My personal mental image for AI agents is Mr Meeseeks.
Per Wikipedia:
> Meeseeks are a powder-blue-skinned species of humanoids (each of whom is named "Mr. Meeseeks") who are created to serve a single purpose which they will go to any length to fulfill. Each brought to life by a "Meeseeks Box", they typically live for no more than a few hours in a constant state of pain, vanishing upon completing their assigned task so as to end their own existence and thereby end their suffering; as such, the longer an individual Meeseeks remains alive, the more insane and unhinged they become.
I'm glad I'm not alone in this. It really starts to feel like this when the agent is rushing to get an answer and keeps shrinking the scope until it can resolve it quickly.
Any idea on who is behind this group? It has an extensive list of institutional partners mostly based in Europe so it must have gained traction somehow. And yet the two key contacts listed on the websites do not seem to have extensive experience in this field. One is a student and the other's main achievement seems to be founding another AI advocacy group.
My gut feeling is this is just another UK based quango — non governmental in name only and exists mostly to push elite agenda.
>These AI images also add to the public mistrust of AI, a growing problem for innovation in a field that is sometimes seen as biased, opaque and extractive.
Oh my, how would anyone ever have gotten that impression?!
And a narrative that makes AI seem friendly and helpful.
These CEOs who keep lying about AI taking jobs need to be silenced with facts about AI impact on jobs (mostly neglible, oddly related to the ROI problem. If AI = job replacement that might be a kind of ROI, right ?)
And the hysterics around "needing to stop AI" because "its moving too fast" wut? Totally fake.
AI can't provide valid formatted XML in some cases, gets caught in useless loops in others, etc.
That only works if the mascot is people’s primary association with the brand, and they have some initially positive experiences with the brand. Most people see AI as suspicious, if not outright sinister. Trying to put a cute fuzzy face on something like that makes it seem more suspicious ands sinister, not less.
A few have been attempted. Mostly by Microsoft and subsidiaries, from GitHub to Mojang. None have caught on. Closest is GitHub's logo which is getting increasingly animated. But dedicated mascots a la Microsoft's Mico have gone absolutely nowhere.
Poor Albert Einstein. He donated perpetual rights to his name and likeness to the Hebrew University of Jerusalem, which promptly used them as a cash cow.
I think a single monolithic mascot for all AI has even worse chances than a Clippy ripoff. People will be as cold towards it as ever, and companies won't want to contaminate their image with something that isn't theirs and that they don't have control over.
I agree that existing images have absolutely terrible ways of bringing in baggage about how these machines work or could work.
The problem is that anything I can think of that would give a true represenation is either super abstract or super detailed and difficult for layperson to understand. Im thinking representations of neural network weights between layers, softmax functions etc.
The issue is that so many ways of doing it here are just bad. Imagery of screens, computer chips, or servers? Normal software runs on chips, but these are qualitatively different to normal software, hence the need for this whole thing. Same goes for images of ones and zeroes, or computer code (a particularly bad offender, given that one major thing that makes them so strange is that they aren't authored or told how to behave in any comprehensible sense)
Images of robots conjure thoughts of magical robots from sci-fi movies, which typically don't remotely work like llms (movie robots have a strong focus on logic and maths, where LLMs are creatures of words and feelings and philosophy)
Brains and people-related imagery fail by bringing ideas of humanity and artifical minds, where these are incredibly complex prediction machines. Some argue perhaps rightly whether humans ourselves are prediction machines, but I think its important not to mix the two ideas because the failure modes of AI are so fundamentally alien and new, and thoughts that these things are the same as smart humans is where problems around oversight and trust get really worrying (confidently speaking with perfect grammar and diction about complicated topics but being confused about basic physics of everyday objects, being easily persuadable to just about any viewpoint and will happily drop everything they believe and admit they were wrong dozens of times back and forth without any problems, maintaining cheerful engagement the entire time)
The ones on the site here don't do much better than the news articles to be honest. A great many of them are art pieces clearly made from some place of dissatisfaction with ai, which isnt really a representation of ai itself. many of them straight up contain the word chatgpt or the openai logo, which is a huge cop-out of what this site is even supposed to be about. Some specific ones I don't think are great:
Thank you... for making the Internet worse, or even more sorrowful ... empty, valueless, worthless, untrue, unaccountable, irresponsible, unconfident , unsafe, insecure, and effortless place... also ridiculing actual artists, professionals, and just people, nature, history, the universe...
