4 ms·

that would be amazing if this could lead us there, and also work in episodic memories and time decay of importance as well as assigning sensory data or semantic

that would be amazing if this could lead us there, and also work in episodic memories and time decay of importance as well as assigning sensory data or semantic emotional data to gauge the sentiment or 'feeling ' such a memory could or should elicit. I say it would be amazing, it'd also be terrifying. The moment you tread past this cliff its going to change our entire existence, for good or bad. I'm anxiously enthusiastically nervous and scared shitless.

mdp2021 3y ago
We do need intelligence enhancing tools, increasingly.
- ChatGTP 3y ago
  Is this the story one needs to themselves to enthusiastic about AGI ? I really don’t understand the enthusiasm for it but I wish I did. I guess if I create this “oh my god we are doomed and the only way out is through AI domination” I can kind of get excited about it? Like I can invent a crisis that needs solving.
  - mdp2021 3y ago
    > needs [and] I wish I did There is no "«need»", if that means "hiding an excuse". But if you want to go "lymbic", as per a parallel post nearby: Intelligence is a Value. It is a fact: individuals and societies are increasingly in need for Intelligence. When Vannevar Bush conceived the Memex, the need was chiefly for making information more available. Now the overwhelming need is to process it. In practical terms, it is quite strong how somebody formulated the point in the recent Kissinger's book "Leadership": -- the world is getting more and more complicated, while administrations are getting less and less capable where "administrations" there reflect individuals in general. But computers, now largely in pockets, finally dramatically available, were supposed to be the Memex, enhancing tools... The promise must be fulfilled, because the stakes remain high (societally, historically). -- Update: > through AI domination [...] invent a crisis This should be a joke. This seems like living on idle fantasies while there is a real world outside them, that will not run according to wishes just through its own natural running. There are decisions to be made - by humans, by individuals, by administrations, by governments. There are real problems out there: drought, floods, diplomacy, resource management ... what to do, who to marry. There are optimization and decision problems that we cannot just solve through paper (consuming pens and eyes). We have been studying AI for decades and we have been using aided decision making tools for centuries - automated, for decades. You are using one now, the browser is a DSS information accessor! That means, calculators, not "Overlords"!
    - ChatGTP 3y ago
      Yup, exactly. So through technology we’ve managed to created situations we cannot manage so the answer is more technology we likely will have zero control over or understand. Can you see see how ridiculous this can sound to some people ? We’ve created a ridiculous situation so I guess you’re right, the only way through it is more ridiculousness. In my opinion that story you’re telling me actually does create a perceived need for AGI. Edit: It sounds like in your world, this is some type of AI without agency or self preservation goals ? Just like an oracle with an interface or something ? I hate to say this but governments have all the information they need to make better decisions, they make them mostly based on lobbying. Not going to fun when AI can extract more oil at 1000x the speed, crush dissidents or capture your attention to sell you more shit even more effectively.
      - mdp2021 3y ago
        And you would think that remaining without tools could be better? Like in Rousseau's ideal? Such as "It is the fault of the people of Lisbon if they were crushed in the earthquake: they should have never left the forest, no buildings to collapse on them"? Rousseau did not seem to notice that forests can be effing damp and cold. If tools started with the Homo Habilies well over two million years ago, it meant that inventing butchery was advantageous. Even "lymbically", as they found it preferable - surely emotionally preferable - than hunger, waste etc. We have been doing AI for a long time: and we quite normally meant "virtual professional". ("Computer" means "virtual computer [(the job)]"; "AI" stands for "virtual computer [(the job)] with less mechanical, routinary tasks".) This "«agency or self preservation goals»" idea is practically not impossible but an "holliwoodian" drift of the debate - entertaining thoughts replacing practical ones. If people act badly because of lack of wisdom, because of the hiatus between technological advance and personal advance (as we have been noticing at least for the past three centuries): they shall cultivate themselves. We do. Ignorance is societally the most massive problem, but it is a different problem. (Edit: sorry, not just and precisely "different": it is a related problem, because in fact we use Intelligence to acquire Wisdom. Intelligence is that thing that dissipates ignorance... That makes it a value. Synthetic, real Intelligence is meant to help against the devastating ignorance around.) That people fall from ladders or use them to hit people in the streets is not a sufficient note to suggest we should do without ladders - and lose the inherent perks.
mdp2021 3y ago
> assigning sensory data or semantic emotional data to gauge the sentiment or 'feeling' such a memory could or should elicit This may not be necessary. It is pretty fundamental in humans, because factual, ethological decision making apparently needs an emotional drive to actually act on a decision, irregardless of the depth of reasoning that ranks optimality of options (this was found by Damasio, I believe). Those who can reason quite well, but suffer from a disconnection to the lymbic system - so they miss a "trigger" "shaking" them in front of the different options - cannot practically decide. But for a Decision Support System, what the operator needs is a lucid, informed consultant to query about the options. The decision remains with the operator. Machines can act out of a function. There is no necessary need in a machine to divide "quick synthetic assessment" and "explicit analytical assessment", if in theory the machine has enough computation abilities to stick to the latter, and if the latter is what we want - as we want reliability. One thing is the implementation of Intelligence, another is emulation (of "Also Potentially Intelligent Systems").

mdp2021 3y ago

We do need intelligence enhancing tools, increasingly.

