I just Googled a term I didn’t know, and Google’s AI Overview informed me that the Second Coming of Christ is a tenet of Judaism.
I’m really worried that people are taking Google’s AI Overview at face value.
Here’s an example where I happen to have the background knowledge to tell me how wildly wrong this is. But I don’t have background knowledge about everything, and students have a lot less than I do.
#generativeAI #Informationliteracy #teaching #AIOverview #Google
And today I learned that you can’t turn it off. #Google is forcing its babbling, useless, and sometimes dangerous experiment onto a public that never asked for it.
Remember when people were LIVID that Apple put a new U2 album on everyone’s iPod? Maybe you like U2 and maybe you don’t, but at least Bono has never recommended that you add glue to your pizza recipe.
#LLMs #generativeAI
@Julie
I'm glad you're pointing it out.
Slightly more disturbing for me are the teachers and professors using AI to *GRADE* papers or the journals using it for peer review.
I've got a scary true story: I was talking to an Ivy-League current professor working in academia during a broader discussion about "what AI can do" at the university. I've been tinkering with these things for a while and I said, "Reliably? Not much". You have to keep knowledgeable humans in the loop and, in at minimum, have narrowly focused AI with data that isn't just whatever can be scraped indiscriminately off the web--that was my basic answer.
The person wanted it to summarize and evaluate student papers. The peer reviewers, likewise.
What kind of world are we making? The students use LLMs to write the paper and the professors use the LLMs to grade them. What do people THINK will be the result of putting the whole world on autopilot?
I'm not even against machine learning, but this is dumb
@IntentionallyBLANK At the high school level, where I work, teachers are excited about how much time LLMs save them. Teachers are overworked and micromanaged, and many feel like this technology is finally allowing them a bit of freedom from the aspects of the job that are pure drudgery. They don’t want to hear about the downsides. I would hope the university is different, but of course, professors and adjuncts are also overworked, if not quite as micromanaged.
The outputs of LLMs require close reading and critical-thinking skills.
Outsourcing that to the synthetic machines, might be pedagogical malpractice, which defeats the entire purpose of the endeavor.
There are no winners in this game.
@Julie
I completely understand the temptation, but it's so poisonous to good outcomes. It kills communication, critical thinking, and reading comprehension at the root.
LLMs are so confidently bad at so many things. The kids are fast to pick up on feedback provided by an LLM but passed off as the instructors own words. The instructors may think people can't tell, but young people and people who work with these things to program/annotate data/review output pick up on the quirks in a heartbeat.
It's demoralizing and encourages bad behavior all around.
It’s Robot Olympics: https://www.superversive.co/blog/hunger-games-for-robots
@IntentionallyBLANK @Julie
And then they will go off into the world with resumes written and optimized by LLMs and then sorted and hired by LLMs.
It will all be robots talking to other robots about what robots want to see and then make. it’s so decadent, so lossy, it takes all the world’s knowledge but with no feedback or context and garbles it into nothing. Might as well just put human knowledge into a paper shredder.
@IntentionallyBLANK yes! It’s just like inbreeding.
I got a laugh the first time I heard of Habsburg AI theory. Everyone thinks they're getting Jarvis/Vision and they're really getting a dumber Eliza with a Habsburg chin
@IntentionallyBLANK @Julie "I'm not even against machine learning, but this is dumb" spot on!
It's exactly as dumb putting the microphone in front of the speaker.
Imho it's totally bizarre that hordes of well educated people completely fail to see that and are happily feeding a country's worth of power into it.
@iwein @IntentionallyBLANK @Julie
"It has always seemed to me that my existence consisted purely and exclusively of nothing but the most outrageous nonsense."
- Thomas Ligotti
Letters of recommendation written by LLMs will be (are being) reviewed by LLMs to choose students who used LLMs to write their application essays.
@Julie I absolutely HATE Google AI. I don't even read it, but make a point of scrolling down below it before I begin to read responses to my searches.
@rebelyn2 Everyone should do that! But Google makes it hard for unsophisticated searchers to ignore.
@Julie The Rapture is an entirely American, 20th century religious belief, to boot. Including it, even if the conversation were about Christianity, is like including Xemu.
But including it in an article about Jews means that, even with Google working at its best, it’s stupidly, toxically wrong.
