4 ms·

ML existed since the middle of the last century but has only really found its legs as of late, but I didn’t see a lot of people make fun of it for not being per

ML existed since the middle of the last century but has only really found its legs as of late, but I didn’t see a lot of people make fun of it for not being perfect within a decade.

graypegg 3y ago
Fair point, but machine learning has also been creating value thru all of that time. There were products that used machine learning in the late 90s for many things. DeepBlue! I think you’re thinking of LLMs which only became possible with the shear amount of conversational data we have today. They obviously share a linage with ML, but it’s a form of training that was impossible several years prior. You can’t say that IBM DeepBlue was an early LLM and measure the timelines like that. If you know of any watershed moment for crypto that needs to happen, I would love to hear it! I just don’t see anything happening on the horizon.
- nwienert 3y ago
  I’ve been getting and continue to get value from crypto for a multitude of uses. There’s tons of people using it actively, so I just wonder why you choose to ignore that. A lot more than the very marginal value ML did for its first decades of existence. DeepBlue came many decades after the field came into being. And I’m very aware of LLM and the different types of ML. It still stands that it was very marginal for decades. But it would be easy to write some dismissive comment in the 80s that the only thing it’s done is play chess for millions of dollars - exactly like crypto haters do today despite it being a completely nascent field.
  - graypegg 3y ago
    Fair! I know there’s lots of wallet activity on any of the big chains right now. I’m a little suspicious of that representing real economic activity (filling the role of a currency, used as units of exchange) and maybe a bit more convinced it’s a casino poker chip ledger. It seems like the simple majority of people trade and transact out of their exchange, to my knowledge. If that simile of ML is like crypto is true, you stand to be quite rich! So best of luck! Best comments have net 0 right, so hopefully you’ll think of me and this thread when you’re a trillionaire eh? ;)
  - giantrobot 3y ago
    ML was not marginal for decades. It's been in regular use in a number of fields for decades. The current hype is an artifact of someone putting fun interface in front of an ML system. Hype around generative imagery and LLMs is new, not the technologies themselves. Crypto has been around for nearly a decade and a half and has been really useful for malware and scammers and not much else. It's fundamentally broken as a currency because most coins are designed to be deflationary so they behave like securities. It's fundamentally broken for micro transactions because transactions are ridiculously expensive and slow by design.
- mattdesl 3y ago
  AI and machine learning has had several decades to mature—and it had multiple “winters” where interest and funding waned. It’s possible in 60 years we might see similar leaps from distributed ledgers, zero knowledge proofs, verifiable computation, fully homomorphic encryption, and other tech being spearheaded by the crypto sector.
  - graypegg 3y ago
    It’s always possible! I’m no fortune teller. Many people have been wrong and maybe one day I’ll be laughing at my naivety. But for my own projects for the foreseeable future as measured right now, I’ll avoid. I just don’t see it becoming more than poker chips at your exchange of choice.
  - giantrobot 3y ago
    None of those things were invented or really developed by the "crypto sector" merely utilized by it. That's not a bad thing but it's disingenuous the crypto sector is really advancing any state of the art. It's implementing existing technologies mostly to facilitate scams and tax evasion.
    - mattdesl 3y ago
      All of these are primarily being funded and advanced by crypto. “Distributed ledger technology” hardly existed before blockchains/Bitcoin. Similar story with ZKP, which has gone from academic theory to real-world application (see: ZKSNARK) in the last decade primarily from blockchain research and funding streams.
      - graypegg 3y ago
        Aren’t zero knowledge proofs cryptographic signatures? I might be missing something, but that’s very much been in active use as part of any flavour RSA cryptography going back decades. Distributed ledger technology, though yes that is clearly a blockchain technology that wouldn’t exist otherwise. Fair point.
nologic01 3y ago
The comparison of incubation periods, adoption curves etc between crypto to ML cannot take us very far. The reason is that "ML" (as a proxy for algorithmic processing of a variety of data) is not particularly adversarial to pre-existing technologies. While this is not entirely true (there is always a hidden or explicit tension between automation and expert assessments) by-and-large this tension can be managed. The new tech gets bolted on the old. In contrast crypto sought to overthrow all pre-existing monetary and financial system technology with proposals that are naive, half-baked and ignore (or reinvent in unacknowledged manner) vital aspects. The practical implication is that crypto cannot carve a legitimate niche and keep iterating. The main outcome of a lot of wasted energy (in all senses) is to point out that indeed, digitization opens up the way to evolve the financial system. So much we knew, but now it has been drilled into the heads of large swaths of politicos, regulators, bankers etc that are completely tech illiterate.

