My first review! Thanks to @tkbeynon for the s and for engaging with the stories in Code Dependent deeply. “A penetrating look at how we’re allowing artificial intelligence to infiltrate all parts of society, from policing, welfare, justice and health”
@madhumita29
-
Join AI Q&A Session: Future Questions with Reporting Expert
By
–
On March 27, come join me to ask all your burning questions about AI and our future, and hear from the frontlines of reporting on this mad story over the last decade
-
Political Bias in LLMs: A Cross-Model Research Analysis
By
–
But you know this is an issue with basically all LLMs, right? Here’s a recent study that looks at political perspectives displayed by a number of models. Surely, that’s the larger story, not Google’s politics. https://
aclanthology.org/2023.acl-long.
656.pdf
… -
Google’s Inclusive AI: Addressing Representation in Creative Tools
By
–
So one wouldn’t have concerns about Google’s mission if all the images of 15th century artists were white Europeans? In any case, this is an imperfect creative tool, not search – why not aim to create more inclusive visions of the world, within reason?
-
Do LLMs Lie? Understanding Intent and Deception in AI
By
–
Lies are deliberate and require intention which an LLM does not have
-
Model Tuning Choices Shape AI Inaccuracies and Outputs
By
–
I guess my understanding is all inaccuracies are in some way a result of tuning choices made by model creators. And this one is just particularly visible and visceral to (some) humans..
-
Unrepresentative Exceptions in Tech Leadership and Workforce Diversity
By
–
Something can be unrepresentative of what usually holds true but not necessarily wrong or impossible. Like women tech CEOs. Or male nannies.
-
LLM Inaccuracies and Representation Bias in AI Systems
By
–
Inaccuracies obviously an issue. But haven’t we all agreed LLMs make mistakes? I haven’t seen this sort of reaction for any other errors. Outrage directed at things like Arab mathematicians in 17c or female soldiers, as if these things are an affront
-
Gemini Image Generation Bias: Race Representation in AI Models
By
–
Impossible to have a nuanced conversation about race on here, but genuine question: why does seemingly everyone have such a huge issue with Gemini generating images of women/people of colour, when AI has been doing this for YEARS in the opposite direction?
-
Historical Accuracy vs Representation in AI Generated Content
By
–
Isn’t the outrage overblown? There may be some historical inaccuracies but people seem annoyed about stuff like prompts for 17th century physicists/mathematicians showing Arab/Asian/African people and not Europeans. Surely that’s not inaccurate, just unrepresentative?