R to @TheAI_Studio: 3. @CaliberAI - AI-powered brand protection from risky content. - Prevents defamation, hate speech. - Suitable for brand management. https://caliberai.net/

R to @moreisdifferent: that's not how defamation law works 🙄 There's this thing called recklessness. Anyway, as I've said before, myself and the @CaliberAI team are more than happy to show @ylecun how to build and effectively deploy a defamation classifier 🤭 h/t @Alanapierreid

R to @GaryMarcus: One for the @CaliberAI team

R to @CharlieBeckett: agreed, though it's equally important to keep pace with regulatory developments, like the @FTC's focus on the kind of defamatory mitigation tech we've built at @CaliberAI 🙋🏻‍♂️ https://arstechnica.com/information-technology/2023/07/chasing-defamatory-hallucinations-ftc-opens-investigation-into-openai/

In today's #AIIreland podcast, Neil Brady, Co-Founder/CEO and Paul Watson, CTO at @CaliberAI discuss use cases of how #AI can track #defamation and #hatefulspeech, how they’re utilizing #NLP and #LargeLanguageModels & much more 🔊Listen now 🔽🔽🔽 https://tinyurl.com/25szpnrp

1/3 Flattered to read @madhumita29's report on how @OpenAI are emulating an approach to synthetic data that @CaliberAI has been perfecting since 2019 🙏🏻

As Co-Founder of the only company around that has built a classifier to mitigate defamation risks from LLMs, myself and the @CaliberAI team would be more than happy to chat @linakhanFTC 🙋🏻‍♂️

R to @neilpbrady: Thank you @neilpbrady @BazWalsh85 🙏🏽

R to @neilpbrady: It is indeed all about the 'expertise'. @EricTopol's summary on medical AI (point 4) published yesterday, and import of fine tuning, also worth reading in this context 👇🏻 If Greg ever needs a hand, at @CaliberAI we're always happy to help 🙋🏻‍♂️ https://erictopol.substack.com/p/medical-ai-is-on-a-tear

R to @BazWalsh85: myself and @CaliberAI CTO @pwatsonCaliber know a thing or several between it, time at @Storyful and at @guardian. Feel free to DM if there's something I can help with.

How accurate are AI-text detection classifiers? 42% accurate. @tomgoldsteincs is right when he says "make it so that the use of AI is not the problem." By contrast, @CaliberAI's defamatory classifier gets 92% recall and 84% precision on our gold standard data. #DiligentPublishing

R to @DrTechlash: can I nominate myself and the @CaliberAI team!? Might warrant a new sub-column under the anti-panic section. Maybe 'Journalist built AI to improve public discourse', or something such 🤔

@HMcEntee ought to read @CaliberAI's FAQ re how we define 'harm', 'defamation' and the interrelationship between the two. Mapping these to NLP moderation tools to nudge better discourse is key but, so far, I see no awareness of this on government's part 🫤 https://www.caliberai.net/faq/what-do-we-mean-by-the-terms-defamatory-and-harmful

I find the phrase "accepting such a law", in this piece by @AndrewSolender & @ashleyrgold, on @HawleyMO & @SenBlumenthal's No Section 230 Immunity for AI Act, striking (though that could be because, at @CaliberAI, we build tech to mitigate defamation) 🤔 https://www.axios.com/2023/06/14/hawley-blumenthal-bill-section-230-ai

1/2 Solid comparison by @barrymward on #rtept, between a Garda using facial rec tech to avoid "watching hundreds of hours of CCTV footage", and using Control+F to keyword filter. This is the same basic value proposition as @CaliberAI's defamation filter, as submitted to

Neil Brady from ⁦@CaliberAI writing in @sundaybusinesspost ⁦@tcddublin⁩ spinout ⁦@Entirl⁩ ⁦@KTIconnect⁩ https://www.businesspost.ie/news-focus/ai-critical-thinking-and-upskilling-crucial-as-journalism-faces-a-new-turmoil/

Nice to see @TateRyMo note how "the newest large language models are much better at interpreting text than previous AI systems." Pretty much the point former @CaliberAI Comp Linguist @Alanapierreid was making in 2021 👉 https://caliberai.net/posts/fit-for-purpose-machine-learning-one-ais-food-is-anothers-poison

R to @neilpbrady: Yes. She’s a lecturer and psychotherapist. You would imagine she should know better . I’m looking at their past timelines . Jeeees Louise 😬

R to @TRF: cc @neilpbrady @CaliberAI