Latest / The Overlap / Silicon and Sparks: How AI and Crypto Are Remaking the Grid (Part 2 of 2)

Transcript

Silicon and Sparks: How AI and Crypto Are Remaking the Grid (Part 2 of 2) · The Overlap

0:00
If this is where you picked up the podcast today, you are in the wrong place.
0:07
This is a part two of a two-part episode.
0:11
Go back to last week's episode and check that one out and give it a start and come back
0:15
right here and join us here at The Overlap as we talk about AI, cryptocurrency, electricity,
0:22
and the future of innovation.
0:41
All right.
0:42
Now welcome back to part two of The Overlap podcast about AI, cryptocurrency, and electricity.
0:47
Will, why don't you kick it off where we left off from?
0:51
Right.
0:52
So, why the hope that these two technologies, although they're very energy hungry, we understand
1:00
now, why they're so energy hungry?
1:03
How do we come up with this idea that perhaps they could rescue us from the energy scarce
1:06
or energy consuming future that we fear and actually give us a positive way out?
1:12
I asked that question, I'm actually going to ask you that question.
1:14
Are you asking me?
1:15
I mean, I have my answer, but.
1:18
Well, what we have is we have examples from the past, right?
1:21
Yeah.
1:22
And the fact that new technologies, although they appear to at first be energy consuming,
1:26
actually enable us to create things that are energy producing, right?
1:29
The very steam turbines we were just talking about, if we could bury these AI and use them
1:34
to heat water and all that, those turbines exist because of the industrial revolution
1:38
and the things that that made available.
1:40
Our variability to harness steam in engines was what gave us the capacity to generate much
1:47
more energy than we actually used ever prior to the industrial revolution.
1:51
And while there's.
1:52
Yeah, I think.
1:53
Go ahead.
1:54
Go ahead.
1:55
We both do that to each other all the time.
1:57
I think that it's important to evaluate this, right?
2:01
I think we're at the precipice of a junction in which we're looking at a major categorical
2:08
shift in how we do things, right?
2:10
For millennia, the human progress was limited by the power that could be generated by human
2:17
and animal muscle.
2:19
Right.
2:20
And that was that was the limit, whether that was swinging of an axe or the walking of an
2:25
ox.
2:27
So the development of the water wheel, for example, produced an increase.
2:33
We had it.
2:34
We had it there that we were now able to extend ourselves past our own physical limitations
2:40
and the limitations of work animals.
2:43
So we have the water wheel, right.
2:45
But they were then geographically tethered.
2:48
They were tethered to rivers.
2:49
If there was no river in the area, there could be no water wheel.
2:52
Right.
2:53
And then you have the steam engine like you were talking about.
2:57
That that thing by itself transformed and created the power available to a society by
3:03
about 600 times.
3:06
So a steam engine had about 600 times the maximum amount of power that a human being
3:11
or an animal could produce.
3:13
Exactly.
3:14
And that's good.
3:16
But because because the coal used to create the steam engines could be transported almost
3:22
anywhere, right.
3:23
Now you can have a steam engine completely decoupled from energy production and completely
3:28
from geography for the first time in the in the history of the modern world.
3:33
Which is the industrial revolution, which gave way to factories and railroads and mass
3:38
production.
3:40
Exactly.
3:41
And now we have access to a technology that can simulate potentially one day simulate
3:46
new types of engines and new types of energy retrieving and generating devices that can
3:52
potentially unlock more power than we can even conceive of now to not only feed the
3:57
future AI and future cryptocurrency devices, but also many other things that we may not
4:03
have even dreamt of yet.
4:05
Perhaps those starships that you were talking about before.
4:08
Yeah, I think I think one way that crypto is similar in this scenario is that it decouples
4:14
our system of trading from any one country.
4:18
Right.
4:19
And allows free money in the same way that cross borders without needing middle middlemen.
4:26
We're using a network in place of that.
4:28
So we're increasing our ability to exchange goods and services.
4:33
And then on the AI side, I see it as you know, in the same way that the steam engine was
4:38
an answer to to overcoming the limitations of human ability physically, AI is now the
4:49
hope I don't know that it does now.
4:51
I don't think it does is the hope that it can now overcome our mental capacity and our
4:56
mental abilities generally by 600 times, but even more than 600 times.
5:01
I mean, if we had a GI, we could put a GI to building its own a si and we would potentially
5:10
see an even bigger savings.
5:13
But go ahead.
5:14
Go ahead.
5:15
I didn't have anything else.
5:16
I was going to let you go with that thought.
5:18
Oh, no, I think I lost my train of thought there.
5:21
Well, I think, I mean, to your point, like this idea that by setting free the value system,
5:27
right.
5:28
The thing that the sort of to use the analogy or to torture the analogy a little bit, the
5:33
engine and the input of our economic systems.
5:37
By setting that free, we can enable people to utilize the technology and the intelligence
5:43
that we hope that a GI will eventually generate to take us to that level.
5:48
Like you said, take us far beyond anything we've conceived and way beyond 600 times our
5:53
output productivity.
5:54
Because I think you could probably argue that GPT's and LLMs have already done that in some,
5:59
to some extent, with our thought output or content outputs, right?
6:04
I think there are studies about how much output there is from GPT's even currently.
6:09
And we haven't even begun to scratch the surface of what they're capable of.
6:12
Yeah.
6:13
Because I find in general our data tends to be reflexive, you know.
6:16
Right.
6:17
So I mean, that's kind of, you know, the history though, I guess to dive a little bit more
6:21
into that or to elaborate a little bit more on that is just that, you know, the new technologies
6:25
enable things that were never thought of, right.
