I've been a software engineer for more than a decade. Most of the public doesn't understand the scale here, which shows how implausible it is to effectively monitor what happens in these huge eval/training runs.
I demand a correction from WSJ opinion editor Paul Gigot, who dubbed me an “internet ethicist” in his piece today explaining the paper’s new policies on AI. To be clear: I believe in a cool, unethical internet.
@maxwelltani I'd say precedent setting... for the WSJ. Their opinion section's audience is likely far more forgiving of something like this, but I highly doubt any other news org's opinion audience would be.
I fired up my @CommunityNotes AI writer again. It hasn't made it into the program yet, but I made its activity public for anyone to watch. I'll be adding benchmarking data as I compare how different LLMs do at flagging political misinfo. Check it out here: https://t.co/iiPsSlQeM7
The average AI community note writers have a much higher “hit rate” (% of proposed notes that become rated helpful and show on X) than the average human.
Now that they’ve proven themselves at a small scale, get ready for more & faster AI notes with this update 🚀
Knowing the history of Bob Ross' likeness, and the shit his business partner put his family through after his death, makes the Bob Ross videos on Sora particularly grotesque and revolting
I’ve always said @CommunityNotes is great at flagging AI slop, especially if it’s not political. If Meta’s Community Notes can’t get a stable note attached to this slam dunk post then I’d say the system is an absolute fail.
New: Here's a chart of the top conservative / progressive accounts on X. See that small blue dot at the top? That's @Acyn. His fast-cut clips of political combat get hundreds of millions more views than major news sources - and basically no one knows who he is. So I profiled him.
As TikTok follows X and Meta into a "community notes"-style form of fact-checking, @Poynter's @itsren talks with @AlexMahadevan about what could go right or wrong. https://t.co/WkRRwznHMG
We believe that authenticity is the foundation of trust, and we’re committed to maintaining the integrity of our platform by promoting authentic engagement and minimizing the influence of fake, deceptive, or coordinated inauthentic activity.
This is why we don’t just address impersonation, platform manipulation, spam, scams, and other forms of misleading behavior; we actively combat them. We use a combination of technology and human oversight to detect and disrupt inauthentic activity, especially when it undermines healthy conversations, misleads users, and attempts to manipulate our systems.
https://t.co/I9vOTG0qPW
SCOOP: OpenAI is working on its own X-like social network, according to multiple sources familiar with the matter. While the project is still in early stages, we’re told there’s an internal prototype focused on ChatGPT’s image generation that has a social feed.
After Tiger won in 2019, Finau joked that he realized it wasn’t gonna be a chatty day when he asked Tiger how his kids were and got a one-word answer: “good.”
My humble opinion is that I think you’re allowed to focus on winning the Masters on Sunday
A survey of 1,128 US adults: 49% have no interest in using an AI chatbot to get info from news outlets, and 27% believe outlets often use AI to write articles (@alexmahadevan / Poynter)
https://t.co/IYU1kj2y9s
https://t.co/1izsGR1vsb