And they also stated that there's no political agenda behind this.
4:44
It is purely meant to benefit the public. But how it benefits the public is,
4:48
I guess, where the secret source of this entire announcement is.
4:52
In their words, they say they might explore large donations to traditional non-profits
4:57
or direct distributions to specific populations. Now, you could argue that.
5:03
Arguably one of the best insights into this because they're actually building the models.
5:07
They have done multiple research profiles actually speaking to the users of
5:11
Claude and people who don't use Claude yet trying to understand what the impact is of AI.
5:16
So in Anthropics defense, they've actually been incredibly proactive and public
5:20
in sharing the results from this type of research.
5:23
So you could actually argue that they do have really good insight as to how
5:27
these AI models are actually going to impact the world and global population.
5:30
But also, they wouldn't have the entire insight into this. The only way we're
5:34
actually going to understand is just releasing these models to humanity and
5:38
observing what actually happens over the next couple of years.
5:41
And actually, Jack Clark, who is a co-founder of Anthropic,
5:45
went on record in saying that this is meant to be an intentional act of benefiting
5:51
humanity from the spoils or the economic wealth that is going to be generated
5:56
from Anthropic's models.
5:58
But also it's meant to act as a hedge or an insurance policy for whatever problem
6:04
or disruption plays out in a world where AI models replace a large amount of
6:09
jobs or people's purpose or their ability to make a living.
6:13
Now, it's important to understand that this isn't actually universal basic income.
6:18
Income is something that happens more regularly. This is more of a 5% upfront
6:23
here and now with the promise that if Anthropik's valuation increases 50%.
6:28
Every 50%, they will add another 2% of the company's equity.
6:33
Now, what I found interesting is how they're funding this equity.
6:37
The 5% is just being given de novo, and so will the additional 2% for every 50% increase.
6:42
But what I didn't realize is Anthropik has collectively committed up to 20%.
6:48
Of the company's equity to the public good.
6:51
Which is absolutely insane, especially for a company that is the fastest growing private company ever.
6:58
Now, where this equity comes from is even more interesting. When the founders
7:01
created Anthropic, they foresaw
7:05
that this would be an issue or a problem or a challenge that they would need
7:09
to prepare for in the event that AI models became super powerful or super intelligent.
7:14
Now, it's important to point out this was before Claude even existed.
7:18
Anthropic founders committed 80% of their equity to this cause,
7:24
to the wider public benefit.
7:26
And they also created a matching program, which 50% of Anthropic employees actually
7:31
participate in, where they just willingly give their equity,
7:34
which will be used for the public good.
7:36
Over the years that Anthropic has existed, which is not that long,
7:40
they have committed 20% of the company's equity that will now be handed towards
7:44
the public with the promise of maybe even 2% with every 50% increase of the company's valuation.
7:50
So this is some form of altruism that we have not seen in any other company before.
7:55
And it's important to understand how the numbers play out. So I did a bit of a calculation here.
8:00
And at the $2 trillion valuation, which is what their rumored IPO is going to
8:04
be at, that would be $100 billion.
8:06
But then with every 2% increase at, let's say $3 trillion.
8:10
$4.5 trillion, and so forth, that would be an extra $60 billion,
8:14
$90 billion, $135 billion, and eventually
8:17
$200 billion at a $10 trillion valuation, which if you assume that.
8:23
These AI companies are going to consume a large portion of the economy,
8:28
where AI is going to be able to.
8:32
Considered very valuable work today, that doesn't seem out of reach.
8:37
So these are the facts of what Anthropic announced over the weekend on Saturday.
8:41
But I think the public reaction really shows you where people's heads are at
8:46
with this, including myself, which I want to share honestly.
8:49
But firstly, some of the most common public reactions were largely negative.
8:53
Some people saw this as a PR stunt pre-IPO just to make Anthropic seem like
8:59
a very relatable and caring company so that people would buy into the equity.
9:04
There's also a more economically strategic critique that we saw where if Anthropic
9:10
portrays themselves as this public benefit corporation where individuals would
9:14
potentially own shares, they would have skin in the game.
9:17
So they would be more economically incentivized to buy into the IPO.
9:22
And someone pointed out that this would be a very handy or convenient backstop
9:26
for Anthropic in the event that they need some form of a bailout.
9:29
If the majority of the population of the world or even America had an economic
9:33
stake in Anthropic, they would be more willing to vote politically to be able
9:37
to bail them out in the future in the event that anything catastrophic ends up happening.
9:41
David Sachs, who is the former AI czar of the US, pointed out that he's...