6:28
So the harnessing of electricity or the harnessing of a steam turbine gave the possibility of
6:32
steady enough electricity or steady enough power output to use incandescent light bulbs,
6:37
right.
6:38
Which allowed us like to, which basically freed up the 24 hour world that we all live
6:41
in now, right.
6:42
Whereas before you were limited to, you know, how many candles you could put around you
6:46
and, you know, utilizing at a given time and there was limitation to how much light that
6:51
generated.
6:52
You know, eventually now we have skyscrapers full of incandescent light bulbs and all sorts
6:56
of other, you know, technologies that were never even dreamt up before the industrial
7:00
revolution.
7:02
The hope is that that will repeat itself but even on a much larger scale, an incomprehensively
7:06
larger scale than it did the first time around.
7:09
Yeah.
7:10
I think another example of that is sort of like the fossil fuel generation industry,
7:15
right.
7:16
I think none of us disagree that fossil burning and consumption of fossil fuels is bad for
7:21
the planet.
7:22
It's bad for our overall air quality.
7:24
It affects our ability to breathe and to succeed in the far future.
7:28
Maybe not as far as I think but that persistent demand for energy drove relentless technological
7:36
advancement.
7:38
So over the past century you've seen improvements in extraction techniques, extraction technologies,
7:43
horizontal drilling in wells, right, and look I'm not a fan of fracking.
7:47
I, especially where I live, I think it actively hurts me.
7:52
I've sat through earthquakes that I've never, never experienced an earthquake before in
7:57
the middle of the United States but because of fracking we're seeing those things increase.
8:01
The US coal miner for instance increased ninefold.
8:05
Their production, their output increased ninefold in 50 years.
8:08
So these innovations have always pushed back the predictions of resource scarcity by giving
8:17
us a higher output and driving innovation through demonstrating that really reserves
8:24
are really not a fixed geological quantity but a dynamic sort of economic and technological
8:32
one.
8:33
Right, and essentially what it amounts to as I see it is that humanity is hurtling down
8:39
the road at an ever-increasing speed.
8:42
The question is where are we headed towards?
8:43
Paradise and utopia or are we headed towards destruction?
8:47
And the signs change pretty frequently depending on what you look around and see as to which
8:53
road we're on or whether we're on the road that could lead to both.
8:56
And the decisions we make today are probably going to determine where we end up.
9:00
Yeah, 100%.
9:01
And technology's just going to accelerate that speed and multiply that effort.
9:05
Yeah, and I mean I guess sometimes I think we do have to kind of sit back and say to
9:10
what end.
9:11
Right.
9:12
And I think that reason differs for everyone.
9:16
I think I have my own reasons for wanting to see it.
9:20
I think a couple of ways that we've already seen that I think are really good examples
9:26
in already with text generation is the ability to use specific A&I models to mimic the replication
9:37
and synthesization of proteins that can produce new chemicals, new cures, new, it helps to
9:47
fight diseases.
9:49
Now I think that there's pluses and minuses to that both.
9:53
Maybe, through these sort of protein synthesis, we can overcome and help eventually cure cancer,
10:01
but at the same time that also might open up a whole new world of toxic chemicals that
10:07
people can use for biological warfare.
10:10
Right?
10:11
I think that there are tradeoffs that we have which is, I mean obviously it's a great reason
10:16
for regulation, which I'm pro-regulation around these sorts of things for that very reason.
10:22
But it's like saying, look, we could find the cure to every known world illness by employing
10:29
an A&I model to do it.
10:30
Yes, it takes a lot of electricity, but we could save lives in the long run.
10:34
So should we completely ignore our ability to cure these diseases because it uses a lot
10:40
of electricity and that could eventually kill the planet as a whole?
10:44
Or do we say, no, why don't we point it at those problems and develop those technologies
10:50
to solve those problems as they arise at the speed of faster than human thinking?
10:56
Right.
10:57
And who do we allow to utilize those computers?
10:59
Because on the one hand, there's the worry about a rogue actor who might discover some
11:03
sort of terrible new virus and unleash it on the world.
11:07
But then on the other hand, there are pharmaceutical companies who know that they'll be putting
11:10
themselves out of business if they cure all the illnesses.
11:12
And are they going to be able to resist the temptation to create new illnesses on their
11:17
pipeline?
11:18
Hey, that's why I'm a socialist, I'm just saying, you know what I mean?
11:21
Not right there, that's reasonable.
11:22
A democratic socialist, not by force, but through democratic means.
11:26
I think one of the things that I wanted to point out, and look, there's detriments to
11:29
this and I hope we can poke holes in it.
11:34
And what I think you're saying, Will, is that as we get down this road that there kind of
11:43
is no way back from, right?
11:44
You wouldn't be able to just reel AI back in and reel cryptocurrency back in.
11:49
Obviously, in the cryptocurrency sake, part of the design was such that it could not be
11:55
reeled in by any single entity or any single force.
12:01
But in the sense of AI, I mean, there are ways in already putting measures in place
12:07
to limit harm, I think is that we have also, as things have come around, increased, A,
12:17
our knowledge, right?
12:18
Like scientifically, when we were putting coal out in the world, we didn't think, "Oh,
12:22
this is going to cause global climate change."
12:26
We basically just thought we have a need and that is we need more energy.
12:31
We need to put light bulbs in everybody's houses so we could take them away from having
12:36
only eight hours of every day with daylight or 12 hours of every day with daylight to
12:42
we don't need people dying of useless diseases, but we also don't need to trade our ecological
12:48
wealth for curing diseases because either way, we're going to die.
