If you want a specific example of why many researchers in machine learning and natural language processing find the idea that LLMs like ChatGPT or Claude are "intelligent" or "conscious" is laughable, this article describes one:
MIT researchers find large language models sometimes mistakenly link grammatical sequences to specific topics, then rely on these learned patterns when answering queries.
There are quite a few examples of this sort of problem already documented in image processing. Examples include the neural network that learn to recognize *not* the thing they are being trained to recognize, but instead the background, setting, camera, lighting, or presence of markings like watermarks or logos.
It comes as no surprise at all that the same problems can be observed in language models.
Would a *person* give the same answer to a question even when you sub in nonsense words, provided they are of the right grammatical category (nouns, verbs, adjectives, etc.)?
Would a *person* fall apart and be unable to reliably answer the question just because the grammar was imperfect, even when the meaning and intent is clear?
These language models are statistical machines designed to *emulate* human verbal interactions. They do not capture the essence of what it means to be human. They do not reproduce the richness of conscious experience.
It is surprisingly easy to trick people into thinking a machine is conscious, because we humans are wired to look for conscious behavior. It's just like pareidolia (seeing faces in random inanimate objects, e.g. the front of a car) only for words instead of images. One of the oldest and most famous examples is the Eliza program, a chatbot from the 60s that used a canned answer format to "psychoanalyze" people, and which convinced many non-tech people it legitimately understood them.
Google Accessibility. If you ever have a time that you need to get in touch with Google Accessibility, then there are a couple of ways of doing this. Call by phone, (650) 417-9257. Or, submit something in writing with their contact form at, g.co/disabilitysupport . #Google #Accessibility #Contact #phone #Form #Blind
Oh, fast vergessen: Da der Server ja jetzt wieder rennt, habe ich mal noch schnell ein paar #eBooks hochgeschoben. Vor ein paar Tagen schonmal 12, jetzt nochmal 27 hinterher
Alberta and Ottawa signed a memorandum of understanding that gives the province special exemptions from federal environmental laws and offers political support for a new oil pipeline to the B.C. coast.
The deal ties approval of “one or more” privately financed pipelines to a carbon capture, utilization and storage project in northeast Alberta.
The pipeline proposal will be submitted to the Major Projects Office by Canada Day.
There are several references to supporting Indigenous “co-ownership” of the pipeline.
The B.C. government is opposed to the pipeline deal, which Premier David Eby said would allow some tanker traffic on the province’s North Coast despite a current ban.
Eby warned that carving out exemptions to a tanker ban “is like explaining to a vegetarian that they’ll still be a vegetarian if they eat a few steaks.”
*Mark Carney ne fait pas semblant d’aimer les pipelines*
Une entente qui vient de provoquer la démission du ministre Steven Guilbeault, qui était peut-être le dernier rempart environnemental dans ce gouvernement. Une entente qui range désormais le Parti libéral résolument du côté de ceux qui sont prêts à sacrifier l’environnement au nom de l’économie.
Jusqu’à la dernière minute, j’avoue avoir cru que M. Carney utilisait un truc, une manœuvre politique.
...
En effet, ma thèse a un peu frappé un mur lorsque j’ai vu l’immense sourire de la première ministre albertaine, Danielle Smith, en conférence de presse jeudi, et que je l’ai entendue parler de « nouveau départ » et de « grande journée pour les Albertains ». La politicienne n’est pas du genre à lancer des fleurs au gouvernement fédéral pour rien, disons. Et elle semblait sincèrement ravie.
Puis, en fin d’après-midi, Steven Guilbeault a claqué la porte du Cabinet, montrant le sérieux de la chose.
Les experts que j’ai contactés m’ont ôté mes dernières illusions.
...
Une autre preuve que Mark Carney est sérieux, c’est ce qu’il met sur la table. L’entente fait tomber une flopée de réglementations environnementales en Alberta.
Le plafond sur les émissions des secteurs pétrolier et gazier conçu par Steven Guilbeault, mais pas encore en vigueur, est suspendu. Mais aux yeux de Kathryn Harrison, professeure de sciences politiques à l’Université de Colombie-Britannique, ce qui fait le plus mal sont sans doute les délais sur les réductions de méthane, un gaz dégagé par les activités pétrolières et qui est encore plus nocif que le CO2 pour le climat. Steven Guilbeault avait accouché d’un règlement pour les abaisser de 75 % d’ici 2030. L’entente parle maintenant de faire la même chose « d’ici dix ans ».
