HNHacker News
TopNewBestAskShowJobs

_wire_

299 karma · joined April 4, 2024

submissionscomments
_wire_··on Ironies of Automation (1983) [pdf]
This article has a lovely premise: that as AI continues its incursion, its very prominent failures, are being institutionalized into a lower class of humans who have to take the blame for all the crazy ways AI screws up.

AI is a failure magnifier, which contrary the most basic idea of intelligence, is incapable of taking responsibility for its mistakes.

"Oh, the AI fucked up again? Who is responsible for this?!"

Looking forward to day when all c-suite has left are AI's that lie and shirk with complete impunity, because how can a computer program be held to any account?!

_wire_··on One Battle After Another Review
Previously posted to HN...

One Battle After Another Review by Mark Cira

https://letterboxd.com/markcira/film/one-battle-after-anothe...

This review is literate, historically aware, personal and very insightful; from the heart of a cinephile.

_wire_··on Steven Spielberg Thinks Aliens Are Among Us
Spielberg: As everyone in my business knows and expects, hype is everything to the box office. I'm risking a spoiler here but, ever heard of the Protocols of the Elders of Zion? (Whispers) Aliens
_wire_··on Most of the AI software YC companies are coming out with doesn't work
Enjoyed that final sentence.

Truly "wow-ed"... Or worked at all?

> "Oh you just have to use that other page and copy the ID from X with Y and we will get back to you about that other bug"

There was a little-known programming language in the early 80s called Z that accepted any input as a syntactically valid program, so the programmer's job was simply debugging.

_wire_··on The Iran War: How America, Israel and Iran Got Here [video]
Ezra Klein show
_wire_··on My AI Agent Knows Who I Am. Not Just What I Want. Who I Am
This article is miraculous for willingness to utterly damn AI agent technology as a bragging point for slavish obsession with it!

The opening:

> The agent works, sort of. It does things. But three months in, it’s making the same categories of mistakes it made on day one. The tool doesn’t improve. It just runs.

But there is a solution...

Dr. Thoughts' patent cure and tonic can relieve your agent's symptoms!

Step right up!

With Dr. Thoughts' tonic, all you have to do is continuously monitor and study the infinite ways an agent can go wrong and upload corrections.

Dr. Thoughts' patented architecture of Observation, Learning, and Context, including python pipeline and non-negotiable meta-systems monitoring, will focus your ADHD into eternal vigilance over your agents constant recurring errors.

Now, let's be totally honest, lots of tonics include python pipelines and feedback. But you've no doubt noticed that something is missing.

• Does your agent lack a personality? • Does it need an identity? • Is it lacking personal context? • Do you wish your agents really knew you?

What sets Dr. Thoughts' tonic apart is a special secret ingredient that no maker of tonics could ever know. You!

Just build an ever-expanding profile of yourself at the level of intimate details only a wife would know and continuously re-upload it into all your agents.

Voila!

You will be not just surprised but amazed by the results of sharing an ever growing database of privileged information about yourself with companies determined to exploit and manipulate every aspect of your life and society for fun and profit.

At this point you're probably thinking "This sounds good but how much does this miracle cure cost?" Now you can use Dr. Thoughts' patent cure and tonic just once to see the benefits. But for maximum effect, Dr. Thoughts' tonic must be taken daily. And with an installment plan, the token costs of providing all your personal information to an alien corporate intelligence bent on controlling the world can be remarkably affordable!

Get started today and ride the AI wave of the future: patent cures and tonics!

(Warning! Tonic consumption can damage internal organs and compromise your social credit score. Only take tonic under the guidance of a licensed Dr. Thought' tonic representative)

_wire_··on The War Trump Doesn't Want to Talk About
I think these user-contributed links to a single archive site are goofy: HN could generate these references automatically, or keep a FAQ-- OMG how radical! It's like modern-day samizdat!

This strange caveman presentation of HN where the same article is posted many times, and vigilant users try to condense previous references, make archive links etc by hand... except who cares bc the half-life of a post is 15 minutes and then immediately dumped down the memory hole!

How can a bastion of elite tech awareness be trapped in web of 1995?

Ye, it's quant. How pure and delightful: we're the most savvy Luddites in the world! Blow me

_wire_··on What it will mean for the economy if the Strait of Hormuz stays closed
Dr. President prides himself on guessing people's shoe size and gifting them shoes. Rubio is stuck wearing the gift:

https://www.aol.com/marco-rubio-pictured-145-trump-144634261...

_wire_··on What it will mean for the economy if the Strait of Hormuz stays closed
The straight stays closed? This can never happen!

Dr. President has thought of everything: as long as the bombing continues, morale stays high, and as long as morale stays high, the straight stays open. Elementary.