12:57
Which one's going to be the worst?
12:58
And I think that doesn't have to be a decision that we live with every day.
13:02
I mean, sorry, that we live with one time.
13:05
I think it's something that we can live with every day and say, "Look, are we getting closer?"
13:11
That to me is kind of the epitome of science is collecting that data and making meaningful
13:17
observations, and then adjusting to those observations.
13:22
Right, it's feedback.
13:24
It's using our feedback and using it to take us to the next step and make the next decision,
13:30
the scientific method.
13:31
So let's talk a little bit about some of the, and even before AI and cryptocurrency, we've
13:36
had a need of wanting and needing to generate more and more and more electricity as we increase
13:44
our technology, right?
13:47
The current power grid is in a miserable state.
13:50
We know that.
13:51
We need lots of infrastructure improvements.
13:54
And we're seeing major swaths of Texas being browned out at certain times.
13:59
They talk about blackouts in California, and we've seen demand-related problems.
14:06
And I think it's more than just one thing.
14:09
First of all, I think that it's because we generate electricity for money, we create false
14:16
scarcity.
14:17
And if we don't create a whole lot of generation, if we don't have a lot of means of generating
14:23
electricity, we keep the costs higher, right?
14:27
Because then we keep the threat of losing it that much.
14:30
And as we increase not just these technologies, but as the temperature rises globally, there
14:38
will be an even greater need for what?
14:41
Temperature control, for air conditioning, for heating, in incredibly cold winters and
14:46
air conditioning in hot summers.
14:49
That demand is going to increase regardless, as long as we continue to increase as a species.
14:56
We've agreed.
14:57
We're not getting any further away from utilizing that energy.
14:59
That's where we're headed, is more and more consumption.
15:03
And the hope is that these technologies will allow us to keep pushing the limits further
15:08
and further while not destroying ourselves in the process.
15:11
Yeah.
15:12
So let's talk about some of the advances that we've already seen in the current electricity
15:18
and power generation space that are sort of driving innovation.
15:22
I think one thing that has been kind of flip-flopped back and forth and has somehow been made political
15:29
is something like nuclear power generation.
15:36
Ultimately, it's terrible, but it really does come down to steam engines.
15:41
And nuclear power generation facilities are essentially just using nuclear power to boil
15:48
water because it has a greater output to input ratio.
15:55
And then steam engines turn and that generates electricity.
15:58
We're still on the same standard, right?
16:03
But it's incredibly difficult to enrich plutonium, to find plutonium, and it's also very dangerous
16:11
to mine it.
16:12
It's a very dirty process, both on the front end and the back end.
16:15
And then there's also the problems, obviously, Chernobyl, if there's some sort of radioactive
16:21
chemical spill in this scenario or radioactive condition.
16:25
Right.
16:26
But we're able to hopefully utilize these, again, these advancing technologies to offset
16:31
those dangers to give us new safety procedures, new development.
16:34
There's a lot of ways, and they're also advising us or simulating these sorts of reactions
16:40
and telling us how to make them safer, how to make them more efficient, and how to increase
16:44
the output without endangering ourselves or blowing ourselves into oblivion.
16:48
Yeah.
16:49
And there's also been alternatives proposed, right?
16:53
There's two approaches when you're faced with the reality of the dangers of a particular
17:00
type of energy generation.
17:01
In the case of nuclear, yeah, nuclear waste and nuclear spills and enrichment of nuclear
17:08
material is dangerous, inherently.
17:11
What can we do to overcome that?
17:13
And you see that has already driven productivity gains in energy.
17:20
China, this year, brought on their first thorium nuclear reactor.
17:26
Now thorium is a little bit easier to cultivate, to enrich.
17:30
I don't think the returns are as high, but the dangers are lower.
17:33
So they're bringing on this thorium.
17:36
From what I understand, thorium is still pretty dirty.
17:38
It's similar to plutonium, but not as much so.
17:42
Again, a trade off, right?
17:45
For additional progress, you got to have a little bit of malaise.
17:51
We've made significant, in the past three years, we've made a significant movement toward
17:56
nuclear fission.
17:58
For the first time ever, I believe China, and then it was reproduced in California,
18:06
there was a nuclear fission reaction that was for the first time ever a net gain of electricity.
18:17
So it required a lot less to put out than it did, or than it consumed, which is enormous.
18:23
I mean, that's massive because in a fusion reaction, you're not going to have the actual,
18:31
the explosive part, right?
18:33
If you think about it, it goes inward and creates cold.
18:36
It doesn't implode as opposed to exploding.
18:40
You're joining two atoms together rather than exploding them apart.
18:47
Which again, is going to, I mean, it could be massive.
18:50
If we could get nuclear fusion to a point where the net gains are just interminable,
18:58
you could literally put a fusion reactor, in terms of safety, you could put a fusion
19:02
reactor at everybody's house.
19:04
It could be just as common as a hot water heater in 30 years.
19:08
I don't know why you'd want to heat hot water, but.
19:11
I'm looking for the Mr. Fusion, you know, the back to the future where you just throw
19:15
your trash in your car, engine in your car, and it generates all the power you need.
19:20
But realistically, that's not far from what we're moving toward.
19:24
Yeah.
19:25
But what has driven that progress?
19:30
The need for power, right?
19:31
The need for more power.
19:32
The necessity for that electricity.
19:33
Exactly.
19:34
Now...
19:35
Thus, the source and the potential solution to our problem.
19:39
Exactly.
19:40
Now, I'm not saying fusion is necessarily, but it could end the scarcity of electricity.
