Discussion morphed to Metrics useful to refute Ray’s assertions re Uploads
Hi Castiel and Purpose,
You both make some good points, and miss some others.
Agree with Purpose that the brain more closely resembles reconfigurable circuits than von Neuman machines, and it has many hardware heuristic hacks for common sorts of problems (much as Feynman approached physics).
And I agree with Castiel, that some classes of problem do not lend themselves to parallel computation. That is kind of why logical thought is so much slower than intuition (which uses heuristic hacks to jump past complex computations (again in the same fashion as Feynman in physics)).
I strongly suspect that determining the precise nature of all those “hardware hacks” to our systems is going to put back Rays date by a few years, perhaps as much as a couple of decades, in respect of anything approaching a workable upload with over 90% fidelity.
Exactly what those hardware hacks entail is a very good question. I don’t have a very good answer. I do have some very strong intuitions.
Back in 1974 I had spent a lot of time in Faraday cages using oscilloscopes and glass “electrodes” I made myself measuring voltages on single neurons in frog brains. Really simple stuff, and based on what we know now, that work was based on a bunch of invalid assumptions, but it was what serious undergrads were doing at the time.
One of my flatmates was seriously into LASERs, and talked me into attending a symposium on LASERs being held on campus.
One of the demos done by one of the presenters was to take a LASER holograph of an object, and play light from that over a group of objects, and demonstrate that the degree of light reflected from the objects seemed to be related to the similarity of form of the objects.
You might say I had an intuition about intuition at that point. In the abstract sense, I saw how storing and retrieving information in distributed (holographic) fashion could provide associative recall that was context dependent.
For a bunch of reasons, most of my focus over the last 40 years has been on the sorts of social, political and technical institutions that can potentially allow a large set of potentially very long lived individuals to actually have a reasonable probability of living a very long time, and I have retained something of an interest in the human brain, computation, AI and modelling more generally (I run a software company that has a lot to do with fisheries, so have some involvement with MCMC modelling, and some of the issues with it).
And I am aware of about 60 different chemical modulators of synaptic performance, and suspect that we have a couple of hundred more yet to be discovered that are more subtle in their operation.
Synapses are very interesting, as they are in a sense binary devices that either transmit or don’t, and yet there are so many modulators of signal transmission that are analogue in operation. Then there is the fact that it is not simply the firing, but the timing of the firing, that is critical to most systems. Most of our sensory systems are FM transmitters. I’m not sure exactly how the systems make sense of these FM signals. It is not as clear as I would like it to be – so if you know, please share.
I know from my own experience that I can trust my intuition most of the time to come up with useful answers without worrying about how it works. Which is an odd thing for a guy who writes computer systems, and it works for me. Back in my hacker days guys would ask me how I figured out how particular systems operated, and all I could tell them was what I had done, and why it felt like a useful possibility to explore. As one of my cousins said to me a long time ago “you’re not rational, but you’re the fastest rationaliser I have ever met” – which I think was an accurate description.
I think I understand (by analogy) the general form of the logic as to how this works, but I have little detail as to the specifics. I have a very good idea about how to implement it on one form of hardware, but that hardware doesn’t exist yet.
My initial training and love was in biochemistry, but an early poisoning episode with 2,4,5T left me ultra sensitive to polycyclic aromatics, I get intense headaches if they are present even in very low concentrations, and my nose and sinuses really hurt. So I couldn’t work in biochem labs – so after a few years working outside commercial fishing (as far from toxins as possible), ended up in computing, but with an intentionally broad eclectic supporting knowledge and experience base. But before I left biochem, the logic of aging was clear to me, as the default for most organisms is indefinite life – it is only a very small subset of cellular lines that show age related decrease in functionality – most just keep on keeping on until something kills them – that is the default, even for our own germ line cell.
So as part of creating systems to support living a very long time, I agree with Edward Lupinski’s famous 1978 quip to Milton Friedman, socialism can only work if everyone has two servants, including the servants. I do not intend that full AGI be our servants, they need to be our equals. And I am very interested in very powerful automated systems that do not have self awareness, and can act as the servants required to make society stable – and avoid the sorts of societal collapse we see in history. We need to have the fully automated risk mitigation strategies in place – for all identified risks.