@Julie
My policy, which I strongly recommend to everyone: Do NOT even look at the "google overview. If it finds articles, papers, any publication listing date and author and gives you a link thereto, look at the reference and decide if it is pertinent to your intention launching the search. YOU decide that. Do NOT let the search engine tell you the answer from its' supposed research.
@Julie but there's a link now, it's reasoning!
/sarcasm
@Julie omg
@cynthiarose @Julie The same “AI” that said pregnant women should smoke 2-3 cigarettes a day, and that geologists recommend eating rocks.
It is absolutely insane that people are treating autocorrect on steroids like it’s Hal.
@MissConstrue @Julie
An autocorrect that changes “its” to “it’s” EVERY FUCKING TIME….
@cynthiarose @Julie Omg, Right!!! Like, the OED is right there y’all. My autocorrect has changed some really strange things, and almost always wrong or grammatically incorrect.
@Julie Life is so much better with DuckDuckGo. I still do my own search and checking of results wizhout any shoehorned "Ai" summaries. There are some, but I haven't seen one yet and their whole DuckAi is a separate feature.
@Julie It's too easy to poison AIs and I am sure it's being done as a tool of propaganda and misinformation. Tech companies only want to somehow "win" this market, and seemingly will do anything to do so, but many don't want their information to be crawled and others abuse the AIs with the complicity of the Ai companies themselves.
@Julie How long before Google starts offering a paid option to remove AI results?
"Here's half a raw squirrel pelt, Google Baby because that's just what you're worth!"
@too_little_caffeine @Julie I keep trying to set this as my chrome homepage (i use Firefox primarily) but Google won't let me
@CosmicTraveler @Julie Might need a chrome extension for this
@CosmicTraveler @too_little_caffeine @Julie I was able to install the Konami page as a web app on my Dock at the bottom of my Mac! I was so happy to find a way to get there quickly. Now I will see if it works better than Duck Duck Go has been working lately. Google has become almost useless to me.
@cobalt @CosmicTraveler @Julie UDM 14 is Google. It just disables the AI bullshit.
DDG sources their results from Bing.
Try startpage.com for anonymous Googling or qwant.com for a completely independent search engine.
@CosmicTraveler @Julie Alternatively, use startpage.com. They source their results through Google but proxy it to anonymize you. It accepts all of Google's Search Operators.
@Julie now I want to share crappy AI results from Google and tag them with #BonoWouldNever.
@Julie If you’re using something like ublock origin you can at least hide that element. It sometimes flashes on the page for a second but so far the block has been working successfully for at least a few months.
@Julie You can turn it off by switching to another search engine. I've been using Duck Duck Go for years on privacy grounds.
@Julie You actually *can* turn it off, but it's hidden. Add '&udm=14' to the search results URL, and those useless summaries and 'people also ask' suggestions disappear, and it starts to look a bit more like it always used to. There's a browser plugin for Firefox that makes the search bar default to it.
@dickon Thanks! I learned you can also use the website https://udm14.com.
@Julie That’s… extremely offensive.
@Julie A group of rabbis should sue.
@Julie What, and I can not say this with adequate emphasis, The Fuck?
@Julie twice now I've google prescribing information for medications I've started and discovered that the medical information that the AI summary provided was worse than inaccurate, it was actively harmful
I love how people expect AI to spit out stone tablets of truth while literally half the nation is drooling out made up gibberish.
@Julie @dangillmor
LLMs have no place in the pursuit of facts. This is why I pay for #Kagi search. I do use LLMs for foreign language practice. They’re good at make-believe. That’s why they’re bad at fact-finding.
Dear @BlumeEvolution
Here (the post above) a wounderful example from your own field of expertise that shows, why I don't share your optimism about AI. Yes, it's just another anecdotal evidence, but that pile grows.
To be precise: I'm sceptical about general AI. There are certainly special use cases, where these statistics tools are useful. Trained for a spcific purpose to be used by experts in the field. Not for public use.
Dear @PiiiepsBrummm , well, that is my critical position on AI, too.
Instead of overpowering General AI
https://scilogs.spektrum.de/natur-des-glaubens/ki-atomkraft-wohin-wendet-sich-europa/
@Julie The technology is sound but the data is not. Garbage in, garbage out. If there are no humans on the loop curating the data, it's useless...