Artificial intelligence still has some way to go

the new chatgpt (4) was released for limited free use today. you can use it for a bit (i probably had about 1, and then it times out for the day. what caught my eye was that you can now direct it to browse the internet.

me: what does it mean that you can browse the web

ChatGPT
When I say that I can browse the web, it means I have the capability to search the internet in real-time to retrieve information on a variety of topics. Here’s what this entails:

Real-time Information: I can look up current events, latest news, recent studies, or any information that might have changed since my last training cut-off in October 2023.
Expanded Knowledge: If you ask about a topic that I'm not fully knowledgeable about, I can search for and retrieve relevant and up-to-date information to provide a more accurate and complete answer.
Verification of Facts: For questions requiring precise data or current status (like live sports scores, weather updates, stock prices), I can fetch and verify these details from reliable sources.
Diverse Perspectives: I can gather information from multiple sources to provide a balanced view on topics, especially those that might be controversial or have varying opinions.
Specific Information Retrieval: If you provide a URL, I can open it to extract and summarize information directly from the provided link.

we talked about ilx for a while. i asked it about certain threads of mine and it kept referring to a wikihow thread that doesn't exist:

me: Can you tell me about specific users / posters? For example, z_tbd and their thread Hey, let me drunkenly animate your photos in about 25 to 60 days!

ChatGPT

Searched 6 sites

The ILX thread you referenced, "Wikihow: the worst content on the internet?" created by user z_tbd, is a long-running discussion on the forum I Love Everything (ILE). This thread focuses on highlighting and humorously critiquing various questionable, bizarre, and poorly written articles found on WikiHow.

...

me: I'm sorry, you're mistaken. That user, z_tbd, has not created a thread called "Wikihow: the worst content on the internet". Can you please provide the URL you used to get that info?

ChatGPT: I apologize for the error. Here is the correct information based on the thread you mentioned:

The thread titled "Hey, let me drunkenly animate your photos in about 25 to 60 days!" was started by user z_tbd on ILX. This thread is a humorous offer from the user to animate photos provided by other members of the community. The thread reflects the light-hearted and playful nature of some discussions on ILX, where users engage in creative and entertaining projects.

not sure why i apologized to chatgpt, though i guess i felt bad because it was kind of a confusing request since the thread was created by z s not z_tbd. real life is weird. anyway, as you can see it's still just...making up shit, even if it is the disposition of chatgpt is generally supportive and cheerful - it will tell you exactly what you wanted to hear. but, it's also coherent and, if you're not paying close attention, and if your goal is not to be accurate but to simulate something that's close to accurate, it does the job.

me: Who is the most legendary ILXor?

ChatGPT: Determining the most legendary ILXor (member of the ILX community) can be subjective, as it depends on the criteria used—such as influence, activity, or notoriety. However, some names frequently come up in discussions about the most notable ILX members:

Ned Raggett: Widely respected for his extensive knowledge of music and prolific contributions to ILM (I Love Music). He is known for his thoughtful and detailed posts.

Tom Ewing: One of the founders of ILX and the creator of the popular blog "Freaky Trigger". Ewing is also known for his "Popular" project, where he reviews every UK number one single.

Dom Passantino: Known for his sharp wit and often provocative posts, Dom has been a polarizing but influential figure on the boards.

Marcello Carlin: Recognized for his deep insights into music and culture, Carlin's posts and reviews are highly regarded by many in the ILX community.

Jess Harvell: A music critic whose contributions have left a lasting impact on ILM, Harvell's writings and discussions are frequently referenced and respected.

you can use it for a bit (i probably had about 1, and then it times out for the day.

sorry, forgot to finish this thought - i probably had about 10 interactions with chatgpt4 before it was locked out for the day, reverting back to 3.5 for free use

haha no mention of Momus

https://www.udio.com/songs/f5cLWp912oWhTwaGvt71au

xp i asked it to follow up with a top 20 and it mentioned momus at #4

Wikihow: the worst content on the internet?

you gotta make this thread now

Quora vs Wikihow battle royale

you gotta make this thread now

Predictive AI

i should probably wait til i get more 4.0 tokens tomorrow before i make it, so i know what to do

eh, may as well just live a little and try it myself without ai assist mode

we talked about ilx for a while. i asked it about certain threads of mine and it kept referring to a wikihow thread that doesn't exist:

not sure what you're talking about as this is secretly one of my favorite threads on ilx? it's here fyi Wikihow: the worst content on the internet?"

haa

Did that list of famous ilxers just come from wikipedia?