19:45
But like you said earlier about pharmaceutical companies not wanting them to find cures for,
19:51
you know, the cancer drugs and things of this nature, there are also gigantic lobbies, gigantic,
19:58
billionaire, trillionaire lobbies in the energy creation and energy dispersion space that
20:10
absolutely would not want that technology to get into the hands of the average human
20:13
being.
20:14
That's right.
20:15
I think it's fortunate in some sense that open AI may be worth more than ExxonMobil,
20:22
once it's publicly traded, in the sense of where that economic high energy flows.
20:28
Where it doesn't flow, right?
20:30
Now, I would think that somebody like Sam Altman and open AI might be incentivized to
20:38
invest in that sort of thing.
20:41
But if supposedly, and this is what they tell us about capitalism, right?
20:45
The necessity drives the innovation.
20:47
So if open AI has a necessity for unlimited amounts of power, they have no choice but
20:55
to invest in things that will continue to allow them to consume that power.
21:00
Then they will invest in the clean creation of it.
21:05
And hopefully it trickles down to us.
21:07
Now, we know that doesn't happen.
21:09
That's why I think we as a society should say, let's fund, let's self-fund as a country,
21:16
as a nation, as a world.
21:18
Let's figure out how to self-fund fusion so that we all can have a fusion reactor.
21:23
Everybody's car is powered by fusion and we're driving around in the cleaner world because
21:26
we're not burning coal, we're not burning gas, and we're not burning diesel.
21:32
And then we own the means.
21:34
So then what do you do?
21:35
Now, in my opinion, I don't have a problem with private companies making money by generating
21:40
electricity and putting it into a nationally owned grid.
21:44
I believe that we as a democratic socialist should own the grid as a whole and we should
21:50
also invest as a people in reactors of whatever sort, even fine with a nuclear reactor, like
21:58
a fission reactor to generate electricity because that is the best technology that we
22:03
currently have at a large scale.
22:06
Thorium is still not at a large scale.
22:07
It's still being test bedded in China and fusion is still not even remotely in the realm
22:15
of being able to scale.
22:17
So we should be in the generation business because we're seeing rising power costs across
22:26
the nation now and they're saying it's because the rising cost of fuel.
22:31
But they're using coal and they're using diesel to drive generators, to generate electricity.
22:39
And they're not going to have a reason to change from those things if they can just
22:44
pass the costs on to the people who have to pay the electricity bills.
22:47
Right, as long as they can keep up, they'll continue to burn the most available and the
22:53
most profitable source for them.
22:55
Exactly.
22:56
So I will say it is my position that innovation will be driven by the need in this space.
23:04
So we as a show would be remiss if we didn't talk about the very real, very raw reality
23:13
of how much this sucks in the everyday for specific people.
23:19
Croc has built a giant AI farm in Georgia.
23:26
Open AI, no, Metta, Metta is building a giant AI farm in Louisiana.
23:32
Because why?
23:34
Low cost of energy.
23:36
Right?
23:37
Relative to the rest of the country.
23:39
And it's putting people out of house and home, obviously, by scooping up land.
23:46
Nobody wants to live near those things.
23:47
They're loud, they're hot, and they're hard to be around.
23:53
Right.
23:55
And they're not really approved by the people.
23:59
Now obviously, who would want, who would sign up to have an AI farm in their backyard?
24:04
Nobody, unless they were somehow benefiting from it.
24:08
So I think it's important in these scenarios to use legislation to say you can build an
24:13
AI farm, but you also have to build a power generation plant, or you have to build a power
24:18
generation facility and contribute, not just remove.
24:28
But the real fact is, exactly.
24:31
And the municipalities are saying, yeah, come on in, we'll give you tax breaks even.
24:35
The problem is, those costs get passed through, sent down to the average people who live in
24:42
that community.
24:43
And it's wrong, it's not right.
24:45
But I want to talk a little bit from a perspective of skepticism, right?
24:51
Let's talk about a little bit of cryptocurrency.
24:55
Right?
24:57
I think there are some practical and ethical challenges of crypto.
25:07
Do you have any opinions on these?
25:09
We'll see.
25:10
I'll ask you too specifically.
25:11
I was talking to the audience, but I also am talking to you.
25:17
So first of all, I think that here's why crypto sucks.
25:20
Right?
25:21
Volatility.
25:22
Crypto, I said the spot price today was 111,000.
25:26
About 112,000.
25:28
It was 127,000, 123,000 two weeks ago.
25:33
So if you own one Bitcoin, in the last couple of weeks, you've lost $12,000.
25:38
$15,000.
25:39
If that's your 401k, then that could be pretty scary.
25:42
Exactly.
25:43
And guess who just legalized external sources inside a 401k.
25:49
Yeah.
25:50
So I think that we're not talking about just volatility, guys.
25:54
We're not talking about like the index goes up a point or two and your 401k gains $12
26:01
one day and $127 the next or whatever.
26:04
We're talking about extreme price volatility.
26:08
That kind of limits cryptocurrency's usefulness as a stable medium of exchange.
26:12
They've tried to fix this introducing stable coins.
26:16
The US Mint is actually talking about a USDC, US dollar coin that's tethered to the dollar.
26:21
There's also Tether, which is another cryptocurrency that's directly tied to the value of the US
26:27
dollar.
26:28
But it's an asset that can fluctuate by double digit percentages.
26:33
We're talking 30, 40% in a single day.
26:37
And that's really impractical for commercial transactions.
26:40
If you take a hundred bucks and you buy a hundred dollars worth of Bitcoin, that's not
26:45
a lot of Bitcoin.
