Amazon founder Jeff Bezos believes that artificial intelligence is going to lead to unprecedented productivity gains which could result in cheaper food, housing, and two income households deciding that they no longer need two incomes. Internally, Amazon employees mock the company’s AI tools, refer to its output as “slop,” and joke about the company’s failed attempt to motivate employees to use AI tools effectively.

The memes are yet another example of the contrast between what AI companies say in public about its potential power and benefit versus the reality of how the people who help create these AI tools use and criticize them internally. Amazon employees told me about these memes after they saw my story last week about Google employees also internally sharing memes critical of Google’s AI tools.

“Now I have everything I need,” says the text over an image of a jet taking off in one meme posted by an Amazon employee. The jet is edited to carry the purple ghost logo for Kiro, Amazon’s AI-powered coding tool. “Narrator: He did not have everything he needed,” says the text over an image of a bunch of people left behind on the tarmac. I've recreated all the memes rather than share screenshots from the Slack channel in order to protect sources.

you are viewing a single comment's thread
view the rest of the comments

[–] smeg@feddit.uk 32 points 1 day ago (2 children)

The problem (well, one of the many problems) is that "AI" is a meaningless buzzword slapped on everything by marketing drones. I'm sure there are some real advances that could be made by what five years ago we'd have called "machine learning", but the chatbots that everyone's currently obsessing over don't have anything to do with that.

[–] Quetzalcutlass@lemmy.world 12 points 1 day ago* (last edited 1 day ago) (1 children)

I miss when machine learning was getting big. Communities were sprouting everywhere and making so many interesting demonstrations using it, and I mean genuinely interesting, not tech bro delusions. These were AI that could actually learn and improve themselves (albeit over thousands to millions of randomized iterations; they were still dumb as rocks compared to even the simplest animals), and stuff like genetic algorithms could brute-force discoveries that humans hadn't found even after decades of trying (for example, I believe a major hurdle in modeling protein folding was finally solved using these types of AI).

It was especially cool in the video game space. Hobbyists were doing crazy shit like getting an AI to reliably play a racing game while balancing vertically on the tip of the car the entire time, or setting up complicated mazes only for the AI to figure out how to cheat by launching onto the walls through physics engine exploits. AIs were making novel discoveries rather than just mimicking and piggybacking off of humanity like LLMs do.

[–] vrek@programming.dev 8 points 23 hours ago (1 children)

You would probably be interested in TAS. It's tool assisted speed runs, basically how fast can you setup a bot to beat a game. Celeste is especially impressive to watch, chaining frame perfect tricks back to back to back for the entire game.

[–] Quetzalcutlass@lemmy.world 1 points 4 hours ago* (last edited 4 hours ago) (1 children)

I love TAS! I used to follow a bunch of SNES TASsers back in the day; haven't really kept up with the more modern runs outside of the occasional GDQ event. There was a culture of "swag" that I enjoyed, where if something didn't cost any time then the TAS was encouraged to show off with it, or chain-spam cool stunts during waiting periods.

I kinda wish they'd stop focusing so much on arbitrary code execution though. ACE setups are interesting to explain but tedious to watch, and the payoff is whatever payload the TAS authors write, not game content. It's basically showing off a game mod, but since you're writing assembly code in the least user-friendly manner imaginable, most (with a few rare, incredible exceptions) are simple skips to the credits so the video is just the bot running around performing random actions until the game suddenly ends.

[–] vrek@programming.dev 2 points 4 hours ago

True, Ace is cool from a technical wow that's possible perspective but I much prefer the traditional chain running cook tricks type of runs.

[–] vrek@programming.dev 4 points 23 hours ago

One of the coolest examples I saw a few years ago was for an automated optical inspection. You would take a picture and identify all the critical features and then manually check like 50 parts, varied between good and bad and marking why the bad ones were bad. Afterwards just throw a part under the camera and it could accurately tell good from bad with about 99% accuracy(biased towards false failures). Granted it was about 35k just for the camera and software but still really impressive.

It could locate the part and rotate the image with no fixturing and each inspection was between 10-90 milliseconds depending on what features were being inspected.