Google says its flagship artificial intelligence chatbot, Gemini, has been inundated by “commercially motivated” actors who are trying to clone it by repeatedly prompting it, sometimes with thousands of different queries — including one campaign that prompted Gemini more than 100,000 times.
pull down to refresh
more proof that AI is going to be as profitable as the local electric utility company, not a world-beating AGI singularity and trillions in profits
my SUSPICION is this is how the chinese chatbots that keep up with state of the art are made.
Only more stealthily. And towards OpenAI/Claude/Gemini/etc simultaneously
They all do it: #1064935
Alright, maybe you're right and they all steal from each other to get even more training data.... 🤨
But the chinese models that come out of nowhere and instantly are state of the art - fishy fishy fishy. At least for the wedtern models we know that they crawl the web and scan paper books like crazy too. A mix of all of these datasources
They're all the same. The "western" and "eastern" divide is something made up by politicians to vote for them and to make you support their statist narratives.
Everyone steals everything from everyone else in the cutthroat tech frontier. That's been going on since forever, especially in software. I've been a founding engineer in a bunch of startups and for every one that was in a competitive space, mostly post-launch, do you get the CEO tell you yo we need to clone
<insert competitor feature>now. Only if you're the frontrunner (and there is none in the LLM model business at this moment) do you get to develop without looking at others - because there are no others.YouTube "invented" shorts and didn't copy it from TikTok? And did TikTok invent it or did they make a successful product out of something that failed on Twitter? Haha. It's all C&P.
For a longer time now, we just say yo Claude/GPT copy
<this shit>for me. From a year ago:source
The frontier guys invented this method and they actively hire their powerusers that are successful in doing things like this.
Stealing ideas from other competitors isn’t the same as stealing implementations though?
But that's not really what is alleged though? You just ask 100k questions to Gemini 3 and then you make a dataset for your own RL out of it and use GPT as a judge (though imho llama3.3 is still a really good cheap judge too, due to great instruct)
Fully automated RL set by spending tokens on Gemini... This is what "AI building AI" means?
The AI wars?
Maybe it's time for a per-prompt captcha. Gotta protect the LLM's precious bodily tokens.