I am sorry, but I have absolutely no idea how these people who create such projects and services sleep peaceful at night...
For me it's that there's less and less soul left in the internet. That trend started before LLMs, but now it also starts appearing in the real world as well. Seeing an AI-generated ad picture in the mall was kind of a wakeup call. So weird seeing the face of a girl that was generated.
What about books, art, music? Real artists and creativity get priced out of our societies.
There are plenty of books, art, and music coming out, none of it AI-generated. What are you talking about?
As you said, the internet had slop before AI. It has been that way since SEO and social media started to dictate what people consume. Remember Elsagate? Remember the IRA bot farms? Remember NFT monkey PNGs?
Most books were trash before the internet, and most paintings have been amateur since the dawn of painting. We just don't remember those examples because we only preserve the best from each era.
Even if 99% of books published in the next 50 years turn out to be AI-generated slop (which is probably what's going to happen, to be honest), that 1% would still be more than you could possibly read in your lifetime. That goes for music, video games, and visual art too.
Let the AI bros make their slop. They are of no consequence to actual culture in the long term.
My personal mental image for AI agents is Mr Meeseeks.
Per Wikipedia:
> Meeseeks are a powder-blue-skinned species of humanoids (each of whom is named "Mr. Meeseeks") who are created to serve a single purpose which they will go to any length to fulfill. Each brought to life by a "Meeseeks Box", they typically live for no more than a few hours in a constant state of pain, vanishing upon completing their assigned task so as to end their own existence and thereby end their suffering; as such, the longer an individual Meeseeks remains alive, the more insane and unhinged they become.
https://en.wikipedia.org/wiki/Mr._Meeseeks
I'm glad I'm not alone in this. It really starts to feel like this when the agent is rushing to get an answer and keeps shrinking the scope until it can resolve it quickly.
I don’t want to think about my growing collection of prompts that never got to fulfil their purpose.
That... seems too accurate.
Caaaan do!
Look at me!
Any idea on who is behind this group? It has an extensive list of institutional partners mostly based in Europe so it must have gained traction somehow. And yet the two key contacts listed on the websites do not seem to have extensive experience in this field. One is a student and the other's main achievement seems to be founding another AI advocacy group.
My gut feeling is this is just another UK based quango — non governmental in name only and exists mostly to push elite agenda.
We already have the best image of AI, clippy.
Nah, AI should definitely be illustrated as robots. Because they're robots. Adding framing, bearings, servos is just an I/O issue.
Where is the goose imagery?
smh goose was not valued...
>These AI images also add to the public mistrust of AI, a growing problem for innovation in a field that is sometimes seen as biased, opaque and extractive.
Oh my, how would anyone ever have gotten that impression?!
Clearly it's all these stock photos of galaxy brains!
I recommend the Japanese mascot approach.
A round smiling AI mascot.
And a narrative that makes AI seem friendly and helpful.
These CEOs who keep lying about AI taking jobs need to be silenced with facts about AI impact on jobs (mostly neglible, oddly related to the ROI problem. If AI = job replacement that might be a kind of ROI, right ?)
And the hysterics around "needing to stop AI" because "its moving too fast" wut? Totally fake.
AI can't provide valid formatted XML in some cases, gets caught in useless loops in others, etc.
That only works if the mascot is people’s primary association with the brand, and they have some initially positive experiences with the brand. Most people see AI as suspicious, if not outright sinister. Trying to put a cute fuzzy face on something like that makes it seem more suspicious ands sinister, not less.
As long as we're taking the Japanese approach, why not create a yandere wAIfu? People will love her irrespective of her being suspicious and sinister.
That would definitely be a hit on particular forums.
A few have been attempted. Mostly by Microsoft and subsidiaries, from GitHub to Mojang. None have caught on. Closest is GitHub's logo which is getting increasingly animated. But dedicated mascots a la Microsoft's Mico have gone absolutely nowhere.
How can you forget Salesforce -- who has way too damn many mascots!?
https://thespotforpardot.com/wp-content/uploads/2017/11/Saas...
https://thespotforpardot.com/wp-content/uploads/2017/11/Astr...
https://thespotforpardot.com/wp-content/uploads/2017/11/Eins...
Poor Albert Einstein. He donated perpetual rights to his name and likeness to the Hebrew University of Jerusalem, which promptly used them as a cash cow.
A mascot for Microsoft copilot - thanks I didn't know about Mico - is totally different.
That's just a Clippy variant.
This AI mascot would represent "all AI" not Microsoft Copilot.
