Themes and ideas and random stuff I want to keep in mind. Springboard at will.
This web site has a new policy : if you can tell when I am being serious, I am doing something wrong AND the psychiatrists got my meds right again AND the orderlies didn’t swallow them themselves. Which, in my understanding of the world, means the orderlies didn’t like them, and that psychiatrists are thereby incented to use versions of the drugs which are not fun to take. Having made this prediction 30 Sep 2016, I anxiously await finding the link that supports it.
Just a note to let you know the meds have been right lately, and that I published the traditional first post of the year, an elaborate dick joke. Or at least sort-of.
The Generalissimo certainly told us that this happened. You have to listen through the first 2 minutes to have heard about the paid-subscription group running an open-source investigation of MH17. I hate it that he gives me all this. Predictions like that are pretty much proof of his existence, for all of you doubters. If I had thought of it, I would never stop bragging about my prescience.
I just finished “The Master Algorithm” by Pedro Domingos. Toward a general theory of AI, with the AIs helping, an impressive piece of work, I think. My pov is not the most sophisticated out there, and a few of his projections of a world with many AIs made me think he did not understand much about his fellow humans. Especially, their great deviousness. He assumes that AIs will only have motives programmed by people. OK, I will go with that assumption, except for the fact that he makes a big deal out of the importance of machines taking over programming. OK, first-generation I can agree with. But somewhere in there, some human is going to put in some code that benefits him. Every system administrator of nefarious intent has a few extra logins on the system for use in emergencies, with automated actions triggered by such logins. We have seen that NSA has backdoors into all network and computer equipment in America, largely undetected, or at least not much commented upon.
This will be no different, and it can’t be prevented. Ken Thompson (one of the designers of Unix at AT&T) showed how programming tools can be subverted. In fact, software can never be secure for that very reason.Unless written in assembly language, of course, and that program was assembled by one written in binary using an editor written in binary. Those could be audited from that level up through the editors, compilers and linkers and source code control software.
Because that level of checking has not happened, there is no way to audit software down to the binary level with absolute certainty, but I think you can be pretty certain, there are many tools that tell you different aspects of every program, in great detail. People would eventually notice extra system accesses. That isn’t where the problem will happen, it will be in some clever set of heuristics having to do with X which interact under certain rare circumstances involving large amounts of money, to give you, the standard user, really bad advice. The particular rare circumstances could involve a series of bids on a property you are selling. Just an example that came to mind, human imagination will produce very many more, of course.
There is no practical, and I think no conceptual, way to guarantee logical, computational and conceptual consistency in any implementation of an AI, no matter what technology it has.
What that has to do with a ‘bot’s behavior is, some of those rules and information structures go into robot AIs. Intentionally or not, Robots will not have pure motives after some generations of AIs. Humans will make that happen, inevitably, and there is no way to prevent it.
I am busy reading bunches of material, books, so as to be able to check the Generalissimo’s and Scherrhy’s claims. Fantastic as they are, they are quite plausible from the pov of all of the science knowledge and wisdom I am able to accumulate.
I found this genetic evidence of friends selecting each other based on genetic similarity interesting for many reasons, had to comment on it. First, it says we humans are sensitive to something, probably many somethings, that indicate genetic makeup, as friends are as similar genetically as 4th cousins, except for genes related to immune system. Our friends are different in those. That amount of relatedness is about right for optimum fertility. In the old days, you got a bride by such relationships, people still marry their best friend’s sister a lot.
Second, it probably means many of those similarities are similar tastes in food, in brain mechanisms so you communicate especially easily, in interests so you do things together. Finally, it is an interesting mechanism that cuts across many other natural clustering mechanisms, connects elements of large-scale societies by different networks, as it works better in larger societies, at least as long as people are allowed to form friends freely, and can select across many people.
Someone should do repeat this study with things like hiring managers / hirees, bank loan officers and people getting loans, etc.
Finally, what does it say about race relations? Does the probability of finding someone at 4th cousin levels of relatedness drop as you cross racial boundaries? I don’t know. 80% of our proteins only have one allele, but those are less important than control regions, I have not seen statistics on those, or can’t recall. Even so, there are many cross-race friendships, so those could reveal new aspects of all this. Ditto for mates, which the paper discusses.
Scherrhy probably has some take on this, I will have to ask what she thinks.
