It slightly depends on what they think their job is, doesn’t it? If it’s “spouting plausible rubbish” then Gen AI is spot on. If they’re actually looking to do some proper work, then probably not.
It slightly depends on what they think their job is, doesn’t it? If it’s “spouting plausible rubbish” then Gen AI is spot on. If they’re actually looking to do some proper work, then probably not.
I have a plausible answer: No.
"Is it your job to do anything correctly?"
youtu.be/HqjhHVUzl8o?...
With your digital futurist hat on - if you were an MP with a case load - not the speeches bit - is there any scenario where AI could or should be used?
Problem with using AI to help with case load is that it would require uploading some potentially very sensitive information. MPs should never, ever use any model that folds uploaded data into its training model. That would not just betray a constituents' confidentiality, it could be dangerous.
Depending on the AI used - the data should be protected. www.computerweekly.com/news/3666141... learn.microsoft.com/en-us/copilo...
Which goes back to what Nick was saying about formalising its use across MPs
Heh.
"Should". Sadly, my trust with this stuff is limited.
Yes, let’s use Microsoft instead of running models locally using Ollama (or similar) 🤦♂️
Yeah, running a local - circumscribed - system seems like the obvious choice, but a) it multiplies the resource problem and b) government culture feels very much geared to push the other way: acquire as much data as possible.
Maybe. It wouldn’t be the Gen variety, but you might be able to do some interesting stuff with a more serious application. Trouble is it’s likely to be messy - like the crime prediction systems which basically replicate our prejudices.
But… traffic flow for public transport? School places? Health risks and hospital beds? All kinds of things seem plausible, with good data and an awareness of what the system can and cannot do with it.
Traffic control? What could go wrong? 🦈
“When driving, ignore Left and Right signs. These are merely political slogans.” - Gerald Hoffnung
Ah. Flow. Design. Not real time control.
I am sure Firewatch said much the same thing at the start... 😅
But it’ll take work and sensible governance! So… maybe.
I’m interested in that dataset analysis. IBM describes it as a form of Generative AI. www.ibm.com/think/topics... At some point (like now), we’ll have to see the limitations, and I think the reaction will be to set some regulations (restrictions) on its use. The genie is out.
The public discourse is so hyped that the terns of art are a bit broken. ML vs AI vs AGI etc. Say, then, that eg an AlphaFold for traffic flow seems plausible to me, but I’m concerned about the quality of human social data.
And certainly I’d be hesitant about whether an individual MP - already under-resourced given the gravity of the job - has access to the necessary staff, data, expertise & compute. Maybe a group?
My biggest concern with MP's using AI for constituency work is that they will be sharing sensitive and personal information with the AI provider, perhaps without consituents realising this is the case. There's a Data Protection element to it that, as you say, individual MP's may not appreciate.
At this point I’d be happy if a government would accept the basic notion of privacy in the digital space.
You could almost treat the question as a litmus test of whether someone is a good MP.
And you should.
The Voight-Cunt Test?
*crying* 🤣🤣🤣
There are a LOT of pro writers - including me - who would give their time to write speeches for an MP they believed in if, let’s say, that MP’s budget wouldn’t stretch to a speech writer even part time.
Of course, that would require MPs to behave with decency or risk losing their writer. You couldn’t… I don’t know, go on Newsnight and spout crap about refugees and trans people, because then I’d bin your speech.
I can imagine a sketch whereby the writer slips a few last minute changes into the final draft.
Maybe we could replace MPs with an elected AI model in each. constituency? Vote for the desired outcomes and model weights, parameters and hyper parameters. A socialist paradise lies ahead!
It would probably do a better job than the legions of advisors they employ. But they will keep the advisors and replace the civil servants with AI.