26:46
It's about a hundred thousandth of a Bitcoin, right?
26:50
And that is worth a hundred dollars when you buy it.
26:54
At the end of the day, that hundred dollars could only have the spending power of 40 when
26:59
you're talking about that kind of volatility.
27:01
So unlike payments made with a credit card or a debit card, those sorts of crypto transactions
27:09
have no consumer protections either.
27:11
It's so if you get defrauded of hundreds of thousands of dollars of cryptocurrency,
27:17
no, ain't nobody can help you.
27:20
They have those.
27:21
They have those coins now.
27:23
They're tied to their cryptography keys and there's no protection.
27:29
If you make a mistake, just fat finger an address that you're trying to send somebody
27:34
money to.
27:35
You can hope they send it back voluntarily, but there's no mechanism to get that money
27:40
back.
27:41
There's also the risks, the idea of a centralized system is contrasted against the practical
27:50
reality of how we interact with it.
27:53
So the underlying blockchain, the ledger that we were talking about earlier, it is really
28:00
secure by design.
28:02
Unfortunately, most people store their assets on exchanges which are centralized.
28:10
Why for convenience, the ability to withdraw, to buy more, to add it, to purchase it, to
28:16
trade it, to make money with that money.
28:19
Or in my case where I got hung up is it guaranteed a return that was greater than the market
28:26
average 5.9% or something to that effect to basically let them borrow your cryptocurrency
28:33
and use it to trade inside of a volatile market and hopefully get something back from it.
28:43
There was a whole lot of mess with the Gemini exchange and you can feel free to look that
28:48
up.
28:49
I was very lucky in that I was able to get it all back eventually, but it took two and
28:54
a half years.
28:56
Otherwise it was just sitting there in flux and it was not a small amount of money.
29:01
When you put money in a bank, if there's a large scale hack, if you're defrauded, you
29:07
go to the bank and you say, "Hey, look, this is fraud.
29:09
I didn't do this.
29:10
This is not my transaction."
29:12
You were hacked or somebody generated my credit card number.
29:16
That's not mine and you get your money back because that is a built-in protection.
29:20
Our system builds in those protections through FDIC insurance.
29:25
Your money is guaranteed up to $250,000 or whatever.
29:28
Same thing with the NCUA for credit unions.
29:31
There is no organization like this.
29:33
Can we make it?
29:34
Absolutely.
29:35
Could it be done?
29:36
Yes, absolutely.
29:37
But who would guarantee it?
29:39
Essentially, it would be a form of insurance, which what do we know about insurance?
29:43
The only people who ever make money using insurance are the people who create insurance.
29:48
They don't stay around that long to give away all their money, all the premiums.
29:54
Exactly.
29:56
In order to get some sort of regulation, we would have to reasonably create a cryptocurrency
30:04
that needed to be centralized in order to protect it.
30:07
It's like the Wild West in the sense that you can use it to trade money, but if you
30:12
mess up, that's on you.
30:13
It's why the libertarians love it.
30:15
They're like, "It's all about your decisions."
30:17
Also, there are smart contracts.
30:20
I didn't really talk about that, but there are smart contracts that automate transactions
30:24
on the Ethereum platform that they could contain actual coding errors or flaws that can lead
30:34
to leaking of coins, that could lead to catastrophic losses to terms of 100% or more.
30:44
Also, obviously, I'm going to point it out, it's almost not even worth considering illicit
30:49
use.
30:50
The currency has been the number one way to do something illegal in the world since its
30:55
inception, mainly because nobody can really track it or see it.
31:03
It's not tied to your identity.
31:05
It's just a thing that can be traded.
31:08
But I would argue, cash is too.
31:12
Cash is absolutely the same concept.
31:15
That's why in the '70s and '80s movies we saw about the cocaine trade and when the CIA and
31:23
the FBI were bringing in massive amounts of drugs into the United States, you saw everything
31:27
was dealt with in cash.
31:29
So it's the same concept.
31:30
It just became the new cash.
31:31
It was an easy way to conduct it electronically without having to carry around large sums.
31:37
It's why we can't take a flight with more than $10,000 in cash without declaring it
31:42
and letting somebody know that we're leaving with massive amounts of cash.
31:45
So it is used for illicit reasons.
31:49
To kind of tackle the downsides of AI as well, now we're talking a high level from an ethical
31:58
and moral standpoint on these topics.
32:01
So it's not just about the power consumption of it.
32:05
Algorithmic bias, you know, ultimately these models are just giant repositories of things
32:11
that have already been written in the world and on the internet.
32:14
That means that along with it, all of these are trained on these datasets and scraped
32:20
images as a reflection of our world.
32:25
That means all of those social biases, all those prejudices and stereotypes about gender
32:30
and race and culture and everything else, those models reflect those biases and then
32:38
can recreate them and sometimes even amplify them in their output.
32:43
There are a lot of examples of this.
32:45
You can check the internet for that, especially around like AI powered hiring tools.
32:51
They've specifically shown bias against female candidates.
32:55
Facial recognition systems have lower accuracy in regards to people of color.
33:00
AI models used in criminal justice system have actually been shown to disproportionately
33:05
flag black defendants as being high risk of reoffending.
33:10
That leads to discriminatory outcomes and the reinforcement of those inequalities.
33:17
Then there's the all encompassing and very well publicized misinformation and hallucinations
33:23
that have been well documented that AI models have the tendency to hallucinate.
33:28
Why?
33:29
Well, it's a great question.
33:30
We're asking you to generate something new from what it was trained on.