And if the straight stays closed, all the better:

> "The United States is the largest Oil Producer in the World, by far, so when oil prices go up, we make a lot of money," Trump said Thursday on his social media site.

_wire_··on Alex Karp: "People are likely gonna have less good, and less interesting jobs"
> "...increasing the power of working-class men."

Who vote Republican.

You know, class warfare.

Nothing to see here, move along

_wire_··on Show HN: What's my JND? – a colour guessing game
Thx for heads up!
_wire_··on Show HN: What's my JND? – a colour guessing game
This test is subject to the display regime on which it's presented. A given display may hide or amplify differences due to lack of conformance with the color space for the test data.
_wire_··on If computers are the future, why are computer users permanently illiterate?
Marshall McLuhan was interested about this in the late 60s and his observation was that electronic media can obviate literacy as the phonetic alphabet, and so literacy, was a technology, not an essential human trait. Language is media, not the contents of the media.

Video retrieves a tribal interaction by bending space and time allowing tribal modalities to manifest at distance.

McLuhan observed in his "Understanding Media" the "Tetrad", 4 laws that guide the uniquely human modalities to media:

1. Enhancement / extension

2. Obsolescence

3. Reversal

4. Retrieval / return

The phonetic alphabet enhanced the continuum of the visual by temporal regimentation, linear thinking and compartmentalization. Corresponding social status structures and modes of participation evolved to high levels of identity and status.

Literacy has been obsolesced by electronic media, which creates a highly connected, free-flowing, always-on, instantaneous mode of involvement. This is returning to previous unstructured tribal arrangements of status and participation.

While literacy enhanced bodily presence and evolved a strong sense of self and identity, electronics are reversing this as the participants are disembodied by the medium. An example is the new reluctance to talk by telephone because of the medium's awkward intensity of vocal cues minus any a sense of the bodily existence of the counterpart as a total person.

More generally, telepresences reverse to isolation and loneliness as one's body and identity becomes ghostly. This may lead to a retrieval or return to an importance of direct social interactions.

(Note that the m Tetrad stands on the figure/ground dichotomy of Gestalt psychology.)

Engaging the topic at hand via the tetrad:

The computer has obsolesced television, but is also retrieving it as we think about the internet as domains or channels, but with an infinite selector. When you have so many channels you need a new way to find and select them. The idea of television "programming" is merging with computer programming through AI.

Notice that the traditional (ouch) art of computer programming is becoming obscured by AI even as more people than ever create programs: the term program has become a euphemism for wish fulfillment, which a more rigorously postured "developer" will term "prompt engineering". The status of this euphemism is supported by the common notion that writing programs is the job of instructing the computer to perform a task (complete one's wishes). As the medium of AI is a sea change in technology for describing tasks, it is obsolescing programming as a literate art of mathematical formalisms for converting algorithms and data-structures into code for a CPU.

_wire_··on Targeting mistake led to US missile strike on Iranian girls school-media reports
Men at Work "It's a Mistake"

https://m.youtube.com/watch?v=I0AxrOUJ62E

_wire_··on Apple at 50: Five Decades of Thinking Different [video]
Looking back at Apple's products, 3/4 of them are impressively bad; very very bad. But the stuff they get right...
_wire_··on Ask HN: What software has improved dramatically recently thanks to AI tooling?
Nothing... But this is because the advances are so fantastic they can't be explained. We need AI to explain how incredible these advances are...

Oh wait, this backwater forum of advanced hyper-aware technologists with billions at stake have no idea about advances; they're behind the curve myopically focused on their ventures.

_wire_··on Iceberg was once the biggest in the world. Now it has just weeks left
When your fridge fails, you will typically find a pool of cold water on the floor... If you choose to regard this as your fridge is cooling better than ever, that's your prerogative. This observation is the essence of the effects we see about the global climate cataclysm.
_wire_··on I'm struggling and I don't have anyone else to share this with except you
My observation neé suggestion is to examine your expectations and what you think it means to be a nice (or good) colleague.

First principle: social milieus run primarily on instinct, so the first rule is know thyself: what are your instincts? What you find attractive in others, what annoys you. What do you want?

Next: You can't always get what you want but we tend to get what we need. Observe silent gratitude when you feel it and make a note of it.

Third: ask not what others can do for you, but what can you do for others, in a manner that fits your feelings, values, and after this, what do you wish for them. Offer your surplus gingerly to the people you like.

Continuing: Respect your own feelings about those you don't like. Understand yourself, your moods, peeves, pleasures, agitations.

Keep work relationships focused on the work at hand. Build trust with your own performance and stay focused on the work. When you have the good fortune to find friendly birds of a feather, treat these relationships with the respect and boundaries appropriate to the workplace. Don't let others suck you into their trips out of obligations of appearing friendly.

