fr GPT-4 has already surpassed the vast majority of low-wage human data workers long ago
It can actually follow basic instructions, while the majority of low-wage humans just outright refuse, constantly “forget”, play dumb, are intentionally trying to scam you at every step.
Should have cut the sh*t and just followed the instructions bro.
Your nonsense won’t be missed.
Low-wage data workers btfo.
It can actually follow basic instructions, while the majority of low-wage humans just outright refuse, constantly “forget”, play dumb, are intentionally trying to scam you at every step.
Should have cut the sh*t and just followed the instructions bro.
Your nonsense won’t be missed.
Low-wage data workers btfo.
👍16❤5👏3😱3🤡2💯2🤣2
Throwback the Katie Bouman & first black hole image fraud
(1) Media gives sole credit for creation of the AI enabling the first black hole image to some young woman
(2) But not just any woman, one who openly brags about being the least unqualified, most incompetent you could imagine
(3) Reality, she did almost nothing, and the vast majority of the people who did the work were guys, who look as you’d expect
(4) Oh, and by the way, turns out the black hole image itself was also a total fraud too.
Frauds all the way down, doubling down when caught.
Only going to get worse.
(1) Media gives sole credit for creation of the AI enabling the first black hole image to some young woman
(2) But not just any woman, one who openly brags about being the least unqualified, most incompetent you could imagine
(3) Reality, she did almost nothing, and the vast majority of the people who did the work were guys, who look as you’d expect
(4) Oh, and by the way, turns out the black hole image itself was also a total fraud too.
Frauds all the way down, doubling down when caught.
Only going to get worse.
👍19👏13😁9❤4🤬4
🚨🧠Mind Reading Achieved: Facebook researchers present method for decoding speech from brain waves
“Our model predicts the correct segment, out of more than 1,000 possibilities, with a top-10 accuracy up to 70.7% on average across MEG subjects.”
“For more than half of samples, the true audio segment is ranked first or second in the decoders’
predictions. Interestingly, these performances can reach high top-1 accuracy in the best performing subjects: e.g. above 80% top-1 accuracy in the best participant of the dataset of Gwilliams et al.”
“For comparison, a model that predicts a uniform distribution over the vocabulary
(‘random model’) only achieves less than 1% top-10 accuracy on the same MEG datasets.”
Paper
“Our model predicts the correct segment, out of more than 1,000 possibilities, with a top-10 accuracy up to 70.7% on average across MEG subjects.”
“For more than half of samples, the true audio segment is ranked first or second in the decoders’
predictions. Interestingly, these performances can reach high top-1 accuracy in the best performing subjects: e.g. above 80% top-1 accuracy in the best participant of the dataset of Gwilliams et al.”
“For comparison, a model that predicts a uniform distribution over the vocabulary
(‘random model’) only achieves less than 1% top-10 accuracy on the same MEG datasets.”
Paper
❤15🤯8😱6
GPTFUZZER: Red Teaming Large Language Models with Auto-Generated Jailbreak Prompts
"GPTFuzz automates the generation of jailbreak templates for red-teaming LLMs"
"results indicate that GPTFuzz consistently produces jailbreak templates with a high success rate, surpassing human-crafted templates"
"GPTFuzz achieves over 90% attack success rates against ChatGPT and Llama-2"
Paper
Github
"GPTFuzz automates the generation of jailbreak templates for red-teaming LLMs"
"results indicate that GPTFuzz consistently produces jailbreak templates with a high success rate, surpassing human-crafted templates"
"GPTFuzz achieves over 90% attack success rates against ChatGPT and Llama-2"
Paper
Github
👍43😈13🫡12🔥9❤8