Mind your language

Peak geek.
Logophile.
Bibliophile.
(Im)mature BA Hons English Language graduate.
Plain English devotee.

[Mind your language @KnottyProblemEn]

@JGrossund @jenatlionra He is a she. But thanks, once again, for proving why the models are so biased in the first place.

[Mind your language @KnottyProblemEn]

@SarahForRuncorn What. The. Actual. F*ck.

[Mind your language @KnottyProblemEn]

@SandraHerz9 @jenatlionra Have you been on LinkedIn recently? The place is flooded with women in tech.

[Mind your language @KnottyProblemEn]

@ClipperCitizen @jenatlionra Intent doesn't mask the fact that 'bro' literally means brother. You may have intended it to be neutral but if your interlocutor is not aware of the sense, or the use isn't common outside of fifteen year-olds, you're failing at communication.

[Mind your language @KnottyProblemEn]

@SandraHerz9 @jenatlionra The feedback loop is critical to ensure that subsequent releases of the models are trained appropriately.
I am a data annotator. I review and annotate model outputs. Models do get retrained.

[Mind your language @KnottyProblemEn]

@operation_rosie @jenatlionra That's an unfortunate typo.
A model might be frozen during a single interaction, but an AI programme is dynamic. If we don't capture corrections when the model is wrong, we lose the exact data required to fine-tune it and prevent the same mistakes in the next version.

[Mind your language @KnottyProblemEn]

@ClipperCitizen @jenatlionra I'm a woman, bro. And it's assumptions and biases like this that are used to feed the LLMs misogynistic BS.
https://t.co/UksFaNQONc

[Mind your language @KnottyProblemEn]

@jenatlionra Agree that the core resp lies with the creators to fix their data sets. My point relates to the broader feedback loop. While the model won't learn instantly mid-session, flagging these errors is what forces developers to update the training data and address systemic bias.

[Mind your language @KnottyProblemEn]

@annex_girl @jenatlionra I'm a straight woman and your bias and assumptions are what propagate LLM's biases.
https://t.co/UksFaNQONc

[Mind your language @KnottyProblemEn]

@jenatlionra Agree that the core resp lies with the creators to fix their data sets. My point relates to the broader feedback loop. While the model won't learn instantly mid-session, flagging these errors is what forces developers to update the training data and address systemic bias.

[Mind your language @KnottyProblemEn]

@stalliamet @jenatlionra https://t.co/UksFaNQONc

[Mind your language @KnottyProblemEn]

@jenatlionra Agree that the core resp lies with the creators to fix their data sets. My point relates to the broader feedback loop. While the model won't learn instantly mid-session, flagging these errors is what forces developers to update the training data and address systemic bias.

[Mind your language @KnottyProblemEn]

@CDrakeFairTrade @jenatlionra It is helpful, it's just not infallible. And we, as humans, do have a duty to understand that.
https://t.co/UksFaNQONc

[Mind your language @KnottyProblemEn]

@jenatlionra Agree that the core resp lies with the creators to fix their data sets. My point relates to the broader feedback loop. While the model won't learn instantly mid-session, flagging these errors is what forces developers to update the training data and address systemic bias.

[Mind your language @KnottyProblemEn]

@Klovn @jenatlionra https://t.co/UksFaNQONc

[Mind your language @KnottyProblemEn]

@jenatlionra Agree that the core resp lies with the creators to fix their data sets. My point relates to the broader feedback loop. While the model won't learn instantly mid-session, flagging these errors is what forces developers to update the training data and address systemic bias.

[Mind your language @KnottyProblemEn]

@xbasslichtie @jenatlionra https://t.co/UksFaNQONc

[Mind your language @KnottyProblemEn]

@jenatlionra Agree that the core resp lies with the creators to fix their data sets. My point relates to the broader feedback loop. While the model won't learn instantly mid-session, flagging these errors is what forces developers to update the training data and address systemic bias.

[Mind your language @KnottyProblemEn]

@Eric_In_Osaka @jenatlionra https://t.co/UksFaNQONc

[Mind your language @KnottyProblemEn]

@jenatlionra Agree that the core resp lies with the creators to fix their data sets. My point relates to the broader feedback loop. While the model won't learn instantly mid-session, flagging these errors is what forces developers to update the training data and address systemic bias.

[Mind your language @KnottyProblemEn]

@Fraccy0 @SandraHerz9 @jenatlionra I am a WOMAN and I do support the work to fix it, so pipe down, lovey.

[Oliver Kamm @OliverKamm]

Two of the five don’t. Orwell’s celebrated essay is deeply misguided (notably his strictures against passives) and his ‘rules’ are plagiarised. Zinsser’s On Writing Well is a terrible book that absurdly claims most adverbs & adjectives are ‘unnecessary’. The book is full of them!

[Mind your language @KnottyProblemEn]

@jenatlionra I do data annotation. I am one of those humans that reviews AI output and flags incorrect responses, bias, etc. so that models can be re-trained.
And, for the avoidance of doubt, I'm not a bro. I'm a middle aged woman, so you can see exactly how the machines learn their bias.

[Mind your language @KnottyProblemEn]

@jenatlionra Agree that the core resp lies with the creators to fix their data sets. My point relates to the broader feedback loop. While the model won't learn instantly mid-session, flagging these errors is what forces developers to update the training data and address systemic bias.

[Mind your language @KnottyProblemEn]

@jenatlionra I love the assumptions that I'm male.

[Mind your language @KnottyProblemEn]

@MadelaineLucyH Now ask an editor.