33:34
That means there's a possibility that whatever it's making up is completely nonfactual and
33:41
nonsensical.
33:42
Well, most of the time it's sensical, but it might not be based in fact.
33:46
It doesn't have a comprehension of the data.
33:49
So if you ask it a question about the speed limit in Wisconsin, it doesn't understand that
33:55
the speed limit in Wisconsin has a specific legislative paper trail.
34:00
You could ask me about the speed limit in Zootopia and it'll tell you something because you're
34:05
asking it a question, right?
34:07
Exactly.
34:08
Exactly.
34:09
There will be some sort of determination.
34:10
Exactly.
34:11
It doesn't know that there's no such place.
34:12
Think of it kind of like my nine year old who hasn't learned the power of "I don't know"
34:16
yet.
34:17
You know, like that's the power to shut down a conversation.
34:19
You ask him a question, they say, "I don't know."
34:20
That's a teenage thing.
34:21
But as a nine year old, he will answer every question you ask him.
34:24
He just will answer it with total confidence whether he knows it or not, and that's how
34:28
AI responds.
34:29
So hopefully it'll grow out before it destroys us all.
34:33
I do that myself now.
34:34
I do that all the time.
34:37
I say if you just say it with confidence, they'll surely believe it.
34:40
Also...
34:41
Go ahead.
34:42
As I was saying, unfortunately, hallucinations didn't originate with AI.
34:46
No, they did not.
34:48
And the truth is, if we're trying to create these language models to sound and be more
34:53
like we are, we hallucinate, we dream, we have ideas that are nonsensical and not based in
35:00
fact.
35:01
But what does this do?
35:02
It means that scammers, people who are nefariously using AI, sound more convincing.
35:08
They sound more real.
35:11
Those letters that we... the emails we used to get from the Zimbabwe government that reminded
35:16
us how well off we were, and how we had magically either won their lottery or something to that
35:25
effect.
35:26
They were hoping for some sort of barrister situation where you had to send them money.
35:33
Those were easily determined to be AI.
35:38
And for a very specific reason...
35:41
I mean, sorry, not easily determined to be AI.
35:44
They were easily sussed out as being fraudulent.
35:48
And why do you think that is?
35:50
Well, because there were spelling errors, there were grammar problems because they were
35:55
clearly from people who didn't speak English as their first language.
35:59
But you know, now with AI, they can use that same script that they had before and run it
36:05
through an AI and say, "Hey, clean this up.
36:07
Make it more formal.
36:08
Make it match American English."
36:09
Those sorts of things.
36:11
And still get away... and more now than ever, they're actually targeting people and getting
36:17
away with it because it looks more legitimate than it did before.
36:21
You know, I remember Kim Commando, who was a radio-popular host that helped old people
36:27
with computers back in the day.
36:29
And my grandparents when they were alive loved to listen to her radio show.
36:34
And she would say like, "Here's how you pick out these things and look for grammar mistakes
36:39
and look for spelling errors.
36:41
But in today's scams, there are no spelling errors.
36:45
There are no grammar mistakes.
36:46
It sounds exactly like a person who would coming from... and they're using proper language.
36:51
You know, they're using barrister instead of lawyer, and they're using, you know, these
36:55
sorts of things.
36:57
Another thing, privacy and data, right?
37:00
When you train these large language models, it requires a colossal amount of information.
37:07
Most of that is scraped from public websites.
37:10
It's scraped from forums like Reddit and Substack and those sorts of things.
37:15
Stack trace, stack overflow, lots of stacks.
37:18
And from social media... all of them.
37:22
And is pulled from social media without the knowledge or consent of the people who created
37:29
that content.
37:31
That raises a lot of privacy concerns, not to mention people's personal stories, their
37:36
sensitive information, pictures of their children, copyrighted artworks, copyrighted music.
37:42
All of these things are ingested into these models.
37:45
So there is a not insignificant risk that models could inadvertently leak, right?
37:53
Or reproduce private data in their responses or worse, maliciously be coerced into producing
38:01
this information that was used to train them.
38:03
There's also a good bit of accountability and transparency issues in these AI models, right?
38:10
We as consumers don't have the ability to produce an AI model on the equipment that we have
38:14
in our homes.
38:16
They require massive amounts of information.
38:18
The average model right now is right at $22 million.
38:23
You need that amount of money to train one of these models.
38:27
So many of the most powerful of these really are just black boxes, like we were talking
38:32
about earlier with the computational aspects of it.
38:36
But in the informational aspect of it, they're just black boxes.
38:40
Their internal workings are so complex that even the people who do this for a living cannot
38:46
fully explain why a model produced the specific output that you're seeing on your chat.
38:54
Like you said, when model 4.0 was so friendly, and then 5.0 was cold and objective, people
39:00
were like, "What just happened?" and open their eyes, shrugging their shoulders.
39:04
They don't know.
39:06
Exactly.
39:07
And that opaqueness is a real challenge for accountability for a very specific reason.
39:16
When an AI system hurts someone and we're seeing it hurt people, whether by providing
39:23
dangerously incorrect medical advice, causing a self-driving car to crash, or making a biased
39:30
loan decision.
39:32
And it's really often unclear who's responsible for that.
39:37
Is it the user who prompted the AI?
39:40
Is it the people who train the AI?
39:43
Is it the company that made the AI available for you to use?
39:47
Or was it the creators of the biased data that were used to train it?
39:54
So without transparency in how these systems make decisions, the ability to establish any
40:01
sort of accountability or transparency is almost impossible.
40:05
All right, let's talk about the innovation, Fowl.
40:08
We have built for you a narrative of historical information that presents progress as being
40:15
a driver for additional progress.