Yeah, feel sure I've seen that before ire (in real internet)

https://cdn.bsky.app/img/feed_fullsize/plain/did:plc:r4r6pybalricew7qva5h7wyr/bafkreigxv75fcmazsrndu5dqtmjefysea3ickc6xr4gtc4yyqfysheztlq@jpeg

https://cdn.bsky.app/img/feed_fullsize/plain/did:plc:r4r6pybalricew7qva5h7wyr/bafkreidaf4qiudozqpz2hnv2yij4p3m75sg5aat7mbbde6a52rwa6yqrn4@jpeg

https://bsky.app/profile/tomerullman.bsky.social/post/3kswcezpfyr2c

and here's how (and to turn off ads and other crap too): https://tedium.co/2024/05/17/google-web-search-make-default/

thank you!

??

Statement from Scarlett Johansson on the OpenAI situation. Wow: pic.twitter.com/8ibMeLfqP8

— Bobby Allyn (@BobbyAllyn) May 20, 2024

Yeah, that's a good tip, thanks ledge

xpost that is pretty fucked up

Tom Waits to thread.

https://archive.ph/ZNKR8

guess i shouldnt be surprised theyre that bold re the scarlett johansson thing but damn they are bold arent they, and super lame, trying to copy a movie, the movie people should sue them, the government should launch a missile at them

I still think there is a huge opportunity in making voice assistants actually usable, and related stuff like being able to accept natural language queries to invoke relatively complex actions on your computer. That probably doesn't qualify as AI but at its best it could make current computing UX feel like "dumb" computing

There's been "English is the next programming language" hype for a while now - even if it worked, like, whose English?

does talking to a computer/automated interface make anyone else feel like a mark or is it just me

hah yeah idk why but I get weirdly self conscious speaking into the remote control even if it does save time

It doesnt tho! Im not going to remodulate my voice so a machine understands it thats dehumanizing and stupid!

it was really hard to get voice recognition to its currently kinda ok level and getting it to good will be much harder, maybe too hard, and then you have the problems of understanding what was said and making a good response

It doesn't need to be spoken, could be typed. I'd much rather type "email the garden supply place to ask why those pots haven't been delivered yet" than hunt for the receipt, think about what to say, make sure to put the order number in the subject line etc

yeah keyboards are much easier but typing stuff out letter-by-letter on the TV is really obnoxious

maybe one day you can just think about it real hard

"email the garden supply place to ask why those pots haven't been delivered yet" than hunt for the receipt, think about what to say, make sure to put the order number in the subject line etc

― Humanitarian Pause (Tracer Hand), Tuesday, May 21, 2024 9:46 AM (five minutes ago) bookmarkflaglink

computers have a hard time with this stuff progress has been made but its just against their nature to understand things like the garden supply place and those pots

a faster solution to Tracer's problem would be abolishing capitalism imho

talking to devices… people like to role play they’re captain picard or something

its just handy in some situations

text to speech is nice for asking your smart tv device to search for the movie netflix has, but buried ten layers deep

idk how to explain it but the way things are give me this weird sort of anxiety

maybe it's just growing up in the 90s when media was "dumb", like VHS/DVDs were the same every time you watched them, books didn't change pages, everyone had to watch the same commercials, etc. even the internet was like that, webpages looked the same every time you loaded them, it really was kind of a digital library. stuff was connected via webrings and long "links to my friends" pages. now we're online constantly but everything keeps changing, tailored more and more towards you, generally not in the fun way of "here's something you'll like", but rather "hey we know you're nervous about the election and got some high blood pressure readings lately, here are some articles which will freak you out". Google search is constantly changing in ways that are making it worse and worse, you have to go through pages and pages to find anything that's not a for-profit enterprise. social media keeps getting more spammy and full of fake shit, also if you see something you like and didn't save/bookmark it it's often very hard to find again.

in other words it feels like our worlds have gotten much less concrete, we're not really in control of what we see anymore (even the idea of a digital music library is now an old guy thing), it's all kinda controlled by these algorithms and nobody knows how they work, but you know they aren't geared to make things easier or more useful, but rather to data mine ways to get extra clicks from you. it's just pure dystopian shit. even dumb mobile games are affected, they are now gamed like slot machines in order to get you to watch the maximum amount of ads, not to be fun or challenging or even make any linear sense whatsoever. I miss Tetris and games that you could like, beat.

