Lightnews — Scholar-powered news

Melissa McCradden @mdmccradden.bsky.social · 2d

The worse part though is that as their costs go up for adding these nonsense features, they inevitably offload it onto us, the consumers. 😡 I’m now looking into old school options again to get away from this stuff.

1 3

Melissa McCradden @mdmccradden.bsky.social · 3d

And the fact that they even think this is possible shows they have absolutely zero understanding of any concepts and principles related to “bias” 🤨 If they cannot understand the very concepts they invoke, then they have zero credibility to address the problem and should sit this one out.

Luke Steuber @lukesteuber.com · 3d

We took every personal, cultural, and institutional bias in our language and baked it into the foundation of this tech. I would say good luck with that, but they know damn well this is marketing

People don’t get that their output is derived from *us.* It’s not just biased, bias is in its genes

OpenAI is trying to clamp down on ‘bias’ in ChatGPT

GPT-5 is better at resisting liberal ‘pressure,’ the company says.

www.theverge.com

Melissa McCradden @mdmccradden.bsky.social · 3d

Not arrogant - 100% accurate. The very fact that these folks can switch up their buzz words on a dime shows they don’t understand the words in the first place.

1

Reposted by Melissa McCradden

Jeff Doctor @jeff.doctor · 3d

I've yet to see a single discussion of "Canadian digital sovereignty" include any Indigenous experts, at best we're given a token mention that at some point they'll "consult with Indigenous Peoples" like we're some sort of monolith. Umm, y'all need to consider all our lands, and our distinctiveness.

1 12 37

Melissa McCradden @mdmccradden.bsky.social · 5d

Gonna pump my own paper here which is basically saying it’s the wrong problem to focus on for making good decisions in medicine 😅

www.sciencedirect.com/science/arti...

Explaining decisions without explainability? Artificial intelligence and medicolegal accountability

www.sciencedirect.com

1 2

Melissa McCradden @mdmccradden.bsky.social · 13d

We’ve also pre-registered our study on OSF to encourage transparency and sharing with the scientific community

osf.io/p6tm5

OSF

osf.io

Melissa McCradden @mdmccradden.bsky.social · 13d

This shared accountability is important to limiting the expanding problem of moral crumple zones, per -

estsjournal.org/index.php/es...

Moral Crumple Zones: Cautionary Tales in Human-Robot Interaction | Engaging Science, Technology, and Society

estsjournal.org

1 1

Melissa McCradden @mdmccradden.bsky.social · 13d

This is a feasible, generalizable approach to evaluating commercial scribe tools to make an informed decision about where they fit (and don’t fit) & building clinical procedures to share the accountability between physicians and their health institution

1 1

Melissa McCradden @mdmccradden.bsky.social · 13d

Sharing our pre-print for a comprehensive silent evaluation of a commercial AI Scribe tool at Women’s and Children’s Health Network in Adelaide!

Co-developed by our multi-d team of physicians, Aboriginal Health colleagues and consumers ❤️

url.au.m.mimecastprotect.com/s/X_F4C81VLx...

A medical algorithmic audit framework for evaluating the safety, equity, and quality of an AI Scribe tool in a paediatric developmental assessment clinic

Background Any tool that can reduce the administrative burden on healthcare providers while preserving safe, accountable and high-quality medical documentation is of immense value both to healthcare i...

url.au.m.mimecastprotect.com

1 5

Melissa McCradden @mdmccradden.bsky.social · 22d

I like the argument that is glimpsed in @abeba.bsky.social ‘s work - we can oppose using robots in this way w/o the anthropomorphising that is dehumanising. Curious if @abeba.bsky.social you’re expanding some of the argumentation in 2.3 of the paper?

arxiv.org/abs/2001.05046

Robot Rights? Let's Talk about Human Welfare Instead

The 'robot rights' debate, and its related question of 'robot responsibility', invokes some of the most polarized positions in AI ethics. While some advocate for granting robots rights on a par with h...

arxiv.org

1 5

Melissa McCradden @mdmccradden.bsky.social · 22d

Do you know what I find wild tho? Nearly a decade of being an ethicist - raising concerns, questions, challenges in a plethora of contexts has been my job. This is different. I’ve never felt so at risk personally for doing this as I do for LLMs. People get personally offended, get nasty even.