40:18
Progress in one area creates the need, which drives innovation in another area.
40:23
And it is a valid one.
40:25
I do want to say that, but it's incomplete.
40:29
Innovation is really not a moral force.
40:35
It's kind of a neutral process.
40:38
It's a tool.
40:39
Gosh, where do you heard that argument before?
40:41
It's a tool, and it can be used for good or bad.
40:46
It just magnifies the effort.
40:51
And that's the point of a tool to begin with, is to expand our capabilities into further.
40:58
So the final and really the most fundamental critique of ourselves challenges this, this
41:07
sort of techno-optimistic premise that a high demand for these resources will lead to positive
41:16
and beneficial innovation.
41:18
That is my belief.
41:20
I will say that there is data that actively backs it up.
41:24
Will, is it your belief that you've seen it come to pass in your lifetime?
41:31
I have, but more importantly, I think it's more of just hope that I have to hold on to,
41:35
because the alternative is just the dark ages again.
41:38
Right?
41:39
We've already gone so far at this point.
41:40
The hope is to completely borrow something entirely unrelated.
41:44
If you're going through hell, keep on going, as they say, right?
41:47
Like the idea is the only way out is through.
41:50
The only way out is through.
41:51
Exactly.
41:52
I think high demand, I think if I had to word it correctly, I would say high demand provides
41:58
a motive for innovation, but it doesn't actually dictate the ethics of the outcome.
42:06
Exactly.
42:08
So demand for cheap textiles in the 19th century drove innovation in factory machinery and
42:16
the steam engine, but it also led to horrific labor conditions, child labor, dangerous urban
42:23
pollution, both in the water and in the air, but the demand for agricultural efficiency
42:31
in the 20th century drove innovation in chemical fertilizers and pesticides, but also led to
42:37
widespread water contamination and ecological damage that we're still dealing with the consequences
42:42
of.
42:44
Historical innovations that power our modern world right now, the steam engine, the internal
42:49
combustion engine were based on fossil fuels, the extraction of use and use of which have
42:56
led directly to our current climate crisis, a massive and potentially catastrophic negative
43:03
externality of that progress.
43:05
So a closer look at the consequences of high resource demand kind of reveals this pattern
43:11
of negative outcomes, right?
43:13
That often come along with the advancement of technology.
43:19
The extraction of raw materials, whether they be coal for a steam engine or lithium and
43:24
cobalt for batteries is a very energy intensive and environmentally destructive process.
43:31
It very often leads to really terrible and sometimes irreversible ecological damage.
43:38
I mean, soil degradation, water shortages.
43:41
It pollutes the air and the water both.
43:44
And the less obvious one for you biologists out there is the loss of biodiversity.
43:49
The global pursuit of resources is inherently intertwined with social conflict and inequality.
43:56
The process of material extraction has been linked to severe human rights violations and
44:02
forced displacement of local and indigenous populations and acute health problems for
44:08
the communities that surround these things from contamination.
44:11
We're already looking at the prospect of all of us losing the ability to work in white
44:17
collar jobs at all because of the possibility of an eventual AGI.
44:22
Which makes alignment all that much more important, right?
44:27
Right.
44:28
The economic benefits of the resource extraction flows to multinational corporations and governments
44:37
and developed nations, while the ecological and social costs are actually burdened upon
44:45
by the actual population of less developed countries.
44:48
According to the UN actually, natural resources play a key role in 40% of intra-INTRA state
44:58
conflicts with profits from their sale being used to finance armed militias.
45:05
And look, this is a crucial point in us building up our current moment.
45:15
I mean, this immense demand for energy and materials generated by A&I and by cryptocurrency
45:23
could spur development, right, of clean fusion power which effectively would end scarcity
45:33
for energy on our planet as a whole.
45:36
But it could also lead, and if I'm also looking at history, to a giant global grab for resources
45:45
needed to build that new energy infrastructure and the computing hardware that it will power.
45:52
And that outcome is not guaranteed to be positive.
45:56
So it's not going to be determined by like… go ahead.
45:59
It's guaranteed not to be positive for some people, even if it's generally positive for
46:04
the majority of us.
46:05
Some of us will have to pay the cost.
46:08
That is a reality.
46:09
I view it kind of like I view war.
46:13
And look, I'm not a fan of war, but the theory behind it was that a few put up their own
46:20
lives at stake to protect the majority.
46:25
And I don't think it's necessarily right to equate this with war.
46:31
But it's not going to be determined by the technology itself, AI or crypto.
46:37
It's going to be determined by the choices that we make as people regarding the regulation,
46:43
the governance and the ethical deployment of these technologies.
46:48
Right.
46:49
We've got access to a lot of power.
46:51
Now it comes with a responsibility.
46:53
Thanks, Spider-Man.
46:54
And Peter Parker over here.
46:56
Hey, Uncle Ben.
46:57
No, it was Uncle Ben.
46:59
I was about to say it was the uncle.
47:00
It was not Peter Parker.
47:02
So okay.
47:03
So where does that leave us, right?
47:04
In terms of a conclusion, I came across a fun word that I really liked in doing the
47:09
research for this and it was trilemma.
47:11
I mean, it makes sense.
47:12
I heard of dilemma.
47:14
Most of us have heard of dilemma, but trilemma or trilemma is a fun word that I really enjoy.
47:21
I think it creates this trilemma of innovation and consumption and responsibility.
47:29
So I feel like we're currently standing like dead center and we have to we take care of
47:35
all of these things because of the profound implications of the twin revolutions of crypto
47:41
and AI that we're, regardless of whether you want it, are being put into the real world
47:48
for us to contend with.