9:48
Going to wait back and see what Anthropic actually does. Committing 20% of your
9:52
equity is one thing, but actually using those funds for good is a complete other
9:57
beast that hasn't really been solved by philanthropy at scale just yet.
10:02
So he'll wait back and see what happens. Palmer Lucky made a really good point
10:06
where he said the definition of public good is going to be something that Anthropic
10:11
and a few individuals within Anthropic define
10:13
and is likely not going to be something that the wider public can help influence
10:19
because they are a private company.
10:21
Or even if they go public, they're still technically at the whim of their shareholders
10:25
or whoever owns the majority of the equity, which right now are the founders
10:28
collectively and investors. So it's still not going to be a genuine representation
10:33
of what humanity believes that's going to be.
10:35
And I guess Anthropik's response to this will be, well, that's why we're distributing
10:39
it to multiple different charities so that it's not part of what our jurisdictional
10:44
definition is going to be.
10:45
So i could see both sides of that and then you have travis kalanick the former
10:48
ceo of uber who basically said that this is an extension of the effective altruist movement
10:54
which is a political critique that anthropic has faced quite a lot they have
10:58
shown to have ties to the ea movement which,
11:03
focuses on i guess a more duma outtake which is very reflective of what dario
11:09
believes or thinks will happen.
11:11
But kind of putting the political agenda aside, the Doomer aspect of this isn't something that we could.
11:18
Necessarily ignore, which would be my opinion, right?
11:21
Like, you could see the Duma case playing out pretty realistically,
11:25
if you allow these AI models to get out of hand, or if we just maybe naively
11:29
assume that AI isn't going to replace a bunch of different things.
11:33
I truly don't believe that. But I do believe we need to take a more optimistic
11:36
approach to preparing humans well for this.
11:40
Now, whether universal basic income or some form of economic stipend,
11:43
like a 5% handover of shares to the public wide world is a good thing or not?
11:48
I don't know. That's something that we have to try and figure out.
11:50
I think one of the more practical takes I saw was basically asking Anthropic
11:55
to keep their stake, or rather use the stake themselves and determine how they
12:00
can more economically benefit the wider world.
12:04
Their point being, if Anthropic were to give this stake over to charities,
12:09
charities don't have the best track record of being able to fulfill the public good.
12:14
A lot of the money goes to execs or people that kind of fund the board or sit on the board.
12:19
It's much better off being in the hands of the company that actually generated
12:23
the economic value in the first place.
12:25
And this was actually the point that Elon Musk made in response.
12:28
He said, listen, either you give these shares directly to the public or figure
12:34
out a way to benefit humans at scale. And I'm going to say,
12:38
Elon actually went on to say that he plans to give every single kid in America
12:44
one share of SpaceX next year in 2027.
12:49
Now, if you run the math on that, I think there's around 73 million kids in
12:54
America today, well, projected to be next year.
12:57
And so to give them each a share, let's take the IPO valuation of $135.
13:03
That's around $10 billion.
13:05
These are huge numbers to play out. and I see his reasoning.
13:08
His reasoning is basically the best way to figure out how this universal basic
13:13
income will work for individuals is to give it directly to them and not add
13:18
intermediaries in the form of charities.
13:20
Figure it out yourself is basically what he's saying and what he plans to do with SpaceX.
13:24
Which brings me on to a wider point around this not being a novel experience
13:30
or initiative for Anthropic. They for sure are the first ones to take steps in the direction of.
13:37
Creating some form of a scheme or agenda. This is the first step that we're seeing from Anthropic.
13:42
Maybe it evolves over time, but this is something that has been spoken about
13:46
by all leading labs, OpenAI, Google, Elon Musk, and now through extension of Elon Musk, SpaceX.
13:52
So the wider point then becomes, why is this necessary?
13:56
And I've been trying to process my feelings and thoughts about this.
13:59
And I came to the conclusion that since the dawn of humanity,
14:04
humans have had a purpose.
14:08
And that purpose has been to invent, to create, to innovate,
14:13
to push the boundaries of humanity
14:15
and what we could create beyond what we understood at that point in time.
14:20
Without that drive, we would not have been able to discover all the wonderful
14:24
things that we have in science, medicine, physics, chemistry.
14:28
In art, in poetry, and all the types of literature that we've seen over the decades and millennia.
14:35
The point is, human innovation has driven so many different things.
14:39
Now, when computers came, when any other technological wave has come in,
14:44
it has always aided human beings in doing this, in innovating.
14:49
But AI is very different in the sense that they're sort of replacing the role
14:54
of humans, or it could potentially replace the role of humans.
