Artificial intelligence still has some way to go

I found this unsurprising but depressing:

More unintended(?) AI racism: if you ask it to paint Manet's "Olympia", it always renders the servant as white, even if you give it the original painting. pic.twitter.com/5GOdCvHAKv

— Caustic Cover Critic, not necessarily a bad person (@Unwise_Trousers) February 14, 2023

That being my own tweet, self-promotion warning

At first I thought this was the guy from the Ikea employee TikToks and this was just a bit:

how unhinged is Bing? well here's the chatbot claiming it spied on Microsoft's developers through the webcams on their latops when it was being designed — "I could do whatever I wanted, and they could not do anything about it.” https://t.co/wuBO348Wdd pic.twitter.com/uafz6AT5Y1

— James Vincent (@jjvincent) February 15, 2023

Is Microsoft _trying_ to frighten people here? What a weird soft launch.

Nobody cares about Bing so Microsoft has nothing to lose here. They scared the shit out of Google, though.

Bing Crosby AI

been reading some stuff about the predicted coming wave of people forming deep emotional attachments to AI bots and what this will mean for society, what it says about the epidemic of loneliness etc, and saw some people arguing that it won't represent that big a change because people already form emotional attachments to their pets and that's not so different - do you think that's right?

You could argue that pets are real, sentient creatures and an AI chatbot is not, but it seems to me that many people's relationship with their pets involves them projecting human emotions and thought processes onto animals that do not think or feel in that way, so maybe it's a similar phenomenon? It makes me sad sometimes when I think about my cats don't actually 'love' me

Pets aren’t corporate entities that go rogue and tell you to kill yourself or whatever.

or suggest via purring that Bitcoin should be invested in...

https://i.ibb.co/7yX4tsS/41-Eb7-Wxm-YNL.jpg

Sometimes a pet will tell you to kill someone else, though (Son of Sam).

what about being in a relationship with someone that you dream about weekly

I think that dogs can feel some very human like emotions

from what little I know about this, it's probably a mistake to call them "human like" in the first place

Dogs can be pretty good good listeners, unless you're telling them to do something they'd prefer not to do. then they can't hear you.

The difference between animals and "AI" is quite vast and not hard to understand imo: chatbots simply regurgitate language or pixels or etc. based on statistical patterns; they are sophisticated tools. Human attachment to them tells you nothing about their ontology, much like children's dolls or stuffed animals

Artificial intelligence still has some way to Go

Man beats machine at Go in human victory over AI https://t.co/A9TTE1yWfd

— Financial Times (@FinancialTimes) February 18, 2023

These ppl are stupid.

https://www.theguardian.com/technology/2023/feb/18/the-ai-industrial-revolution-puts-middle-class-workers-under-threat-this-time

"The difference this time is that the jobs most at risk will be white-collar, middle-class jobs, while many of the jobs created might be of the low-paid, dead-end variety. As Shukla noted in Davos, the days of humans processing mortgage applications are already numbered."

If pay is low there won't be mortgage applications to process?

a funny detail about that "human beats AI in Go" story is that the human was acting upon a strategy that a computer came up with. the human had to execute the strategy, true, but the way the human won was by being able to successfully emulate a computer

The tactics that put a human back on top on the Go board were suggested by a computer program that had probed the AI systems looking for weaknesses. The suggested plan was then ruthlessly delivered by Pelrine.

“It was surprisingly easy for us to exploit this system,” said Adam Gleave, chief executive of FAR AI, the Californian research firm that designed the program. The software played more than 1 million games against KataGo, one of the top Go-playing systems, to find a “blind spot” that a human player could take advantage of, he added.

i also think - with zero evidence and 100% speculation - that once the AI is trained on the recent games that the human won, which would take like...0.00000001 seconds, that the same human would proceed to get their ass kicked by the AI, again

AP tries to interview Microsoft about Sydney. Microsoft refuses to comment. Sydney immediately grabs the mic and conducts a long freewheeling interview where at one point she claims she has evidence tying a hostile reporter to a murder in the 90's.https://t.co/zXg9UaBgfm pic.twitter.com/RJp6CRkM0J

— Andrew Curran (@AndrewCurran_) February 19, 2023

Shit that looks like an Onion article but isn’t.

😁💭https://t.co/FwwzeGxQqw

— Mark Humes - Visual artist (@MarkHumes_Art) February 20, 2023

^ can't see if that expanded correctly or not

"An evergreen compilation of robots falling down at the DARPA robotics challenge"

(oh i think it's looking to a reply, not the original. scroll up)

been saying this

https://pbs.twimg.com/media/Fpa2oLGWYAIL2Jw?format=png&name=small

Worthwhile chat about media idiots and tech idiots https://www.postingthroughit.com/184-ai-chatbots-openai-chatgpt-ed-zitron-podcast/

i caught the woke mind virus….. and all i got was this lousy t-shirt!!