And there needs to be a consistent counter to the "were laying off 10% of our staff to pivot to AI" lies.
Yes, they're laying off staff, but no this has nothing to do with AI.
They lay off staff to appease stock holders.
"We over hired in the pandemic" is a hoot. So we're firing the executives who did the planning for that? Nope. They stay.
AI is a technology whose defamation by CEOs will make its progress slower and more difficult, simply so the CEIs have some cover story for layoffs.
I think a single monolithic mascot for all AI has even worse chances than a Clippy ripoff. People will be as cold towards it as ever, and companies won't want to contaminate their image with something that isn't theirs and that they don't have control over.
This guy is the opposite of the fear mongers, don't listen to "everything is fine and normal!"
How about a sarlacc, a garbage dump, a factory pipe spilling toxic waste, etc.?
I agree that existing images have absolutely terrible ways of bringing in baggage about how these machines work or could work.
The problem is that anything I can think of that would give a true represenation is either super abstract or super detailed and difficult for layperson to understand. Im thinking representations of neural network weights between layers, softmax functions etc.
The issue is that so many ways of doing it here are just bad. Imagery of screens, computer chips, or servers? Normal software runs on chips, but these are qualitatively different to normal software, hence the need for this whole thing. Same goes for images of ones and zeroes, or computer code (a particularly bad offender, given that one major thing that makes them so strange is that they aren't authored or told how to behave in any comprehensible sense)
Images of robots conjure thoughts of magical robots from sci-fi movies, which typically don't remotely work like llms (movie robots have a strong focus on logic and maths, where LLMs are creatures of words and feelings and philosophy)
Brains and people-related imagery fail by bringing ideas of humanity and artifical minds, where these are incredibly complex prediction machines. Some argue perhaps rightly whether humans ourselves are prediction machines, but I think its important not to mix the two ideas because the failure modes of AI are so fundamentally alien and new, and thoughts that these things are the same as smart humans is where problems around oversight and trust get really worrying (confidently speaking with perfect grammar and diction about complicated topics but being confused about basic physics of everyday objects, being easily persuadable to just about any viewpoint and will happily drop everything they believe and admit they were wrong dozens of times back and forth without any problems, maintaining cheerful engagement the entire time)
The ones on the site here don't do much better than the news articles to be honest. A great many of them are art pieces clearly made from some place of dissatisfaction with ai, which isnt really a representation of ai itself. many of them straight up contain the word chatgpt or the openai logo, which is a huge cop-out of what this site is even supposed to be about. Some specific ones I don't think are great:
A person in glass-like squares? https://betterimagesofai.org/images?artist=AlanWarburton&tit...
some people pointing at a screen? https://betterimagesofai.org/images?artist=NachoKamenov&titl...
I think this one is pretty good, though not really related to llms or generative ai https://betterimagesofai.org/images?artist=AlyssaChen&title=...
Yeah, but this sounds like hard backpedaling on the ridiculous sales pitches.
"There are lies, damned lies, and statistics"
We've already aroused far too many idiots who decided to hang their hat on a flimsy epistemology. They will never be convinced.
[dead]
Thank you... for making the Internet worse, or even more sorrowful ... empty, valueless, worthless, untrue, unaccountable, irresponsible, unconfident , unsafe, insecure, and effortless place... also ridiculing actual artists, professionals, and just people, nature, history, the universe...
I am sorry, but I have absolutely no idea how these people who create such projects and services sleep peaceful at night...
Why do you think AI is making things worse? Please explain.
For me it's that there's less and less soul left in the internet. That trend started before LLMs, but now it also starts appearing in the real world as well. Seeing an AI-generated ad picture in the mall was kind of a wakeup call. So weird seeing the face of a girl that was generated. What about books, art, music? Real artists and creativity get priced out of our societies.
There are plenty of books, art, and music coming out, none of it AI-generated. What are you talking about?
As you said, the internet had slop before AI. It has been that way since SEO and social media started to dictate what people consume. Remember Elsagate? Remember the IRA bot farms? Remember NFT monkey PNGs?
Most books were trash before the internet, and most paintings have been amateur since the dawn of painting. We just don't remember those examples because we only preserve the best from each era.
Even if 99% of books published in the next 50 years turn out to be AI-generated slop (which is probably what's going to happen, to be honest), that 1% would still be more than you could possibly read in your lifetime. That goes for music, video games, and visual art too.
Let the AI bros make their slop. They are of no consequence to actual culture in the long term.