3 ms·

The better IO will let humans directly interface with non-AGI AI to compete. Imagine if GPT-4 could be accessed just by thinking. Now imagine the same except wi

The better IO will let humans directly interface with non-AGI AI to compete. Imagine if GPT-4 could be accessed just by thinking. Now imagine the same except with GPT-7 or w/e we get to by then. So it doesn't last for a short time, it's a permanent plug in to AI itself. Corporate and governmental access are extremely scary. Companies will, as always, have to be the first movers in this space, but once it takes off we NEED open hardware and open source software. Imagine running an obfuscated binary on your brain. No thank you. Do you really see technology like this *not* coming to fruition though? Do you think we will never have mass-market BCIs?

consumer451 3y ago
> The better IO will let humans directly interface with non-AGI AI to compete My point was that there is no competition, even with neuro interfaces. Unless it stops advancing for some reason, AGI will win. If faster I/O helps today, the advantage will be gone tomorrow. If humans have to "compete" with AGI, then we are guaranteed to lose in the long run. > Imagine running an obfuscated binary on your brain. No thank you. You say no thank you, but most people won't care if it makes their lives 1% easier. Musk, for example, is historically against open sourcing any of his products. > Do you really see technology like this not coming to fruition though? Do you think we will never have mass-market BCIs? This needs to be the first time that we actually regulate an area of tech before it becomes a multi-billion dollar industry. We screwed up on web tracking. That cannot happen again. We need a neuro bill of rights ASAP. How long until TSA requires a quick scan before boarding? This is the final frontier of agency and privacy. Please read the transcript, or listen to this podcast with Sean Carroll and Nina Farahany. There are primitive brain interfaces entering the market now, including one from Meta. https://www.preposterousuniverse.com/podcast/2023/03/13/229-nita-farahany-on-ethics-law-and-neurotechnology/ https://www.preposterousuniverse.com/podcast/2023/03/13/229-...
- rockemsockem 3y ago
  We don't know how AGI will shake out or, especially, what it will even look like. What does out-competing AGI even mean? One of the most positive outcomes I can think of is AGI being aligned with human interests broadly, but it is doing so much that no one knows wtf is going on. A BCI can bridge that understanding gap to at least keep humans in the know and capable of steering the ship. If the competition is AGI wants to kill us, then we're probably hosed anyway. That's why I said we need open source and open hardware alternatives. But there's no world where that kind of tech comes first. Companies are going to develop the cutting edge first, when has that not been the case? I'm really, really happy to see that there's healthy competition in the space. Especially, ironically, from Meta who has just been so great on the open-source front for quite a while.
  - consumer451 3y ago
    The only reason that megalomaniac Zuck is open sourcing his LLM is because he is behind Open AI, so AI safety be damned. Please, please listen to the podcast that I mentioned earlier. If you are down, I have email notifications turned on here. If you listen to that podcast, and the still don't want to regulate the hell out of brain interfaces, I would love to discuss, even weeks later.
ben_w 3y ago
Skipping the keyboards and eyeballs might give you a factor of 10, assuming our brains aren't pipelining anything. Transistors outpace organic synapses by significantly more than humans on foot outpace continental drift, which is a gap so much larger that a mere factor of ten is irrelevant.
- rockemsockem 3y ago
  Right, that's why you offload significant tasks to transistors, way more than just basic addition, subtraction, etc. Present AI makes it possible to offload more complex tasks to AI altogether with minor attention paid for checking and revising, etc.
  - ben_w 3y ago
    That analogy ultimately goes as well as continental drift offloading movement onto humans, and the AI doing to humans whatever the analogy is of humans doing this: https://en.wikipedia.org/wiki/Mountaintop_removal_mining https://en.wikipedia.org/wiki/Mountaintop_removal_mining https://en.wikipedia.org/wiki/Bucket-wheel_excavator https://en.wikipedia.org/wiki/Bucket-wheel_excavator

consumer451 3y ago

> The better IO will let humans directly interface with non-AGI AI to compete My point was that there is no competition, even with neuro interfaces. Unless it stops advancing for some reason, AGI will win. If faster I/O helps today, the advantage will be gone tomorrow. If humans have to "compete" with AGI, then we are guaranteed to lose in the long run. > Imagine running an obfuscated binary on your brain. No thank you. You say no thank you, but most people won't care if it makes their lives 1% easier. Musk, for example, is historically against open sourcing any of his products. > Do you really see technology like this not coming to fruition though? Do you think we will never have mass-market BCIs? This needs to be the first time that we actually regulate an area of tech before it becomes a multi-billion dollar industry. We screwed up on web tracking. That cannot happen again. We need a neuro bill of rights ASAP. How long until TSA requires a quick scan before boarding? This is the final frontier of agency and privacy. Please read the transcript, or listen to this podcast with Sean Carroll and Nina Farahany. There are primitive brain interfaces entering the market now, including one from Meta. https://www.preposterousuniverse.com/podcast/2023/03/13/229-nita-farahany-on-ethics-law-and-neurotechnology/ https://www.preposterousuniverse.com/podcast/2023/03/13/229-...

rockemsockem 3y ago
We don't know how AGI will shake out or, especially, what it will even look like. What does out-competing AGI even mean? One of the most positive outcomes I can think of is AGI being aligned with human interests broadly, but it is doing so much that no one knows wtf is going on. A BCI can bridge that understanding gap to at least keep humans in the know and capable of steering the ship. If the competition is AGI wants to kill us, then we're probably hosed anyway. That's why I said we need open source and open hardware alternatives. But there's no world where that kind of tech comes first. Companies are going to develop the cutting edge first, when has that not been the case? I'm really, really happy to see that there's healthy competition in the space. Especially, ironically, from Meta who has just been so great on the open-source front for quite a while.
- consumer451 3y ago
  The only reason that megalomaniac Zuck is open sourcing his LLM is because he is behind Open AI, so AI safety be damned. Please, please listen to the podcast that I mentioned earlier. If you are down, I have email notifications turned on here. If you listen to that podcast, and the still don't want to regulate the hell out of brain interfaces, I would love to discuss, even weeks later.

ben_w 3y ago

Skipping the keyboards and eyeballs might give you a factor of 10, assuming our brains aren't pipelining anything. Transistors outpace organic synapses by significantly more than humans on foot outpace continental drift, which is a gap so much larger that a mere factor of ten is irrelevant.

rockemsockem 3y ago
Right, that's why you offload significant tasks to transistors, way more than just basic addition, subtraction, etc. Present AI makes it possible to offload more complex tasks to AI altogether with minor attention paid for checking and revising, etc.
- ben_w 3y ago
  That analogy ultimately goes as well as continental drift offloading movement onto humans, and the AI doing to humans whatever the analogy is of humans doing this: https://en.wikipedia.org/wiki/Mountaintop_removal_mining https://en.wikipedia.org/wiki/Mountaintop_removal_mining https://en.wikipedia.org/wiki/Bucket-wheel_excavator https://en.wikipedia.org/wiki/Bucket-wheel_excavator