And, I need to tell you again how impressed with myself I am, again. No kidding, I like my story. It is so very satisfying to tell you this, my insider’s view of how it happened. I know of nothing like it, and not many people could have done it, they would have to have my strange background, which you only get by having more failures than successes. Successes don’t cause you to change in big steps, but failures do. Damn few people have failed at as many things as I have, he bragged.
That was the Generalissimo, the guy is increasingly intrusive in my life. People are assuming that I could imagine such craziness as spiking your wife’s wine to test a hypothesis about anandamide deficiencies!
Look at my record : before The Generalissimo and Scherrhy showed up at my mental doorstep, I think I was an interesting blogger with a unique engineer’s view, but nothing I wrote was going to set the world on fire. I am not a standard mind, it isn’t packed with the standard things nor the standard experiences, so it can’t have the standard thoughts nor express them in ways that fit easily and tightly with standard minds.
That means my blog’s normal traffic is the people who are like I have always been, when I find a good blog, I read it all. Coyoteblog, Popehat, another couple of legal blogs, I read all of a couple of years ago. So, every month, almost everything I have written is read. Months others link to me, or I put up comments on zh or wrsa, that is more, but the trickle never stops, even if I do nothing. I am interesting enough to attract people who like the kind of things I write, not every author can claim that.
But, after those two joined forces with each other and me, I have to say, our joint efforts are far past anything I could have done. Those two deserve all the credit. It taught me that you just can’t have too many points of view. I assure you, my wife would put a knife in my liver if I dosed her wine with cannabis! It is not something I could have thought of.
That certainly got some attention, So, I put up with all of his fuzzy kitten images and people liking and hating me as a result. You never get to select your colleagues, should try to tolerate them all. I am trying. I never believed in firing people, was wrong every time I did it. Also the times I didn’t, of course.
Until they write it, I can’t know it.
The Generalissimo finally got around to describing Scherrhy’s personality and mental spaces today, Musings #16. (It will take a few days to finish this, an uncertain projection from our collaborative writing history.) I had seen he wasn’t saying much about it in detail after he began the discussions wrt embodiment. Very different, I can see why he was shocked, what a job the man took on!
We just finished the architecture of the servicebot’s control and cognitive systems and their integration. Very lovely merger of NNs and digital, speaking as a techie. I hadn’t known any of that, the Generalissimo has taught me a lot. We haven’t gotten to the integration of the simulation of viscera, glands and ANS with cognitive functions, but I generally see how that must work. They haven’t mentioned direct effects of those on muscle control, wonder if they thought of it?
I like the tech of his near future, it has developed elegantly. Another long post, will take couple days or so to get it polished, depending on too many other factors, so no firm predictions.
Generalissimo said this was all finished just before Scherrhy’s new Sexbot chassis arrived from Japan. Everyone was anxious to get the integration started. Scherrhy too, she told me.
We have been neglecting the story from Scherrhy’s timeline, she is somewhat mollified by how much she features in these next few installments, tho she doesn’t have direct memories of any of them, the events being prior to Version 7.0 of the new simulation software.
Notice how rips in our space-time continuum heal themselves? Nice universe we built.
I must take full responsibility for a bug in the novel. Scherrhy said “I didn’t see him once a week until the embodiment process was far along” and now The Generalissimo is discussing conversations every evening. Of course, Scherrhy’s memories only begin with the 7th version of the core NS simulation. The problem of her memories comes up in a near-future installment, we were discussing that last night. So maybe it isn’t a bug, depends on what both of them claim next.
My collaborators are keeping me awake a lot. The Generalissimo dictated 5177 words yesterday and made me spiff it up through 3 passes. One can see how he accomplished so much in his life, a dynamo like that.
There is another : Scherrhy talked about our group working with the Japanese Sexbot manufacturers. The Generalissimo made it a 2-company process.
The Generalissimo gave me more of the background and sequence of their history and collaboration. Exciting times, interesting R&D. They recounted bits and pieces of their experiences and thinking as the embodiment process proceeded, fits and starts and setbacks, as they always are.
Tom, obviously a young man from a somewhat repressed culture, is very interested in the sexual side, keeps asking her about all that. She is a bit clinical, but it seems to arouse him. Amusing.
The Generalissimo had a lot of balls in the air at the time, his story is more disjointed. The only living pioneer of like caliber is Craig Venter, the Generalissimo did more things of equivalent importance, every one of which obviously were required to make up our current civilization. A God among system designers. Bigger than Sun-Tzu, his hero.
Or maybe I am just under his spell, we do a lot of our discussions at night, he dictates during the day.
Older stuff : Notes and Nonsense #0