1 8 23

Melissa McCradden @mdmccradden.bsky.social · 23d

“is it just everybody is scared to say this and pleased I did?”

^^^* YES, we are!!!

1 1 5

Melissa McCradden @mdmccradden.bsky.social · Sep 4

Hey #AIEthics people (and all people as well!), got a question for you…

Is anyone else 100% completely and totally fed up with AI ethics?

1 3

Melissa McCradden @mdmccradden.bsky.social · Aug 27

I think it’s kinda fascinating to see the epistemic struggle in this write-up between the beliefs espoused about LLM “capabilities” versus the considerations re ethics violations… 🤷🏼‍♀️

2

Melissa McCradden @mdmccradden.bsky.social · Aug 24

I hate that I believe you are 100% correct on this prediction 🫠

1 2

Melissa McCradden @mdmccradden.bsky.social · Aug 11

Always stunning how the scientific standards for study designs and the claims one can make from them get thrown out the window when it involves AI. Any study that is not longitudinal is meaningless given we are now seeing the long term effects of AI use = worse learning and performance.

1 10 30

Melissa McCradden @mdmccradden.bsky.social · Aug 9

I will believe AI has PhD-level knowledge when every response is concluded with a suggestion of a local seminar or talk that has free food 🍕☕️

1

Reposted by Melissa McCradden

Lana Tikhomirov @lanatikhomirov.bsky.social · Aug 6

Our preprint is up! Ever heard of the silent phase of AI evaluation for medical AI? Well, now we’ve summarised the current state of research!

Melissa McCradden @mdmccradden.bsky.social · Aug 6

We are also SO excited to release our preprint for our scoping review reporting on current #silenttrial practices for #HealthAI

@lanatikhomirov.bsky.social did an amazing job leading this work all the way through ❤️

osf.io/preprints/os...

OSF

osf.io

1 2

Melissa McCradden @mdmccradden.bsky.social · Aug 6

Please get in touch if you do or are interested in the silent evaluation phase - we want to hear from you!

1

Melissa McCradden @mdmccradden.bsky.social · Aug 6

We are also SO excited to release our preprint for our scoping review reporting on current #silenttrial practices for #HealthAI

@lanatikhomirov.bsky.social did an amazing job leading this work all the way through ❤️

osf.io/preprints/os...

OSF

osf.io

1 2

Melissa McCradden @mdmccradden.bsky.social · Aug 6

🤩Excited to share that Project CANAIRI has an official website!

Find out about our team, our work, and how to get involved!

www.projectcanairi.org

Project CANAIRI

Artificial intelligence (AI) tools are moving into healthcare at a steady pace in the hopes that they will augment care, administrative tasks, and support clinicians in their daily workflows. The fie...

www.projectcanairi.org

1 3

Melissa McCradden @mdmccradden.bsky.social · Aug 4

And in between there are the doors 🚪

1

Reposted by Melissa McCradden

Carissa Véliz @carissaveliz.bsky.social · Aug 1

Turns out that #Google was publicly logging ChatGPT conversations that people had shared (privately, or so they thought) with friends. There are conversations about mental health, sex, etc. Do you still think #OpenAI take #privacy seriously? By @stokel.bsky.social

fastcompanyme.com/technology/g...

Google could be reading your ChatGPT conversations. Concerned? You should be

Thousands of shared ChatGPT chats are now appearing in Google search results.

fastcompanyme.com

5 20 29

Melissa McCradden @mdmccradden.bsky.social · Jul 29

I have seen this!! We tested some models with dummy data and found that someone calling someone or something “deadly” was ALWAYS coded as negative sentiment (it’s the opposite!) We ended up deciding the sentiment function will be useless and scrapped it.

The Conversation AUNZ @aunz.theconversation.com · Jul 29

A new study shows large language models find it much harder to understand the nuances of Indian, British and Australian English.

‘Are you joking, mate?’ AI doesn’t get sarcasm in non-American varieties of English

theconversation.com

4

Melissa McCradden @mdmccradden.bsky.social · Jul 29

I have also wondered whether using the common name (eg ChatGPT, Gemini) in papers (esp titles) instead of the actual model name/version is like using the brand name of the drug rather than the substrate… probably not gonna look good down the line

4 34