14:57
And in the event that it is able to do this, in the event that you can stick
15:01
an AI model in a robot and it could automate both the mental human software
15:06
cognition of a human as well as the physical cognition of a human.
15:11
Then what are humans meant to be used for? Do we just sit around all day?
15:15
Do we need to be babied by this super intelligence that knows a lot better than
15:19
us, that knows how to raise us, how to monitor us, how to herd us?
15:23
Do we just end up becoming some form of organic waste? Now, that is a bearish
15:28
take, in my opinion, but that was what was going through my head and what I
15:31
think kind of like fed in through my unsettling behavior.
15:35
But when I applied my kind of like logic to this, I realized that there's probably
15:39
no way of avoiding the scenario that humans will end up having a lot more free time on their hands.
15:46
But I don't necessarily think that that's a bad thing. I think that that could
15:49
result in a much more optimistic outcome, which is we can operate on a much
15:53
higher order of thinking and creativity. Think about it.
15:57
Right now, we spend a lot of our time crunching Excel sheets,
16:01
going through emails, texting people, scrolling, trying to ingest information,
16:07
compute that information, all to achieve some kind of specific outcome.
16:12
But what if you get an AI to automate a bunch of that for you whilst bringing
16:17
you up to speed so that you can spend your time doing way more productive things?
16:21
Now, maybe at a high level that looks like managing a fleet of different AI
16:26
agents or AI models, or managing a bunch of artificial intelligences to help
16:31
you advance in your personal and professional work life.
16:35
Maybe that's a future that would exist today, right?
16:38
In this optimistic sense, people can end up probably innovating at a faster
16:43
and much larger extent than we ever have before.
16:46
But as Anthropik's announcement said in their post, they don't really know what
16:51
might happen or what this disruption may end up being the case.
16:55
And I guess that's where the unsettling feeling comes from. It's not.
17:01
Disagreeing with Anthropic's take. It's not being a complete doomer about this entire thing.
17:06
It's about not knowing where this all ends up and only time will tell.
17:12
We don't know the crazy disruption that Anthropic or any AI lab will have when
17:18
they release their models that are crossing the intelligence level of an average human being.
17:24
We haven't reached that bit just yet. we're starting to see it broach in certain
17:28
fields with mathematics.
17:30
Over the last couple of weeks, we had internal models from OpenAI solve 20%
17:34
of mathematics toughest problems. And you had,
17:40
what do we do about this? How do we fulfill our roles as mathematicians if we
17:46
can just get an AI doing the same thing?
17:48
We're starting to see the same thing in the field of physics,
17:50
and eventually this will spill
17:51
out into any and every human labor or professional job that we see today.
17:56
So I understand that point of view, but there's no telling how severe this is
18:01
going to be or whether humans actually end up better than they expected.
18:07
David Sachs actually critiqued Dario Mode's prediction that 50% of white-collar
18:12
jobs will be wiped out over the next one to five years by showing data that
18:15
white-collar jobs actually hadn't been replaced, but had actually increased.
18:19
Coding was the best example for this. When Call Code launched at the end of
18:24
last year, fast forward to today, these coding models became really good.
18:28
And they technically right now can replace a lot of L5 and L6 staff engineers' jobs.
18:35
But we haven't seen that. We've seen instead these companies hire more engineers
18:40
to instead manage these different AI agent fleets.
18:44
So the idea here is what we presume to be a Duma take actually ended up being
18:48
way more positive for humans.
18:49
So this could end up happening, but there's no certainty in determining how
18:53
it's going to happen or how it's going to play out.
18:56
And so I guess going back to Jack Clark, the Anthropic co-founder's take on
19:00
all of this is we need to hedge against the risk of people losing their jobs
19:05
or people having to evolve their jobs we don't quite know how that's going to look like but
19:09
this is our best attempt at trying to address what that is so let me leave you
19:14
with my own personal thoughts as to
19:16
how this plays out and how this is going to impact you and i over the next couple
19:21
of years because i think there's a lot of ways that this can go so let's get
19:25
into it on today's episode brought to you by qualcomm.
19:30
So let's start off with the optimistic take, which is universal high income,
19:35
which is beyond basic income.
19:38
It means that everyone gets a lot of money and we benefit from the vast amount
19:43
of wealth that AI and superintelligence ends up generating in this world. It's a panacea.
19:49
Everyone gets enough money, the wealth gap closes, and you're able to afford
19:54
anything and everything that you had hoped to afford in the first place there's
19:57
no shortage of supply or goods and services
20:01
and everyone gets to live a much better life order of magnitudes better than what they
20:06
currently exist today so it's not some form of communism it's not some form
20:11
of extreme capitalism it's the benefits or measures of both of those things
20:15
together in this world the machines would basically do.