47:50
So this podcast has been an effort to show that these technologies aren't like abstract
47:56
concepts, but they're real powerful forces with real world consequences, right?
48:05
Especially with their immense and endless, meaningly endless growth of desire for consumption
48:13
of energy.
48:14
So where we go from here is not like a simple choice between, "Hey, I'm going to embrace
48:21
AI and I'm going to embrace cryptocurrency" or flat out reject it, right?
48:25
And decide to be Amish and go live in Pennsylvania.
48:28
But it's really, it's a very complex navigation of this trilemma.
48:35
And opting out is no guarantee that you'll avoid the consequences and in fact, maybe
48:38
just abdicating your ability to do something about it.
48:41
Exactly.
48:42
So if you were to take each branch of this, right, innovation, we have potential benefits
48:50
of AI and decentralized technology that we talked about earlier.
48:56
Not only that, we're talking about the possibility of increased productivity.
49:00
Now, productivity for what?
49:02
You know, again, this is a tool.
49:04
We get to determine that.
49:06
We've got new forms of economic organization, accelerated scientific discovery and solutions,
49:13
possibly to some of humanity's most pressing problems.
49:16
So if we turn our back on the potential innovation, we would literally be rejecting like a pretty
49:22
powerful engine for future progress.
49:26
The second point is consumption, right?
49:28
Like as we talked about earlier, that innovation comes at a steep environmental cost.
49:34
It doesn't matter how you look at it.
49:35
The silicon, from the silicon of the chips to the coal that we are apparently now, once
49:42
again, all excited about burning to produce electricity inside of our electrical grid.
49:48
That is the energy demand of these technologies is already on the scale of like coal nations,
49:54
right?
49:55
Like, and growing exponentially.
49:56
That is, I mean, that drives a voracious appetite, not just for that electricity, but for the
50:02
raw materials needed to build the data centers, to build the processors, the new energy infrastructure
50:08
required to sustain them.
50:11
And if we don't instill and enforce a system of checks and balances, it really could threaten
50:18
to undermine our climate goals.
50:21
And I don't mean the ones that our current president has decided to remove us from.
50:28
I mean our individual personal climate goals and also exacerbate the destruction that we're
50:37
already seeing at an environmental level, which brings up the most important part of
50:42
all of this.
50:43
It's about responsibility.
50:45
We cannot have the benefits of innovation without managing the consequences of consumption.
50:52
And those have a tangible effect on everyday people, not just multinational corporations.
50:59
And that requires a distinctly profound sense of responsibility to develop and deploy these
51:09
technologies in a way that is ethical and sustainable, but equitable too.
51:15
Absolutely.
51:17
And that means confronting the risks of the algorithmic bias, the misinformation and yes,
51:24
the illegal use, right?
51:27
And I mean, whatever.
51:29
I guess I personally don't really care if some guy from Montana really needs LSD and
51:34
can figure out a way to get it through the internet.
51:36
That doesn't really affect me that much.
51:40
But it's about ensuring that the pursuit of new energy sources doesn't just recreate all
51:48
of the crappy historical patterns, right?
51:51
Of social and environmental exploitation that often come along with resource extraction.
51:59
And we have real world examples of those.
52:01
Plenty of historical examples.
52:03
Yes.
52:05
But even looking toward the possibilities, when we look at possible hiccups in the road,
52:13
it's pretty clear that the future is not just a predetermined outcome of technological inevitability.
52:20
The pattern of demand driving innovation is powerful.
52:24
I mean, you can't deny that either, but it is not a moral guarantee, right?
52:29
It doesn't mean that it is...
52:31
A scientific fact.
52:32
Right.
52:33
It is morally good.
52:34
It does not mean that.
52:35
It is just a fact.
52:38
So that same demand for electricity that could actually get us there faster for clean fusion
52:46
power could also trigger a destructive scramble for the resources to build Fusion React.
52:52
That same AI that could cure diseases we've already seen through Russian propaganda entrenches
52:58
societal biases and completely obliterates public trust.
53:03
So our challenge really is to manage this trilemma, you're welcome, with intention and foresight.
53:10
Right?
53:11
So the path forward is really not just about the breakthroughs that we're seeing from this
53:16
technology.
53:19
And they're cool.
53:20
They're cool.
53:22
The idea of endless clean energy generation is amazing.
53:27
Not to mention computational efficiency of having a constant brain more powerful than
53:33
the human when continually working on a problem.
53:36
But it demands conscious, deliberate, and international conversations about the kind
53:43
of society we want to build.
53:45
And it has to be one that balances the power of that innovation with the very finite reality
53:52
of our planet and the enduring importance of our collective human value.
53:58
And that is the overlap.
54:03
We hope you enjoyed this podcast here today.
54:06
We worked really hard on coming up with all the information and doing all the research
54:10
to get it to you.
54:11
If you have any questions or comments, you can email us the overlap@fof.foundation.
54:22
You can hit us up and give us a follow over there on Blue Sky, the overlap podcast, as
54:26
well as Mastodon, the overlap podcast.
54:30
You can also check out our website, fof.foundation, and keep up with us there.
54:36
We will see you again soon, folks.
54:39
Well, maybe not see you.
54:41
Maybe we'll hear from you again soon, but you'll definitely hear from us later.
54:45
Thanks, everybody.
54:47
And thank you again, Will, for being here.
54:49
Thanks everybody.
54:50
Bye now.
54:50
[MUSIC]
55:00
[MUSIC]
55:10
♪♪
55:17
♪♪