MSNBC host Lawrence O’Donnell laid into news outlets on his show Thursday night, comparing their collective failure to call out Donald Trump’s falsehoods to 2016.
“It was 2016 all over again, today,” he told his viewers. “Donald Trump spoke at his home in Florida for over an hour and all the major cable news networks, including this one, carried it live. Just like they all did repeatedly in 2016. It would be hard to find a sentence that Donald Trump said today that did not include at least one lie.”
O’Donnell acknowledged that some networks attempted to fact-check the MAGA leader, but only after he finished his speech.
“That of course is way too late and utterly useless,” he said. “No network tried to fact-check every lie Donald Trump told.”
Don’t these networks have AI? Voice to text from the feed, send it through ChatGPT to summarize fact check, quickly verify the stats and add the result to graphics overlay.
I’m surprised a streamer hasn’t done this yet.
https://www.reddit.com/r/LocalLLaMA/comments/173jqpe/realtime_fallacy_detection_in_political_debates/ has something a bit similar
Large language models (chatgpt etc, basically any of the AIs you’ve seen recent headlines about) aren’t especially good at distinguishing true from false claims. It’s one of the biggest weaknesses that AI researchers are actively trying to find solutions for.
Great idea, validate bullshit with the bullshit generator.
I did a quick sampling with copilot, it seems to lie less than trump. Good enough for me!
you could tell an LLM to specifically lie and it would probably lie less than him. everything would be trustworthy if compared to the bullshit man
Copilot appears to even refuses to tell me that these are true, either when I encourage it by telling me to “correct thr statement” on the true statement, or when I explicitly tell it to tell me otherwise.