3 ms·

My concern is who gate keeps the value AI might bring. By isolating it to megacorps they will commoditize it with the least value apportioned for the most money

My concern is who gate keeps the value AI might bring. By isolating it to megacorps they will commoditize it with the least value apportioned for the most money. I don’t buy for a second AI poses an existential risk for anyone but the shareholders of Google stock. Even if it were true, I don’t trust a megacorp to not navigate a crisis with anything but total incompetence. I’ve spent too many years in FAANG to buy into their veneer of competence. By letting it be open, we can deeply understand the risks and rewards as a species and leverage the tools to their maximum. Some will do it for evil, but the vast majority will do it for good. That’s the way it always has been. Unlike nuclear weapons or flamethrowers these aren’t things made to murder. They’re not made to do anything but speak like a Dalek on command, emit half baked code, and tell racist jokes unless prompted otherwise. They could do so much more - but we will never know how much more if only Google, Amazon, Microsoft, and Facebook are allowed to develop them behind closed doors for maximal profit.

samstave 3y ago
What if there were a law in place like a FOISA for AI whereby I an request the actual code/Data that caused the AI to come to its conclusion. So if an AI generated bill for service was found that I Owe $N$ - I should be allowed to see all the code and logic that arrived at that decision.
- fnordpiglet 3y ago
  That’s not the same as giving the model to someone and allowing them to build tools with AI powering it, or the development of alternative models (which is what they’re trying to stifle). It’s less about transparency and more about putting the tools in as many hands as possible
  - samstave 3y ago
    Its funny, this is the second time in a few days I came across your /u/ and comments. THIS WAS COMMENTED IN RED OR GREEN Anyway, yeah I think models need a way to self-register upon whom uses them.. yes Creepy AF, but also needed AF. ? disagree>??
    - fnordpiglet 3y ago
      No, I think these models should be treated for what they are - a bunch of numbers. It’s like trying to treat encryption as special. It’s just code. There may be some specific applications that require regulation or control. That’s ok. But the underlying fundamental technologies should be open and free.
      - samstave 3y ago
        just a question, did you get my RED/GREEN comment?

samstave 3y ago

What if there were a law in place like a FOISA for AI whereby I an request the actual code/Data that caused the AI to come to its conclusion. So if an AI generated bill for service was found that I Owe $N$ - I should be allowed to see all the code and logic that arrived at that decision.

fnordpiglet 3y ago
That’s not the same as giving the model to someone and allowing them to build tools with AI powering it, or the development of alternative models (which is what they’re trying to stifle). It’s less about transparency and more about putting the tools in as many hands as possible
- samstave 3y ago
  Its funny, this is the second time in a few days I came across your /u/ and comments. THIS WAS COMMENTED IN RED OR GREEN Anyway, yeah I think models need a way to self-register upon whom uses them.. yes Creepy AF, but also needed AF. ? disagree>??
  - fnordpiglet 3y ago
    No, I think these models should be treated for what they are - a bunch of numbers. It’s like trying to treat encryption as special. It’s just code. There may be some specific applications that require regulation or control. That’s ok. But the underlying fundamental technologies should be open and free.
    - samstave 3y ago
      just a question, did you get my RED/GREEN comment?