No one who has even a vague understanding of present day ML models should not even entertain the idea that they are sentient, or thinking, or anything like it.
AI models don’t resynthesize their training data. They use their training data to determine parameters which enable them to predict a response to an input.
Consider a simple model (too simple to be called AI but really the underlying concepts are very similar) - a linear regression. In linear regression we produce a model which follows a straight line through the “middle” of our training data. We can then use this to predict values outside the range of the original data - albeit will less certainty about the likely error.
In the same way, an LLM can give answers to questions that were never asked in its training data - it’s not taking that data and shuffling it around, it’s synthesising an answer by predicting tokens. Also similarly, it does this less well the further outside the training data you go. Feed them the right gibberish and it doesn’t know how to respond. ChatGPT is very good at dealing with nonsense, but if you’ve ever worked with simpler LLMs you’ll know that typos can throw them off notably… They still respond OK, but things get weirder as they go.
Now it’s certainly true that (at least some) models were trained on CSAM, but it’s also definitely possible that a model that wasn’t could still produce sexual content featuring children. It’s training set need only contain enough disparate elements for it to correctly predict what the prompt is asking for. For example, if the training set contained images of children it will “know” what children look like, and if it contains pornography it will “know” what pornography looks like - conceivably it could mix these two together to produce generated CSAM. It will probably look odd, if I had to guess? Like LLMs struggling with typos, and regression models being unreliable outside their training range, image generation of something totally outside the training set is going to be a bit weird, but it will still work.
None of this is to defend generating AI CSAM, to be clear, just to say that it is possible to generate things that a model hasn’t “seen”.
This kind of thing is where ML/AI can really shine. Data which is consistent, regular, where there are deep, hidden patterns that are not easy for humans to recognise. It’s very interesting that this came from an LLM, there are so many interesting and surprising applications of them that go beyond asking ChatGPT to write python for you.
I saw a talk at a conference about a ML model designed to write chemical synthesis instructions. We have tons of systems able to predict synthetic pathways and the like, but not necessarily to predict the best solvents, extraction techniques, etc… and an LLM might provide a way to get there, which I think is amazing.
Stealing from an individual is deplorable. I can understand why someone might want to respond aggressively (although to be clear I still don’t think it’s justified) if someone steals medication from an old lady… But from a shop?
I don’t think there’s any shortage of Japanese samurai protagonists in games. From a representation standpoint I don’t think there’s any issue… And Yasuke is an interesting character that’s worth exploring. There’s enough mystery to his story that he fits perfectly into an Assassin’s Creed game imo.
Probably because dogs only live like 10-15 years. Their little gnashers don’t have to last quite so long… Then again they also like to chew up wood n shit so I suppose they go through a lot.
This falls into a common trap. Because we cannot succinctly define a salad in one sentence we decide that it cannot be defined at all. This argument effectively reducto ad absurdums itself by coming to the conclusion that all foods are salad.
If we start from a position where we discount nothing from being a salad, and we have only salads (and soup, seemingly) to base our analysis on, how can we ever identify the boundaries of salad? The whole argument is based on the flawed premise that anything could be a salad.
I don’t need just a bit of free time one time. I need more free time always! The individual blocks of free time one tends to get end up being useless because you’re busy recovering from the rest of the time.
That’s a great resource, but good god I can’t imagine a worse environment for me to work in than one where I’m constantly being questioned by children.
adherence to the rule is the 4th law of robotics (lemmy.blahaj.zone)
“CSAM generated by AI is still CSAM,” DOJ says after rare arrest (arstechnica.com)
I tried to give my old Wii and games to my 11 year old son, he refused them. (lemmy.world)
Oh well. I’ll just leave it at that.
People want 'dumbphones'. Will companies make them? (www.bbc.com)
New AI generates CRISPR proteins unlike any seen in nature (bigthink.com)
me_irl (lemmy.world)
Woman Stuck in Tesla For 40 Minutes With 115 Degrees Temperature During Vehicle Update (www.ibtimes.co.uk)
Sainsbury's staff beat up shoplifter after dragging him into the back room (metro.co.uk)
Unexpectedly Wholesome (i.imgur.com)
Uh oh the new AC character is woke (lemmy.world)
Nvm that Yasuke was a real person.
Images leak of Valve's next game, and it's an Overwatch-style hero shooter (www.eurogamer.net)
Though the way the leaker describes it makes it sound more like a 3rd person MOBA than a hero shooter....
Unforgivable (lemmy.world)
It's not fair. (lemmy.world)
Indiana judge rules tacos, burritos are sandwiches (www.wishtv.com)
I just turned into a puddle of goo. Thanks for asking. (lemmy.dbzer0.com)
50 million rendered polygons vs one spicy 4.2MB boi (lemmy.world)
You REPLENISH (cdn-cf-east.streamable.com)
Then and Now (i.imgur.com)
Stormy Daniels wore a bulletproof vest to Trump’s hush money trial, attorney says (www.independent.co.uk)
Earlier this year, porn star said she accepted hush money so Donald Trump could not have her killed...
Lemmy: Linux is my boyfriend (i.ibb.co)
This would be the time to plan something without Australia (lemmy.world)