Not in the field, so genuine question: what is the evidence/theory to support the notion that deep learning is at all a reasonable route towards AGI? As I understand, this is nothing like how actual neurons work - and since they are the only "hardware" which has ever demonstrated general intelligence hoping for AGI from current computational neural networks feel like a stretch, at best.
I cant talk with your customers but my initial thoughts are that you should focus with laser-like attention to your customer needs. There are a lot of "ands" in your product. Who is your ideal customer? Is it whales tracking multiple cryptocurrencies?
Is your ideal customer dev teams keeping their coin alive with DCA arbitrage across multiple exchanges?
Absolutely. We actually run our company on a discord with both public and private channels, just so we can strike up conversations with people.
We have three target customers: the experienced retail trader with multiple exchange accounts, the long-term hodler who’s looking for better indexing tools, and the crypto newbie who just wants to learn how to get started.
In terms of having a laser focus on a single area, honestly I both agree and disagree. Most of the unique and interesting paths we can take are based on having a wider foundational breadth of features (best price trades require multiple exchange connections, auto rebalancing requires a complete view of your portfolio, tax prep requires order management). All these things are interweaved within each other.
Totally agree that it’s a lot, but it’s also much easier to make a general product specific, than it is to make a specific product general. The tough thing is trying to strike the right balance between that foundation and really nailing a single aspect.
Right now, that single focus is custody and best price trades.
The biggest argument is the other coin networks that have been running for years in production without proof of work.
Nano and EOS have interesting consensus models. Nano manages to remove all inflation and transaction fees and is just a base level currency. EOS has an account fee and you have to rent resources but can do code execution on transactions (smart contracts). They both rely on accounts voting for representatives and have had similar problems with spam.
Nano has been running for > 4 years and EOS for > 1 years.
Can you elaborate on the network partition issue that you found with nano? As far as I know nano has been live for years without a doublespend or downtime and thats pretty impressive.
Yeah this is kind of where distributed systems theory comes up against practice, and I think is very interesting. On a theoretical level Nano is just fundamentally flawed, as are all coins which use this "everyone has their own blockchain" eventual consistency concept. The problem is this, as far as I can recall (I looked at this three years ago): imagine a network partition occurs, where two different exchanges are in different partitions. For whatever reason Eve still has network access to both partitions. She broadcasts a different transaction to each partition, each transaction sending all of her Nano to a different exchange. Since Nano uses proof of stake, these transactions have to get a certain number of votes. I don't recall the exact threshold (I think 1/3 of all extant Nano has to vote?) but it is possible for both transactions to be confirmed as long as the partition stays in place. Possibly long enough for Eve to convert her Nano to a different cryptocurrency and transfer it out. Once the partition collapses one of the transactions is rolled back and the exchange is left holding the bag. IOTA solves this problem by just saying "the partition containing the coordinator is the real partition" lol. Humorously the more centralized Nano voting power is the less likely this scenario is to occur, too - although then you get the existential threat of those big-vote accounts getting hacked! They have to stay online with their private keys accessible, after all.
As you say, this has never happened (to my knowledge). So it's possible that even though the problem is real, it will never actually occur in practice. In the end the real measure of success is how well these systems perform in the real world, so we could say this analysis of the theoretical problems doesn't hold much water. On the other hand if something like this were to happen, it would probably tremendously undermine confidence in the currency & destroy it overnight. So this could be more of a thanksgiving turkey situation where each day seems better than the last until thanksgiving arrives.
Thank you for your writeup. I understand the idea behind the network partition and I've been doing some digging. It looks like there are numerous commits attempting to address this edge case (they call it frontier cementing).
Here is a writeup I found on reddit that further elaborates this:
"""
There is a (currently hypothetical) edge-case where this could be reversed (not exactly the same as PoW chain reorganization due to forking, but similar.) In this edge-case:
We imagine the Internet has, right now, been carefully split by a malicious attacker into two exact halves
The Internet halves would be carefully contructed to each hold >60m online Nano votes. (This Internet fracture has never actually happened in Nano's lifetime. There are not 2 x 60m votes currently online.)
The attacker double-spends on a coffee (or house) in two of your stores worldwide just now
They then allow the Internet to heal
One of their transactions will be reversed by majority voting.
So while Nano transactions are fully-confirmed, they are not currently immutable.
Since there are not >120m votes currently online, it has not happened and cannot currently happen either. No previous-to-today Nano transactions will ever be reversed.
""" -- throwawayLouisa https://www.reddit.com/r/nanocurrency/comments/bgdshq/what_i...
GAN improvements that lead towards finding global minima for any given distribution mapping. GANs are versatile but still suffer from mode collapse and instability, drastically reducing their utility.
Capitalism can be thought of as a complex learning system. We, as a society, want people to contribute the most that they can to the overall goals of humanity and capitalism acts as a vehicle for that.
Comparing it to another popular complex learning system, neural networks, might provide insight on failure conditions.
Income taxes can be thought of as gradient regularizers.
Basically limit the amount of weights accumulated per training session (wealth per tax year). Estate taxes can be thought of as weight regularizers (that occur once per lifetime). Maybe it would be more palatable as a tax on wealth yearly.
One of the ways that networks fail (and capitalism can fail) is with extreme values on weights (wealth) on small portions of the network. There are ways to counter this on a neural network - some are heavy handed (l1, l2 regularization), but some are just clever without direct analogies (dropout, batch normalization, etc).
I wonder how the conversation would change if we had to calculate our income tax as a percentage of total wealth each tax year.
>We, as a society, want people to contribute the most that they can to the overall goals of humanity and capitalism acts as a vehicle for that.
For the point of view that it's not a very good vehicle at all, it's worth reading Oscar Widle's essay The Soul of Man Under Socialism, which is free to read online (of course!)
Here's a quick excerpt:
Now and then, in the course of the century, a great man of science, like Darwin; a great poet, like Keats; a fine critical spirit, like M. Renan; a supreme artist, like Flaubert, has been able to isolate himself, to keep himself out of reach of the clamorous claims of others, to stand ‘under the shelter of the wall,’ as Plato puts it, and so to realise the perfection of what was in him, to his own incomparable gain, and to the incomparable and lasting gain of the whole world. These, however, are exceptions. The majority of people spoil their lives by an unhealthy and exaggerated altruism – are forced, indeed, so to spoil them. They find themselves surrounded by hideous poverty, by hideous ugliness, by hideous starvation. It is inevitable that they should be strongly moved by all this.
If this is interesting, Kropotkin's The Conquest of Bread is also a really engaging read.