Lastly, most of who we are is unconscious and instinctual. Being nice all the time might seem appealing from a pragmatic view, but it's vibe that can truly annoy others subconsciously. If there's "a game" we all play socially it's to conserve the privilege of our solitude and time with people we truly like. Work is full bores.

(A bore being someone who deprives you of solitude without offering company)

Regarding destiny or a personal curse:

It is possible you are a true social outlier having an unusual and unfortunate experience that few others have seen nor can they appreciate. It's possible that workplace life is now generally insane. Consider your own temperament and values and how you will walk through hell.

Observe that while you might be in hell, by your own conduct and attitude, anyplace you end up can potentially become better because you're there!

If you aren't finding what you need, try building it for those around you.

_wire_··on Surgical Repair of Collapsed Attention Heads in ALiBi Transformers
Interesting that lack of engineering discipline in LLM design has industry vernacular veering into medically pathologized forensics and remediation.

From a C.A.R. Hoare perspective, this looks like a horrible development: Human bias in training of LLMs produces effects that convince credulous users that automata are alive, credulous users think that as we don't understand why living organisms do what they do, then as an LLM seems like a living organism there's no expectation of understanding why LLMs do what they do...

We put our own ghost in a machine, confuse the machine with our ghost, then rely on dark arts to cope with our lack of understanding of the machine.

So expect LLMs to be further mystified, treated as specimens for study and symptoms to cure, then categorized behaviorally by medically appropriated syndromes.

We can already see the burgeoning new high priest career-track Doctor of AI, with all the attendant quackery, patent cures, leaching and horse-whispering.

_wire_··on AI is to software as power tools are to woodworking
A drill press doesn't spontaneously turn into a table saw that ejects its blade because although you secured the bit in the chuck you didn't secure the blade on the arbor.
_wire_··on Codex, File My Taxes. Make No Mistakes
"Codex, File My Taxes. Make No Mistakes"

(clank clank)

"Codex, File My Taxes. Make No Mistakes. I really mean it!"

(whirr)

_wire_··on Ad-tech is fascist tech
Doctorow begs questions in the manner typical of self-hosted and/or expatriated journalists, who must be evangelists, but religious work tends to disregard the labor of research as it's hard work that doesn't pay and it might uncover a contradiction of the orthodoxy.

Let religious voices like Doctorow indicate places where we may examine policy and why we should be interested to look there, but sermons aren't vehicles to carry meaningful analysis.

Sermons must have a righteous tone and term "fascist", used correctly by Doctorow, has a long-standing colloquial connotation of teutonic allegiances during WW2, which emotionally overloads his diatribe.

Yet, he's correct to invoke it, even as his rhetoric is misplaced for audiences whose forbearers were sacrificed to the trauma of WW2, and whose generational scar tissues suppress remembrance that the patterns of evil which exploded across the Axis did emerge within our sacred bastions of liberty and freedom, and that these evils are fomenting again right now.

So I appreciate Doctorow's polemics even as I too regard the details of his claims with skepticism.

_wire_··on After outages, Amazon to make senior engineers sign off on AI-assisted changes
Like bombing a building full of little kids? Oops too late...
_wire_··on After outages, Amazon to make senior engineers sign off on AI-assisted changes
> because the time will just go into tests that wouldn't otherwise have been written

Writing tests to ensure a program is correct is the same problem as writing a correct program.

Evaluating conformance is a different category of concern from ensuring correctness. Tests are about conformance not correctness.

Ensuring correct programs is like cleaning in the sense that you can only push dirt around, you can't get rid of it.

You can push uncertainty around and but you can't eliminate it.

This is the point of Gödel's theorem. Shannon's information theory observes similar aspects for fidelity in communication.

As Douglas Adams noted: ultimately you've got to know where your towel is.

_wire_··on Something feels weird about this economy
What's weird is the stats are all manipulated. The measures and the ground are disconnected.
_wire_··on Behavioral Effects of High Peak Power Microwave Pulses (1992) [pdf]
Conclusions from studies on monkeys:

//Behavioral alteration was characterized by decrements in response rate but not by changes in reaction time, yellow light errors, and postreinforcement pause time. There are two distinct explanations for the singular effect of microwave pulses on response rate. First, the reduction of response rate may be due to an auditory sensation during microwave exposure and subsequent distraction of the animal from producing the normally consistent rates of responding during the red light. People have reported a hearing effect produced by pulsed microwaves. An early report described the effect as "clicks" or "hissing" located behind the head (10), that required pulsed energy for its occurrence. Several important studies have verified the hearing sensation (11) and measured cochlea microphonics produced by pulsed microwaves (12). Foster and Finch (13) suggested that the hearing effect was due to a thermoelastic expansion of tissue during absorption of a microwave pulse with a subsequent thermoacoustic pressure wave propagating through tissue producing the hearing effect by stimulating receptor cells located within the cochlea of the ear. It is reasonable to assume that the energy delivered to the head of our monkeys was well above the threshold for the hearing effect (14).//