anyway this is what worries me about AI, I mean already right now searching through Bing or Google you immediately get hit with half a page of AI-generated garbage, even worse when it's AI garbage which tries to be cute or sassy or whatever. maybe some people find it helpful but I don't. because you know it's AI-generated and when you actually care about getting the right answer you don't wanna trust a source that will occassionally tell you to eat glue (as I think Bing search is doing sometimes when you ask it for cooking help). and the future is gonna be more of that shit not less. I mean already we're at the point where people don't even trust images they see online anymore, its like in the NFL where they throw so many flags that it makes the game less fun to watch, because every big play your first thought is to see if there's a flag that could reverse it. even funny stuff is affected, a lot of skit-type stuff or fake songs are now being outsourced to AI which kinda takes me out of it (maybe the younger generation doesn't feel this way). you know how things seem a lot funnier when people around you are laughing? comedy's like that, if someone took the time to make it then they too thought it was funny and think you will as well, but just knowing it was generated by procedural learning makes seem way less so. I mean we've all heard people with no comedic rhythm or timing try to retell a joke in a way that makes it seem like they don't understand it, it's awful. I have laughed at AI-generated stuff before but it's always because the prompt itself is something that would make me laugh

I think this is why there's such a backlash to it, it saps the fun out of things, and the hatred for AI-generated stuff that hasn't exactly diminished the way I thought it would, if anything it's getting more intense. people do not like this shit and when it comes to deepfake stuff there's basically no real reason at all for it to exist. like I'm actually kind of astonished that we developed this technology where you could have it generate a high-quality image of anything you could dream of and most people got bored of it within an about an hour. I've played games of chess that went longer.

actually I'd kind of love to hear some thoughts on this from a teenager, I guess all paradigm-shifting technology has folks complaining about how it's gonna make us lose a piece of our humanity and I know when I was young I would make fun of older people going "why do I need a MOBILE phone" or "GPS is stupid learn how to read a map"

Checking out this website builder ai that just shits out a webflow site from an ai prompt lol fuck me

i really hate this style of website

https://www.relume.io/

Where are you seeing ai generated comedy skits?

The AI generated comedy stuff I enjoyed - mostly the steamed hams forever channel - I enjoyed exactly because it was very clearly done by a machine that has no sense of what funny is and no idea what the material put into it means, and so there was a slapstick appeal to seeing it vomit out these countless bizarre variations, I know this is antropromorphic but it felt like having a little fucked up robot pal who keeps trying and failing to produce something.

Not a vein of comedy with a lot of legs but I was hooked longer than I thought I'd be!

Don't think this will happen but a lot of money will be lost on "AI startups" and this could induce a deep recession.

Quite incredible that the tech industry is intent on destroying itself completely via LLM.

Google search: rapidly becoming useless. Windows: Copilot+ renders it unusable. The whole internet is being turned into tie-dye scrambled egg thanks to the stupidity and hubris of AI bros. https://t.co/6CTC9U5DFr

— Gareth Dennis (@GarethDennis) May 23, 2024

maybe one day you can just think about it real hard

― Humanitarian Pause (Tracer Hand), Tuesday, 21 May 2024 14:49 (two days ago) bookmarkflaglink

time to volunteer as a test-subject at neuralink!

Where are you seeing ai generated comedy skits?

― m0stly clean (Slowsquatch), Thursday, May 23, 2024 1:00 AM (six hours ago) bookmarkflaglink

i'm not, just saying on podcasts I'm seeing graphics and songs get outsourced to AI sometimes which is fine in service of a dumb joke but it always sorta takes me out of it. maybe its just me though. I guess I'm fine with some of this stuff just don't want it to become bigger than it is.

I was (still am) pretty interested in media literacy in the early 2000s, like I thought with the rise of Wikipedia and online news articles being able to link to their sources so you could read more/check they were representing it accurately (particularly science journalism but hoped it would filter through), I guess I naively hoped that with everyone finally realising that just because it's on your computer screen doesn't mean it wasn't made up by some random person, your average person might pause for a second before taking something at face value because it's "in print" somewhere.


I sort of expected that we'd be forced to get better at filtering out junk from legit content through necessity. Obviously you'd still have loads of people manipulated by YouTube videos or ranting Twitter feeds but twas ever thus.

and the onslaught of AI generated gobsmackingly WRONG content (I got Google gemini to do a very basic maths question and it just... kept lying to me about the numbers I'd given it) makes me realise, nope, belief in human-sounding text strings is being prioritised at all costs by the uber-tech guys and it totally depresses me.

Seems the origin of the Google AI’s conclusion was an 11 year old Reddit post by the eminent scholar, fucksmith. https://t.co/fG8i5ZlWtl pic.twitter.com/0ijXRqA16y

— Kurt Opsahl @k✧✧✧@ms✧✧✧.soc✧✧✧ (@kurtopsahl) May 23, 2024

https://udm14.com/

frogbs longer post otm

https://www.theguardian.com/sport/article/2024/may/23/michael-schumacher-family-win-legal-case-against-publisher-over-fake-ai-interview