Kate Hartley

Image shows a screenshot of Meta Llama text, which reads: Fact-checking is crucial in mitigating the risk of bias and misinformation in AI. Here are some reasons why: Benefits of Fact-Checking Verifying accuracy: Fact-checking helps verify the accuracy of information used to train AI systems, reducing the risk of perpetuating misinformation. Detecting bias: Fact-checking can detect biases in data, algorithms, or AI-generated content, enabling corrective action. Building trust: Fact-checking helps build trust in AI systems by ensuring that information is reliable and accurate. Improving decision-making: By providing accurate information, fact-checking enables better decision-making and reduces the risk of adverse consequences.

I just asked Meta's AI about the risks of bias in AI, and it tells me it's important to have diverse and representative data and human oversight/review to mitigate risks including erosion of trust. Also that fact-checking is critical in mitigating bias, including human fact-checking. The irony

Bias in AI amplifies our own biases Artificial intelligence (AI) systems tend to take on human biases and amplify them, causing people who use that AI to become more biased themselves, finds a new study by UCL researchers.

Researching AI bias, and came across this research from UCL. Short version: if you look at biased data from AI, you're more likely to take on that bias, and you may use it to train AI, leading to more bias. And repeat. www.ucl.ac.uk/news/2024/de...

We need diversity more than ever, and particularly in managing a crisis. Bringing together different experiences and thinking leads to better outcomes for business

This is fascinating. Thank you @shayonislynn.bsky.social for spotting it

Independent Thinking Politics Podcast · 3 Seasons · Updated weekly

This is a really good analysis of what’s next for Syria, from the @chathamhouse.bsky.social team including the always brilliant @haidhaid22.bsky.social Worth a listen podcasts.apple.com/gb/podcast/i...

Thank you Alan!

78% of employees bring their own AI to work (BYOAI), says ex No 10’s Peter Heneghan at the Athens Corporate Comms conference - fairly shocking stat that shows how badly we need policies and strategies about which tools we use and how

Fantastic to hear from industry colleagues who I admire hugely at the Corporate Communciations conference in Athens @farzanabaduel.bsky.social @richardbagnall.bsky.social and @jrgerlis.bsky.social

Thank you Jon!

💯 agree with this!

Thanks Chris! It’s a fantastic book and a great example

Why every crisis team needs a challenger, holding the team to account | Polpeo When we run simulation exercises, almost every team says the same thing: “We were too slow to act. Time ran away from us!”

One of the things teams always say to us during a crisis simulation is “We were too slow - we thought we had more time!” There’s a reason for that. And it’s why we all need a challenger in our crisis teams - to keep time and hold us to account polpeo.com/why-every-cr...

It’s terrifying

That looks amazing.

A picture of a boxer puppy taken with a phone, with an auto suggestion that it could be a photo of Donald Trump

I think my phone suggestion of who might be in this photo is unfair to the boxer puppy

Hello! Lovely to see you here. It’s so much nicer than X (for now at least)

Fascinating insight into what we should be worried about by the amazing @lucygobag.bsky.social - her book When the Dust Settles is the best crisis preparedness book I’ve ever read.

Does Brio look secretly relieved?

Oh. That’s not good

Can’t see enough of Brio. How is cat nemesis?

You know I always have a crown handy

I’m having that put on a business card!

Gorgeous but flooded commute today

Both on there!

Definitely not. And we need more dog pictures in our lives right now

Absolutely love this. And what a weekend to be in FT Weekend!

Yes to all of this. It’s a massive project (one we’re doing right now) to fully understand the implications of team/company-wide adoption, and balance benefit with risk (bias, hallucination, environmental impact, security) - totally different to finding ad hoc uses and experimentation. Great piece.

Great call. It's such a toxic place

I’m starting a crisis comms starter pack here (go.bsky.app/Nb3ZjhF) - it’s early days but recommendations very much welcomed! A lot of crisis experts are not in Bluesky yet so I’ll keep adding as they join

Love spray mounts at dawn 😂