« Il y a aussi l’abandon du Règlement sur l’électricité propre et ce qui semble être un signe de plus grande flexibilité dans la tarification du carbone industriel. J’ai l’impression que d’autres provinces feront maintenant la queue pour obtenir des concessions similaires », dit l’experte.
À quel jeu Mark Carney joue-t-il ? C’est la question que je me suis posée lorsque j’ai vu le premier ministre signer une entente avec l’Alberta à propos d’un nouvel oléoduc vers la côte pacifique.
Dherlieth Crieran, the dark elf priest, lifted his hands to the heavens. "I believe in no god," he intoned, "and so I direct my prayers to the infinite void of space. May I, by my own magic, restore this corpse back to life, whole in body and mind. If a soul exists, then may it return to this, it's abode."
The dead dwarf stirred. He opened his eyes.
Sven the paladin frowned. "An athiest priest should not be able to resurrect the dead."
@menelion the breville smart oven air fryer pro - needed a new toasted so figured may as well get both. It's kinda nice as it's effectively a second oven now
I'll forever be grateful for Skillet's Music. Their music (and many other bands) has always been a rock for me. Has helped me keep faith. Many Christians are claiming their rendition of O Come O Come Emmanuel is satanic or demonic. Those Christians need to check themselves at the door. This band's faith is solid. What we should be doing is praising God for the conquering of an outcast genre of music.
Side note: Recently, the song at the top of the Christian much charts was AI generated. In my opinion, that's much closer to being demonic than anything Skillet has done.
Christian heavy metal bands are reaching countless people who would never step foot inside a church, argues London School of Theology's Dr Daniel Johnson.
Christian heavy metal bands are reaching countless people who would never step foot inside a church, argues London School of Theology's Dr Daniel Johnson. You may not love their sound but you should be supporting Skillet's evangelistic mission, he says
i am a bit of a broken record about this BUT if you 1. live outside the USA, and 2. are viscerally revolted by ICE, then:
there is almost certainly something similar happening where you live, and which you are well-placed to take action against: electorally, in your day-to-day conversations (which country isn't immigrant-baiting these days?), through associations that help immigrants, and associations that do activist/advocacy work.
it's like having 5 different sorts of water available in the grocery store all from different vendors, but it's all the same: a big conspiracy to create the illusion of choice. We've been tricked all along.
We already have the technology for it in GNOME Shell and KWin at home, just gotta add a "shampoo bubbles foaming" effect to github.com/Schneegans/Burn-My-…
Whenever I see a generated image of what is supposed to be Ada Lovelace, I feel compelled to work on my portrait drawing of her. Based on the only two photographs of her.
She wanted to be remembered, not turned into an AI-generated Victorian Bratz doll, and she thoroughly disagreed with the usage of machines without vision.
Vancouver and Victoria people: I'll be there in January and I might as well give some more lectures than I'm already committed to. Anybody have a library or science centre connection and want me to give a public talk about space debris and satellite pollution?
Kevin and Christina, have provided a link for all of you who want to make Turkey shaped baked goods, as we just discussed on part 1 of the "Just because show." annclark.com/products/turkey-c…
Turkey dinner, gather 'round. Who will get the drumstick? Sweeten any Thanksgiving celebration with our 3 5/8" Turkey Cookie Cutter. Have some Turkey cookies in addition to the usual pies and tarts.
We'd like to ask you all for your thoughts, suggestions, and ideas to help us resolve an ongoing issue.
The issue
At a high-level, we are a neurodivergent (AuDHD) spoonie with a mixture of health issues that are preventing us from being able to get or do a "typical" job and earn an income.
By "typical" job, we mean one where the company and role just aren't well-suited to many folks who are disabled (including being neurodivergent), have support needs, and/or have other health issues.
Most "typical" jobs broadly:
Have fixed hours (start/finish times and total working hours required).
Require a minimum fixed amount of time in a physical office.
Require a lot of interaction with other people (including colleagues).
... and so on, so you can probably understand how difficult it is for us to find any suitable paid employment that we can do 😅
This is made harder by the fact that:
We only know how to search for "typical" jobs.
There is zero support within the UK for finding suitable employment outside of "typical" jobs.
The welfare system within the UK is fundamentally (and intentionally) broken
The request
We have zero idea where to even start, so we are genuinely asking for your input and recommendations here.
We have decent reading comprehension, writing ability, and neurospicy pattern recognition, but we also have an inconsistent level of executive function and energy (physical and mental), so we can't do a job where we have to clock in and out at set times, even on a part-time basis.