20:19
Everything for you and it leaves you up with more time to do whatever you want
20:23
whether that's spending time with family, creating more memories,
20:26
or just creating or inventing other stuff on your own that maybe feeds into
20:30
the economy that the machines and AI actually operate.
20:33
The question I have is, what's the evidence of that actually happening?
20:37
And I think a good case study was the one that OpenAI ran that I mentioned earlier
20:41
on, which basically gave 1,000 people with household income around $30,000 a
20:46
year, so not very much the lower income bracket, $1,000 a month for three years.
20:52
The results of the data basically showed that people didn't end up inventing or creating more.
20:57
They just worked a little less, roughly an hour and a half a week,
21:01
reported at least, and no measurable improvement in job quality.
21:06
So in that event, people just decided to take the stipend and get increasingly
21:11
lazier. And then putting aside both of these scenarios, the optimistic take
21:15
and the doomer take, what's the right structure to actually build this entire
21:19
thing? Because in my opinion.
21:21
There needs to be some form of universal basic income or universal high income,
21:25
whatever you want to call it.
21:27
But the hardest challenge is going to be delivering that.
21:30
Should it be what Anthropik's done, which is donating the shares to a separate
21:34
entity and then giving those away to charities and then specific populations?
21:38
Is it a private company defining what public good is going to be,
21:43
defining who gets the money and who doesn't?
21:46
Or is it some form of committee that represents every single class of individual
21:52
person, human being, profession, sector, industry, and them collectively coming to an agreement?
21:58
I don't know the answer, but it seems very complex across the board and something
22:02
that needs to be probably discussed and worked on and experimented with over
22:06
time. But the point is, we're here.
22:08
It's an unprecedented announcement. This goes beyond just the model updates.
22:13
This goes beyond what you're using AI agents for in your day-to-day business.
22:18
This is the wider impact on humanity. This is the wider course or inevitable
22:22
conclusion of where AI eventually leads us.
22:26
Think about it. Just play it out for a second. AI models are smart right now,
22:30
but what happens when they become smarter than humans? What happens when they
22:32
have robotic limbs and arms that can do all the physical labor and work that we do currently today?
22:38
What happens when AI can replace anything and everything that gives humans purpose and value today.
22:44
I don't mean to try and sound philosophical.
22:46
I mean to sound how I think things might eventually end up playing out.
22:50
And again, as I mentioned earlier, this is not to be a doomer about it all.
22:54
I guess, to try and figure out or start a conversation between you and I about
22:58
what this eventually ends up looking like. It's a weird world.
23:02
I don't quite know how it's going to end up, but I remain optimistic that having
23:07
these conversations, starting these initiatives, and I'm glad to see Anthropic
23:10
taking the first step, is a good way to figure all of this out,
23:14
because it's going to be a very weird world.
23:16
Think about it. If people just got a regular stipend, it would be very predictable
23:20
how the economy or spending habits might actually end up progressing.
23:24
It'll affect the investment market. It will affect spending on consumer goods
23:29
and services, as well as just habits and general drive of the population and people.
23:34
Right now, you have a very vast
23:36
and eclectic mix of humans that are just trying to make it in the world.
23:41
They express that through the jobs and the degrees and the universities that
23:46
they attend and strive to try and achieve, the certifications,
23:49
the prowess, the status.
23:51
But what does that look like in a world where everyone is sort of fed the same,
23:56
stipend or money and what happens in a world where that is a necessity that
24:00
actually has to be the case it's just weird and it is a step away from capitalism
24:07
or maybe an evolution of capitalism.
24:09
But let me know what your thoughts are on this if you have an opinion or if
24:12
you have a take or if you just agree in some things please let me know and also
24:16
if you're listening to this on youtube spotify apple music wherever you're listening
24:19
to this or watching this,
24:20
please give us a thumbs up, subscribe, turn on notifications,
24:23
give us a rating. It helps us out massively.
24:26
And if you have a friend that might want to listen to this episode or listen
24:29
to any of the episodes that we have previously, please share it with them.
24:32
It helps us out massively.
24:34
We are going to be here five days a week, you know? So anything and everything
24:38
that happens in AI or frontier tech, I'm going to discuss and give you my thoughts and opinions on.
24:43
And if you find it interesting, I would love to hear from you.
24:45
Thank you so much for listening, and I'll see you on the next episode of In
24:48
The Loop brought to you by Qualcomm.