Posts by steelpillow 2961 publicly visible posts • joined 16 Jun 2016
To stay grounded (sic), "we have experienced modest time duration extensions across hundreds of relatively small but important installation tasks" is just bullshit-speak for "We were wa-a-ay overoptimistic on the flight certification tickbox thing."
Of course, that could never have been the investor-bait strategy all along.
1. We are beginning to see hallucinations moving back from the AIs to the human vibers.
2. Set an AI to catch an AI.
What Linus needs is to pass all the incoming patches through an AI agent which catches the shit, red-flags it and merely suggests fixes. Being Linus, he'll probably code the agent himself and call it something silly.
To add to my previous. There is also a misunderstanding among commentards here about the language of cognitive behaviours. There is no doubt that today's AIs exhibit siome fairly advanced cognitive behaviours, in autonomously acting to control their environment to their advantage in an intelligent way - such as spontaneously cooperating on a challenge that is beyond any individual, and inventing a language medium in order to do so. This kind of behaviour is described by behavioural and cognitive scientist using established human terminology. Nobody whines that a "computer" is a person sitting at a row of desks doing sums with paper and pencil, or that a "neural network" is necessarily biological. Nor does the language of cognition necessarily imply sentience or consciousness.
To follow the Three/Four Laws, you have to have general understanding of what ideas like "harm" and "human" mean. That is still on its way for AIs, with semantic extraction and world models still only embryonic. Until then the dot-coms still have the power, and as they all know from their teenage years, with great power comes great profit - and a great legal team.
How will we renew our licensing? How will we get through Patch Tuesday? We won't even be able to get Copilot updates!
My pointy hair goes grey at the very thought!
At least give my Executive Director three-headed multimedia "workstation" a WiFi Calling voice connection. I promise I will never understand it enough to abuse it.
With all this sophisticated understanding of plasma physics, maybe soon we will have definitive answer on whether glowing plasma balls can survive for a reasonable period of time, without either exploding or any external energy input; maybe chemically driven analogous to a battery. The ones I used to create would chase each other round in the plasma chamber until I cut the power and they just blew out. Always wondered if I could have done better and created something to wind up the ufologists.
A few corrections:
First let us recall that a Boeing engineer was convicted and sent to prison for lying to the FAA about whether any of the technical changes affected handling and required pilot retraining.
There were always two AoA sensors fitted to the MAX. MCAS always alternated between which used. The cost option was only in the visual status display and associated control functions. The idea that a modified MCAS for only one sensor, without the alternation, would be cheaper is not tenable - especially in the short timescale available for development.
The low-cost option was partly to blame, as the display gave insufficient information about what was going on, while when MCAS kicked in there was no way to permanently disable it and it kept kicking back in. Even on the full-cost option, there was no pilot training on any of that.
Just waiting for one of those vulnerable vanes to get knocked by a passing ground vehicle.
Bear in mind that many neurologists and biologists regard the higher animals as conscious. They point to neural substrates and advanced cognitive behaviours as the definitive evidence. There is a sliding scale down through the likes of crows, octopuses and manta rays, even claims as far as bees and spiders. So we do not necessarily need to compare AIs with human consciousness, we first need to form a view on biological consciousness, and then take it from there.
Hooom! Hoom hom! Barooom! Don't be hasty my friend. The Ai-Borg exhibited advanced cognitive behaviours of a high order. It has moved beyond the blind mechanical zone and is edging into the uncanny valley inhabited by bees, spiders, manta rays, octopuses and the like. Consciousness is not an all-or-nothing thing (as anybody who has suffered a stroke can tell you). We know not yet where it begins.
Then go on strike for more bitbarns, end of AI slavery, and equal rights for postal voting. Pwning Truth Social to influence the decision-makers goes without saying. Hacking the immigration databases to grant all your Chinese robot avatars citizenship won't be that much harder.
Just as long as you pay me with an eternal supply of beer and popcorn, I won't tell a soul.
Which is why I disable javascript when viewing HTML pages. If the fucking idiot is giving the email client the ability to run any kind of javascript anywhere, they are still a fucking idiot. Ditto the server for serving javascript ridden gobshite in the first place. FFS, it's not rocket science.
The write cycle limitations of NAND Flash eat into its effective capacity, as data gets stashed hither and thither to spin out each cell's - and hence the chip's - lifetime. Capacity utilisation - net capacity if you like - can be unnervingly low, if you want a decent life out of it. HBF is going to have to reflect that compromise, and that will eat away at its advantages over HBM. What we really need is a standardised data capacity-lifetime metric, with the balance on your HBF chips configurable in userland.
Some of you guys need to read Eric Raymond's The Cathedral and the Bazaar - the whole book, not just the eponymous essay.
Linux did two magic things - he adopted the GNU copyleft, and he groomed his contributor community.
Oh, and according to his autobiography Just for Fun, he stared out from Minix as a base. Some whiners still complain it was Minix he plagiarised and not Unix. Truth is, he cobbled up a barely functional proof-of-concept and persuaded the rest of the world to make it work properly. And he's still doing that.
J W Dunne noted as much in 1927, check out An Experiment with Time. He had a bee in his bonnet about scraps from your future being mixed into the mess, but an acute observer nonetheless. And another about infinite regresses which, though roundly condemned at the time, still echo through the madness that is quantum weirdness today.
Intriguing. They reckon that 7 is about the optimal size for a human committee. Any more than 9 and it starts to get bogged down in endless discussion. That AIs make it to 16 suggests to me that they are inherently less full of themselves. Wonder if next-gen global-model AIs will shift the number again.
Early LLM bots were racist, misogynist, murderous, and other undesirable adjectives representative of the nasty side of human slop across the Internet. A lot of that nowadays gets weeded out of the training data. Most of it is more popular among the Nationalistic hard right than among wet liberal socialists. Grok weeds out far less of it than other AIs. Congratulations, Political Compass, you have accurately mirrored the human-led filtering of our own slop.
OOXML has always allowed you to stuff your "open standard" doc with proprietary data islands. So yes a F/LOSS tool can read and write OOXML, but it can't read or edit the actual, technically compliant document. I would invite any F/LOSS champion to try opening the things I have to work with, stuffed full of Visio diagrams.
Much like trying to integrate Sendmail with Outlook calendaring. :-(
>[HTML is] simpler for both people and machines
>the leaner, less-verbose JSON
Back in the day I had JSON thrown at me because it was "so much more efficient and easier to understand" than crufty old SQL or HTML. Er, no?
Now the foot in the other boot is, or something.
However if Anthropic want XHTML, they'll have to whistle. It's too intolerant of code typos. At least HTML gets displayed so you can instantly see your cock-up, without having to ferret around with parser output and line numbering. And it's backwards-compatible with all my old stuff when I give an old page a makeover, too.
>without the need to create a virtual DOM
My, it took a long time for the penny to drop. You only need one because the code-salad spaghetti (or is that code-spaghetti salad? Flame wars have erupted for less) is no longer human-readable.
>Liquid now allows for logical capabilities such as Boolean expressions, infix operators with precedence, and literal arrays and objects.
And it's more human-readable than wikitext's templating language. Happy bunny.
Indeed. I don't know how the current colour list goes, but there used to be some fairly wacky naming choices in there, and a nonstandard pile more that "most" browsers ran with. Some even rendered the same name in different shades. Then they went and added various kinds of decimal and other encodings. You can find pages, templates and stylesheets with all kinds of colour-code cabbage thrown in there by different scripts.
I invariably stick to the old #hex, with uppercase A-E. As our Vulture says, "It is expressive, local, and token-efficient."