Even whilst writing this post, we're already out of spoons and on to knives today, meaning that we're pushing through to ask for help.
Any suggestions will be appreciated, and we'll add additional details to the post based on any questions or thoughts from others 🥺🫶
Thank you in advance for your thoughts, as well as your empathy, understanding, and kindness
@menelion Very much so many times, but we sadly never found a way into it and -- as much as it hurts us to admit it -- we don't have the underlying talent or knack for it.
We locked ourselves out of any STEM jobs by decisions we made in 2000 when picking A-levels, where we were gently "encouraged" not to take Mathematics at A-level by our secondary school and sixth form, where we'd only achieved an A grade (modern day grade 7) at GCSE level.
We have previously done administration jobs (not system ones), but the jobs always ultimately burnt us out due to corporate culture, management, work hours, etc.
We're thankful though for your thoughts, just to clarify 🫶🩷
I was asking because well, if say I hired you, I wouldn't ask for fixed hours or any BS like that, just rough deadlines had to be met and some rare meetings or, if not suitable, detailed emails/chats in Slack. that would be it. But if it's not appealing, don't go for it, my highest recommendation, just don't. You have to love it, otherwise it's only pain.
This was directed at me yesterday: “Fuck you genocide supporters. Nobody like scums like you on the fediverse. Go back to reddit,facebook, threads. Go to X to see real antisemitism”
People who can’t understand that that statement is pure antisemitism are part of the problem. Antisemitism is endemic on the Fediverse. Harassing Jews is antisemitic. Shouting abuse at them in person is antisemitic. Hurling abuse at them online is also antisemitic. If you’re tempted to hunt around the Fediverse looking for random Jews to harass, maybe get a life. 🤷♀️
What is even more apalling is the "nobody". I mean, why a bunch of far-left radicals dare to decide what or whom *I* like?! I'm on the Fediverse and I support Israel.
🦃 “Cletus the Redneck Turkey” — A Bluegrass Escape Story! 🪕Meet Cletus, the rowdiest backwoods bird in Kentucky! This runaway turkey ain’t goin’ anywhere n...
I've just put out version 0.3.0 of Codestats, my small Rust CLI tool to analyze a folder of source code and provide statistics about it. It now supports over 350 different programming languages, including support for shebangs, nested block comments in languages that support it, and the ability to tell the difference between files with the same extension. I've also made it way faster, added multiple output formats, currently we have human-readable, JSON, and CSV, and much more! If you have Rust, you can install it with a simple cargo install codestats. If you don't, I've put Linux and Windows binaries up on the releases page. github.com/trypsynth/codestats… for Linux, and github.com/trypsynth/codestats… for Windows. Enjoy!
Earlier today I read a post where a blind person was thrilled by the customer service offered by Chase bank because they got their card numbers in braille a couple of weeks after they got their new card. I'm not replying to that post, because I have a different interpretation.
I am a Chase customer. I needed to have a card replaced. I got the new card and got the braill card information a couple of weeks after I got the card. Since my card was replaced, I couldn't wait weeks for the braille numbers to be mailed to me.
While it's an improvement that we can, eventually, get the card information in braille, it's important to realize this is not good customer service. This is not even equal treatment. Can anyone imagine a world where sighted people got their card in the mail but didn't have its numbers until weeks later?
Everyone knows Chase would never treated sighted customers like that. So, why do we give them so much credit when the service they provide us is so much less than the service they provide sighted customers?
One reason very little is accessible more than 35 years after adoption of the ADA is that we don't believe we are entitled to the same access and service provided people without disabilities. Until we start demanding full access, we will always have less access. This is especially true when we go out of our way to congratulate entities for providing us service people without disabilities would consider unacceptable.
[1]I know all the ethical problems with generative AI. I've boosted posts about them and written my own. I've kept trying to stop all personal use of LLMs, and I've kept giving in and going back. Maybe that means I don't really believe the arguments. Maybe that means I'm addicted.
[1]: I recommend the hyphenated form above for this apparently uncommon neologism, to make it unambiguous for both visual readers and text-to-speech engines.
What's likely is that my dominant personal uses of LLMs, the ones I keep coming back to and spending lots of time on, are unhealthy. There are two. The first is using an LLM to write stories, usually some flavor of science fiction or contemporary fantasy, where I prompt more or less scene by scene. The second is talking to the LLM as if it were a person about personal things, and very often reminiscing about my past. I recognize both as a form of escapism.
