Sometimes I despair at the continuing naïveté of scientists. From those who think we’re going to use new weapons responsibly to those who trust that grant money will be given out on the basis of which research holds the most promise to benefit the earth or humanity, researchers in any field can be startlingly ignorant of how their work will ultimately be used.
Ah, Headmistress of Hussies, Mentor of Molls, and Didact of Doxies! You are truly the best of bimbos, the fairest of flirts and floozies. Knowing your eternal struggle in maintaining an incoming supply of vibrators appropriate to your outflow, I wished to make sure that this advance in sex toy technology was placed in just the right spot to get your attention:
FtB generally and Pharyngula in particular had a longtime commenter nymmed What A Maroon. (Miss you! Haven’t seen you! Hope you’re well!) This, of course, comes from the classic Bugs Bunny insult, “What a maroon!” which I have always assumed to be a child-friendly corruption of “What a moron!” but whose etymology I had not examined.
Well, I still have not examined that etymology, but Janelle Shane has a new post up that does dive into the nature of human maroons and other colors. As always, her methodology is to feed data of a certain type into an AI and then see what the AI can create along the same lines. In this case, it’s an experiment in color/personality comparisons that have been popular memes for some time. Shane tells us that
It seems that the communications security protocol flaws and lack of physical override/bypass or escape mechanisms we build into male chastity devices have been discovered. Look at how much the men now know:
The Cellmate, [an internet-connected male chastity device] produced by Chinese firm Qiui, is a cover that clamps on the base of the male genitals with a hardened steel ring, and does not have a physical key or manual override.
A security flaw in [the device] could allow hackers to remotely lock it — leaving users trapped, researchers have warned. … “An angle grinder or other suitable heavy tool would be required to cut the wearer free,” [British security firm Pen Test Partners said Tuesday].
The MRAs do not yet know it was engineered this way by scientists at NOW Labs. We must keep that information contained at all costs. In the meantime, get as many men fitted for Cellmates as soon as possible before masculine distrust sets in. Further updates on the secret channel.
You may already be familiar with the very fun person whose writing includes the book, You Look Like a Thing and I Love You. She tests AI networks by training them on large samples of certain kinds of writing, then takes a look at the output when they try to reproduce such things. In the olden days of 3 or 4 years ago, the “writing” was extremely short. AI couldn’t manage much in the way of good grammar, so things spun off the rails very quickly if you asked for even 5 words in a row. As a result, much of the fun she originally shared on her blog consisted of things like AI-generated names for hypothetical metal bands. Later there were song titles. Still later things got longer, although that worked best when working with very predictable data sets, like AI-generated recipes.
But as time has gone on, the AI networks with which Shane experiments have gotten more capable, and she has finally decided to go full meta. She asked an AI network to generate a blog post for her own blog by training it on her past blog posts. The very first one was quite sharp:
Welcome to AI Weirdness! In today’s blog post, I wondered what would happen if I trained a neural net to generate numbers in the range [-1,1].
As you might expect, the neural net started generating numbers that were more than a little odd: 0.9286401128496419 0.06491227721694329 0.9879906435756054 0.745677946535383 0.679450170988766 0.1065339560745793 -0.2977483969426606 0.05673549481890381 0.4304206725753945 0.7544217043730188 -0.8960721608636865 0.7224255864247317 -0.2139997613483644 0.0295771285597738 0.619233475195753 0.05578493537857566 -0.630712498931699 -0.0506476255464803 0.3256973044703828 0.0988207888549295
Isn’t that one awesome? I mean, when I saw “-0.6307…” I just knew it was going to be hysterical, and then it ended “…31699” and I just completely lost it. OMG 2 FUN-E!
Yet now I am left with the tragic notion that in the future, I will never know whether a post on Shane’s blog was written by her with a bit of AI output placed where she consciously chose to put it, or written by the AI from scratch. I have lost Shane to meta-space. Make sure you go and get your last look at her blog before Skynet takes over.
One or both of my readers my remember my long time fondness for the work of Dr. Janelle Shane, an optical physicist who delights in exploring the capabilities and limitations of neural networks – computer systems that attempt to identify what certain words or images have in common and then generate novel members of the inferred set. While she has used neural networks for many delightful things, this week she has literally used it to release the Kraken!