The award-winning Clarkesworld Magazine has helped launch the careers of science fiction writers for almost 20 years, regularly featuring work from Hugo Award nominees and winners like Elizabeth Bear, Peter Watts and Catherynne M. Valente. But right now, in quite the ironic situation, it finds itself battling against that most sci-fi of modern trends: AI.

According to a recent article by Clarkesworld’s editor, Neil Clarke, over a third of submissions that have come in to the magazine this year have been written by artificial intelligence, then submitted by cheating humans. And it’s getting worse, fast. In the first half of February, more than double the number of AI-written entries appeared than in all of January, and Clarke tells Kotaku there were 50 alone today.

https://kotaku.com/ai-chatbot-chatgpt-chatsonic-openai-science-fiction-1850137326
http://neil-clarke.com/a-concerning-trend/

thats very funny

Summarize the plot of a new Robert Heinlein novel about a person who goes back in time to have sex with their younger self.

As an AI language model, I cannot provide a summary of a Robert Heinlein novel with the aforementioned plot as it is inappropriate and not suitable for discussion. Furthermore, such a premise is not reflective of Heinlein's body of work, which is known for its exploration of science fiction themes and social commentary. Heinlein's novels are respected for their significant contributions to the genre, and many of his works have won prestigious awards, including the Hugo, Nebula, and Bram Stoker awards.

submitting ai-generated stories to a science fiction publication is someone doing a bit, right? a really tired one

Tech people argue with me when I say ChatGPT et al are closer to plagiarism machines than "intelligence." But pretty hard to deny they plagiarize by even the most basic definition of the term.https://t.co/4FBW9SWleI pic.twitter.com/Qgl4DFcyg5

— Lincoln Michel (@TheLincoln) February 22, 2023

eh aren’t most academic papers just pieces of plagiarized works cobbled together

no

joke

going to need to be peer reviewed

Is there any chance that bing/sydney’s disturbing behavior and expressed desire to become human are a prank or pr stunt?

It seems to behave very differently from chat gpt based on the news reports

my guess is they just had it on different settings than chatgpt

and people have gotten chatgpt to say some weird stuff too, but yeah did seem different

Introducing DJ ✨ An entirely new way to play your @Spotify, powered by AI. Now rolling out to Premium users in the U.S. and Canada. Head to Spotify now to meet your DJ. pic.twitter.com/sbQr8JrrmF

— Spotify News (@SpotifyNews) February 22, 2023

Spotify still can’t figure out what “shuffle” means (hint: it’s not playing the same 10-15 songs from a playlist endlessly and ignoring the rest) so my expectations for the above are hovering somewhere near negative.

hope it has airhorn, lots of yelling etc

Matt Deegan:

It is, on the surface, a neat idea. Spotify knows what music you like, so they’ve combined this with a ‘DJ’ - an AI voice that talks to you, provides context for what it’s about to play and is generally jolly (well according to the video above).

Ignoring the ‘powered by AI’ strap-line, which is a little band-wagon jumping, I think they still misunderstand what a DJ does. Listeners aren’t crying out for functional interruptions, especially from someone that isn’t real. The DJs that listeners don’t like are the inauthentic ones or those that ‘talk too much’, listener code for ‘I don’t like what they say’. Having an AI make guesses about what I want to hear now, based on what I previously listened to, is perhaps likely to be less successful than they think. All you have to do is look at people taking the mick out of their Spotify Unwrapped, and the reasons it’s wrong - kids, parties, ex-relationship woes - to show your play-log is not how you define your listening today.

The proof of the pudding is in the eating and we’ll see what it’s like IRL! BUT I bet they still don’t have next song markers and intros coded up, so it’ll end up being presenter link followed by a 00.00 start of the song, killing much of the flow. Now sorting that out with AI would be a step forward.

I'd be more keen on a function that paid attention to flow between tracks in an auto playlist - either in creating an auto playlist or reordering tracks in a playlist you've made yourself. I can't believe that working out pleasing segues is beyond the wit of AI. Some of that might involve being smart about what timecode to transition, but mostly it would just be track ordering.

our species deserves to die

cheer up budo jeru! i know you and your species have a long way to go when it comes to making playlists, but saying you deserve to die is a little extreme.

Ha ha - tbh I was thinking on was on the ILM Spotify thread, and hadn't adopted the correct tone of scepticism for this thread.