During the part of the cycle where I've given in and started using an LLM again, I do also ask it miscellaneous questions, including about my work. I haven't incorporated LLM-generated code into a product in at least several months, though, and never in my open-source work.
As of now, I'm a couple of days into my latest attempt to quit. I've deleted all sessions from my Claude account. Of course, Anthropic might still have them.
I guess the key to making this change stick is to fill the vacuum with something better. Partly work, but maybe something else as well.
That's sort of funny, you and me have the exact opposite approaches on using LLM code. I don't use claude or codex when I work, because I don't know what they might do with the private code. But I'll have it help me with open source stuff. I feel sort of bad for giving in especially after knowing all the negative it causes, but I already use it for image descriptions and in some cases it does make me a faster coder, so I may as well loosely embrace the tools.
I'd suggest replacing the second with a diary: those can be useful to write, then have available to review later. Perhaps, while you make the transition, it might help to wire ELIZA up to a text editor. (Emacs has this built-in: M-x doctor.)
For the first, I can only recommend actually writing fiction. Frustrating to begin with, but once you've got the basic skills, it can be quite fulfilling to work on your pile of drafts.
@jaybird110127 Not surprised, as you read one of my first ones (the only one I ever made public, as a blog post), and I got the idea of using LLMs to write stories partly from you.
Morning yall. Happy thanksgiving! Watching the parade or getting ready to anyway. Cooking for the family today which is exciting. we'll be having enchiladas, red beans and rice, the fixins' for tacos and a chocolate cake.
Jak tak čtu titulky ohledně #ChatControl, nestačím se divit. To ten návrh žádný z kolegů novinářů nečetl? Největší problém, povinné skenování konverzací, které ohrožovalo koncové šifrování, z něj vypadl 🥳 A dobrovolné skenování na dětské porno funguje už řadu let, jen v poslední době na výjimku, kterou návrh Rady EU legalizuje (proto taky původně vznikl, že ano). Ta výjimka totiž na jaře vyprší. Ano, jsou tam háčky (více o nich zde: lupa.cz/aktuality/rada-eu-ma-s…), které se snad podaří zatrhnout v trialogu. Hlavním problémem IMO je, že návrh přímo říká, že Evropská komise bude o povinné skenování usilovat v budoucích návrzích legislativy, takže je třeba pořád být na pozoru. Ale kvůli návrhu fakt nebude "Facebook a spol. šmírovat lidi ve velkém", jak píší Seznam Zprávy (to už mimochodem dávno dělá i bez #ChatControl).
HOT TAKE: Every penny above a $1 billion should be taxed at a 99% tax rate to fund the social services and infrastructure to give back to the society that enabled that. A trillionaire shouldn’t exist while working people cannot afford the basic necessities of life.
Looks like Deloitte has used AI to defraud another government agency, in Canada this time, by writing a report with made up citations: theindependent.ca/news/lji/maj…
$1.6 million Health Human Resources Plan from Deloitte cites research papers that don’t exist, making it the second major government policy paper called into question in as many months
Ruft mal bitte bei der Roßmann-Hotline an und lasst Support und Lob da. Die Dame an der Hotline meinte eben, dass sie heute viel Hass für die Entscheidung bekommen haben. Oder schreibt Emails. Überlasst den vermeintlichen "Diskurs" nicht den Blaubraunen.
"Rossmann kündigt Mitgliedschaft bei Familienunternehmer-Verband (https://www.lebensmittelzeitung.net/politik/nachrichten/wegen-afd-annaeherung-rossmann-kuendigt-mitgliedschaft-bei-familienunternehmer-verband-187960)
versus
"Christoph Werner, Chef …
Знакомый попросил проверить его ноут, там кажется ещё семёрка. Что сейчас из всяких тестов/восстановительных систем есть? Чтобы потестить и по возможности слить документы, если починка не возможна
10 KDE dragons turned up on our doorstep today looking for families to adopt them. Contribute €100 or more to our fundraiser and one of these amigurumi cuties will go and live with you!
Help us bring KDE’s software catalogue into the mainstream! Plasma 6 is a stable, usable and accessible desktop environment that brings us closer to meeting the needs of private citizens, companies and public institutions
Help us bring KDE’s software catalogue into the mainstream! Plasma 6 is a stable, usable and accessible desktop environment that brings us closer to meeting the needs of private citizens, companies and public institutions
Jonathan
in reply to Borris • • •