//From this study, we can say that the threshold for exposure of the head to pulsed microwaves lies between 16 and 26 W/kg. The threshold for changes in behavioral performance during whole-body exposure to microwaves has been reported to be near 4 W/kg. The safety standards (1,2) employ a safety factor of 10 placing the maximum permissible SAR at 0.4 W/kg. The safety standards also employ a localized maximum SAR [specific absorption ratio] in any 1 g of tissue at 8 W/kg. The threshold range determined in this study for a localized exposure to the head to produce a behavioral effect of 16 to 26 W/kg suggests that the 8 W/kg safety standard does not provide a safety factor of 10.//

My favorite interlude:

//DOSIMETRY: The rhesus monkey carcass was mounted in the Plexiglas chair and positioned in the anechoic chamber to simulate behavioral test sessions. Four temperature probes were inserted into the carcass at various depths followed by microwave exposure at 9 pps for 180 s. Small holes were drilled in the back and top of the skull and small plastic cannulae inserted into the brain. Two locations along the vertical central axis and one along the horizontal central axis of the cranial cavity in the head were sampled at 10, 16, and 26 mm as measured from the surface of the brain. The fourth sample site was the center of the neck. Tissue temperatures were recorded every 10 s before, during, and after exposure to microwaves. Identical measurements were made after two different microwave exposures. The local SAR was calculated using the following formula: SAR (W/kg) = 4.186cT/t where T is the temperature change in degrees Celsius, c is the specific heat for brain tissue in cal/g/°C, and t is the exposure time in seconds. Average SAR of the whole head was estimated with the saline-filled balloon by measuring the temperature of the saline and quickly placing it on the Plexiglas restraint chair in the location of the monkey's head and irradiating it for 180 s. The balloon was then quickly agitated to provide mixing, the temperature was remeasured, and the SAR was calculated using the formula given above.//

//BEHAVIORAL: The monkeys were placed on restricted food rations and then trained by successive approximation to operate two levers for food pellet reward. This task, diagramed in Fig. 2, required the monkey to respond on one lever in the presence of a red light. The first response on this lever at the end of a variable time interval (30 s average, 2-140 s range) turned off the red fight and turned on either a green light (75% of the trials) or turned on a yellow light (25% of the trials). The green and yellow lights were illuminated for 1 s. A response on a second Iler while the green light was illuminated resulted in a food pellet reward. A response on the second lever in the presence of an illuminated yellow light was followed by a 30-s time out with the anechoic chamber lights and the stimulus panel lights (red, green, yellow) turned off. The sequence of red light responding followed by either a yellow or green light was typically repeated 100-120 times in a single behavioral test session. The behavioral session lasted 77 min and consisted of three 25-min components separated by 1-min extinction periods. During the extinction periods, the levers were ineffective, and all lights were turned off. Microwave exposures were always given during the middle 25-min component allowing three distinct behavioral evaluations: premicrowave, microwave, and postmicrowave.//

Suggested thought experiment: If you believe AIs are at least as sentient as rhesus monkeys consider the discipline and ethics implications of this typical monkey study when handling the AI for testing and applications.

_wire_··on AI Didn't Destroy Your Company, Your Processes Did
Guns don't shoot people...

Oh, wait, now they do

_wire_··on AI Needs Management Consultants After All
Can't this be done by AI?!
_wire_··on I can no longer write technical blogs due to coding agents
This comment captures this AI era of oddly verbose and empty observations:

"I recently wrote a blog about how I'm prompting." (skips reference to blog)

"I don't think it's boring."(lazy self-centered antecedent for 'it': prompting? blogging? blog about prompting? self?; operative terms are 'I don't think' and 'boring')

"I think we don't have the correct language to talk about the techniques/processes..."(lacks language but keeps going; adds empty concern about correctness for unspecified techniques)

"You can almost tell from the types of complaints people have that they are using it very differently." (repetition that doesn't clarify; but you can almost tell there is a point from some of the words)

"So, I think it's worth discussing." (go on...)

"...but I bet that not everyone prompts/guides AI the same way." (I bet a betting man bets; thank you; you're a a fabulous audience; I'll be here all week; try the veal!)

Slop is a very human trait, which explains why AI emits so much of it.

_wire_··on If AI is so good, why don't we have an infinite supply of 10x engineers?
Via Doctorow a day ago:

https://singletrackworld.com/2018/01/collision-course-why-th...

← PreviousPage 7 of 27Next →