ML existed since the middle of the last century but has only really found its legs as of late, but I didn’t see a lot of people make fun of it for not being perfect within a decade.

graypegg 3y ago

Fair point, but machine learning has also been creating value thru all of that time. There were products that used machine learning in the late 90s for many things. DeepBlue! I think you’re thinking of LLMs which only became possible with the shear amount of conversational data we have today. They obviously share a linage with ML, but it’s a form of training that was impossible several years prior. You can’t say that IBM DeepBlue was an early LLM and measure the timelines like that. If you know of any watershed moment for crypto that needs to happen, I would love to hear it! I just don’t see anything happening on the horizon.

nwienert 3y ago
I’ve been getting and continue to get value from crypto for a multitude of uses. There’s tons of people using it actively, so I just wonder why you choose to ignore that. A lot more than the very marginal value ML did for its first decades of existence. DeepBlue came many decades after the field came into being. And I’m very aware of LLM and the different types of ML. It still stands that it was very marginal for decades. But it would be easy to write some dismissive comment in the 80s that the only thing it’s done is play chess for millions of dollars - exactly like crypto haters do today despite it being a completely nascent field.
- graypegg 3y ago
  Fair! I know there’s lots of wallet activity on any of the big chains right now. I’m a little suspicious of that representing real economic activity (filling the role of a currency, used as units of exchange) and maybe a bit more convinced it’s a casino poker chip ledger. It seems like the simple majority of people trade and transact out of their exchange, to my knowledge. If that simile of ML is like crypto is true, you stand to be quite rich! So best of luck! Best comments have net 0 right, so hopefully you’ll think of me and this thread when you’re a trillionaire eh? ;)
- giantrobot 3y ago
  ML was not marginal for decades. It's been in regular use in a number of fields for decades. The current hype is an artifact of someone putting fun interface in front of an ML system. Hype around generative imagery and LLMs is new, not the technologies themselves. Crypto has been around for nearly a decade and a half and has been really useful for malware and scammers and not much else. It's fundamentally broken as a currency because most coins are designed to be deflationary so they behave like securities. It's fundamentally broken for micro transactions because transactions are ridiculously expensive and slow by design.
mattdesl 3y ago
AI and machine learning has had several decades to mature—and it had multiple “winters” where interest and funding waned. It’s possible in 60 years we might see similar leaps from distributed ledgers, zero knowledge proofs, verifiable computation, fully homomorphic encryption, and other tech being spearheaded by the crypto sector.
- graypegg 3y ago
  It’s always possible! I’m no fortune teller. Many people have been wrong and maybe one day I’ll be laughing at my naivety. But for my own projects for the foreseeable future as measured right now, I’ll avoid. I just don’t see it becoming more than poker chips at your exchange of choice.
- giantrobot 3y ago
  None of those things were invented or really developed by the "crypto sector" merely utilized by it. That's not a bad thing but it's disingenuous the crypto sector is really advancing any state of the art. It's implementing existing technologies mostly to facilitate scams and tax evasion.
  - mattdesl 3y ago
    All of these are primarily being funded and advanced by crypto. “Distributed ledger technology” hardly existed before blockchains/Bitcoin. Similar story with ZKP, which has gone from academic theory to real-world application (see: ZKSNARK) in the last decade primarily from blockchain research and funding streams.
    - graypegg 3y ago
      Aren’t zero knowledge proofs cryptographic signatures? I might be missing something, but that’s very much been in active use as part of any flavour RSA cryptography going back decades. Distributed ledger technology, though yes that is clearly a blockchain technology that wouldn’t exist otherwise. Fair point.

nologic01 3y ago

The comparison of incubation periods, adoption curves etc between crypto to ML cannot take us very far. The reason is that "ML" (as a proxy for algorithmic processing of a variety of data) is not particularly adversarial to pre-existing technologies. While this is not entirely true (there is always a hidden or explicit tension between automation and expert assessments) by-and-large this tension can be managed. The new tech gets bolted on the old. In contrast crypto sought to overthrow all pre-existing monetary and financial system technology with proposals that are naive, half-baked and ignore (or reinvent in unacknowledged manner) vital aspects. The practical implication is that crypto cannot carve a legitimate niche and keep iterating. The main outcome of a lot of wasted energy (in all senses) is to point out that indeed, digitization opens up the way to evolve the financial system. So much we knew, but now it has been drilled into the heads of large swaths of politicos, regulators, bankers etc that are completely tech illiterate.