ChatGTP 3y ago
Is this the story one needs to themselves to enthusiastic about AGI ? I really don’t understand the enthusiasm for it but I wish I did. I guess if I create this “oh my god we are doomed and the only way out is through AI domination” I can kind of get excited about it? Like I can invent a crisis that needs solving.
- mdp2021 3y ago
  > needs [and] I wish I did There is no "«need»", if that means "hiding an excuse". But if you want to go "lymbic", as per a parallel post nearby: Intelligence is a Value. It is a fact: individuals and societies are increasingly in need for Intelligence. When Vannevar Bush conceived the Memex, the need was chiefly for making information more available. Now the overwhelming need is to process it. In practical terms, it is quite strong how somebody formulated the point in the recent Kissinger's book "Leadership": -- the world is getting more and more complicated, while administrations are getting less and less capable where "administrations" there reflect individuals in general. But computers, now largely in pockets, finally dramatically available, were supposed to be the Memex, enhancing tools... The promise must be fulfilled, because the stakes remain high (societally, historically). -- Update: > through AI domination [...] invent a crisis This should be a joke. This seems like living on idle fantasies while there is a real world outside them, that will not run according to wishes just through its own natural running. There are decisions to be made - by humans, by individuals, by administrations, by governments. There are real problems out there: drought, floods, diplomacy, resource management ... what to do, who to marry. There are optimization and decision problems that we cannot just solve through paper (consuming pens and eyes). We have been studying AI for decades and we have been using aided decision making tools for centuries - automated, for decades. You are using one now, the browser is a DSS information accessor! That means, calculators, not "Overlords"!
  - ChatGTP 3y ago
    Yup, exactly. So through technology we’ve managed to created situations we cannot manage so the answer is more technology we likely will have zero control over or understand. Can you see see how ridiculous this can sound to some people ? We’ve created a ridiculous situation so I guess you’re right, the only way through it is more ridiculousness. In my opinion that story you’re telling me actually does create a perceived need for AGI. Edit: It sounds like in your world, this is some type of AI without agency or self preservation goals ? Just like an oracle with an interface or something ? I hate to say this but governments have all the information they need to make better decisions, they make them mostly based on lobbying. Not going to fun when AI can extract more oil at 1000x the speed, crush dissidents or capture your attention to sell you more shit even more effectively.
    - mdp2021 3y ago
      And you would think that remaining without tools could be better? Like in Rousseau's ideal? Such as "It is the fault of the people of Lisbon if they were crushed in the earthquake: they should have never left the forest, no buildings to collapse on them"? Rousseau did not seem to notice that forests can be effing damp and cold. If tools started with the Homo Habilies well over two million years ago, it meant that inventing butchery was advantageous. Even "lymbically", as they found it preferable - surely emotionally preferable - than hunger, waste etc. We have been doing AI for a long time: and we quite normally meant "virtual professional". ("Computer" means "virtual computer [(the job)]"; "AI" stands for "virtual computer [(the job)] with less mechanical, routinary tasks".) This "«agency or self preservation goals»" idea is practically not impossible but an "holliwoodian" drift of the debate - entertaining thoughts replacing practical ones. If people act badly because of lack of wisdom, because of the hiatus between technological advance and personal advance (as we have been noticing at least for the past three centuries): they shall cultivate themselves. We do. Ignorance is societally the most massive problem, but it is a different problem. (Edit: sorry, not just and precisely "different": it is a related problem, because in fact we use Intelligence to acquire Wisdom. Intelligence is that thing that dissipates ignorance... That makes it a value. Synthetic, real Intelligence is meant to help against the devastating ignorance around.) That people fall from ladders or use them to hit people in the streets is not a sufficient note to suggest we should do without ladders - and lose the inherent perks.

mdp2021 3y ago

> assigning sensory data or semantic emotional data to gauge the sentiment or 'feeling' such a memory could or should elicit This may not be necessary. It is pretty fundamental in humans, because factual, ethological decision making apparently needs an emotional drive to actually act on a decision, irregardless of the depth of reasoning that ranks optimality of options (this was found by Damasio, I believe). Those who can reason quite well, but suffer from a disconnection to the lymbic system - so they miss a "trigger" "shaking" them in front of the different options - cannot practically decide. But for a Decision Support System, what the operator needs is a lucid, informed consultant to query about the options. The decision remains with the operator. Machines can act out of a function. There is no necessary need in a machine to divide "quick synthetic assessment" and "explicit analytical assessment", if in theory the machine has enough computation abilities to stick to the latter, and if the latter is what we want - as we want reliability. One thing is the implementation of Intelligence, another is emulation (of "Also Potentially Intelligent Systems").