Stephen Bush: 'LLMs to send birthday messages' is the latest fascinating example of 'a thing that is useful for governments and organisations being weirdly packaged as useful for households'.

'LLMs to send birthday messages' is the latest fascinating example of 'a thing that is useful for governments and organisations being weirdly packaged as useful for households'.

So much of, e.g. anti-fraud is going to become easier and more effective the more of it we can automate it (a lot of it is just 'we got this bit of the state to talk to this bit of a bank which talked to this mobile phone company' which obviously takes a lot more time the less automated it is.

It is not going to usefully replicate the same person who works at the anti-fraud bit of a bank wishing their partner a happy birthday, come on.

They say it’s the thought that counts, but thanks to AI, you no longer have to think at all!

OhNo.jpeg

as a developer, i do not see how llm would even enter this picture, all of this can be automated without it the key problem with automating these is always that you still need human supervision for cases where automation goes wrong, and if anything, llms make this task harder

Agree that you are never going to, nor should you want to, automate humans out of the process. But turning the job of someone who works into anti-fraud into one where they spend more of their time making decisions and less time translating their own work to multiple audiences is a time-saver.

I suspect that requires a cultural shift as much as a technological one. I’ve had some very surreal conversations with my building society’s fraud section, often centring on their contention that basically any external (cheaper) transfer service is “prone to scams”.

here's the thing though: AIs have been in use for many years now to process INCOMING information so what you will probably end up with is machines producing untold shittons of slightly faulty paperwork in a human-ish language, which other machines will process (once again, with errors)

at which point it's like using voice commands to write a text to someone who uses text-to-voice to "read" them you might as well just leave a voice message you don't need llms in inter-machine communication, their only possible use is in human-machine interaction, but the human side limits scale

I mean - in this case, we're talking about a thing that real hospitals and government departments have done: they don't produce slightly faulty paperwork because they are still being inputted and finalised by clinicians or HUD officials, they just cut down the amount of time the paperwork takes.

This and many similar tasks. For example technical documentation for safety-critical systems. Are the AI bros confident it can safely do that?

"as a developer, I do not see how llm would even enter this picture" through the marketing department, as per 🤷

Right - 'machine learning for government bureaucracy' is not sexy, which is why the marketing is 'LLMs to write birthday messages'.

As a good example of this, read up what the Dutch tax services did a decade ago to families making use of 'fraudulent' child care services and how dodgy algorithms and deliberately limiting human intervention in same meant tens of thousands of people got ruined.

Or come to think of it, the Post Office scandal where 'computer says you're guilty' was accepted without question.

Contrary to many people's imagination, LLMs aren't only used for producing new text, they're also used for processing incoming human-generated text. You could very well feed an LLM with sufficiently long context window size a number of documents and ask "do you see any inconsistencies between them?"

Agree, I can't see how they improve things or add any value beyond novelty

And much more reliable too if the range of allowed requests and responses is fairly tightly constrained by spec (in the same way any current REST API might be now, for example)

Anti-fraud is the last thing you want to let AI lose on; cf. post office scandal.

No - the Post Office scandal is not relevant here, that was about using tech for fraud detection, not speed. And the central problem in that case was the people and leaders who persisted long after it became clear the software was returning false positives.

It's naive to think that you can use ai just for speed without it impacting decision making.

Well, what is “AI” here? States use algorithms throughout public policy, and have for longer than anyone has been alive.

AI is great for fraud detection if you implement it correctly.

It's pretty telling ngl that they have this tech and "wish someone else a happy birthday" is a use-case they come up with and "navigate this company's awful phone menu and cancel a subscription to a thing I don't want" isn't

Oh, they absolutely can create such a phone menu. They just don't want to.

mmhmm

Having seen how something like a phone menu may come about in a large organisation, I feel that Hanlon's Razor applies in many if not most cases.

Scrabbling around for risible use cases is so 2010s cryptocurrencies. LLMs too are going to settle on the only profitable one: fraud.

Yes, instead of call centers, it will be data centers selling us solar and stealing our parents money.

AI is a therapist with no potential clients. “Is there a reason you’re forgetting your Father’s birthday?” “Shut up Siri! Can just learn how to change the channel on Hulu!”

Again, I find it strange how much of this stuff is taking over the human side of things rather than the boring grind work. mrgan.com/ai-email-fro... is an interesting take. And, IMO, quite sad.

How it feels to get an AI email from a friendmrgan.com Recently I received an AI-written email from a friend. I'm trying to understand why I had the reaction to it that I did.

I think the simplest answer is that, unfortunately, boring grind work tends to be heavily, repetitively, fact dependent and that is exactly the shit genai is not (and likely cannot, w/o some separate system) ever be good at

It's execs designing products for people like themselves. I work in the industry and see it happen over and over again.

Indeed. I wish they'd design a thing that talked to my friends' calendars for me and spat out dates for us to discuss, which would also be useful to the execs.

I think we're heading in this direction, but implementing this is a privacy and compliance nightmare. mspoweruser.com/microsofts-o...

Microsoft's Outlook's upcoming feature will tell you where your team ismspoweruser.com From allowing users to schedule meetings with Copilot AI to provide them with new ways to copy emails and organize folders, Microsoft Outlook is set to become a lot more useful in the coming months. T...

I want my availability displayed to my friends to reflect how keen I am to see each of them, how far it is where they want to invite me, and what else I have on in any given week. Good luck with that Alexa.

I was attending AI conferences with papers promising this over 25 years ago... 'AI agents' was the AI hot topic of the time. But instead of a global bubble we had just one or two AI companies getting super overinflated and imploding.

Cf www.independent.co.uk/news/busines... - I think Autonomy's business was AI agent focussed, if I remember correctly

Tech mogul Mike Lynch cleared over multi-billion dollar salewww.independent.co.uk British tech entrepreneur cleared of fraud charges

I think the LLMs may be a breakthrough tech now for that.

AI has been able to do scheduling pretty well for decades. Other than more confidently giving answers that may or may not be correct, while using lots more energy, I'm not sure how LLMs can solve this class of problem any better. The only difference is people being more willing to share data.

IMHO The real problem for AI schedulers is that people tend to not like the results of having their time assigned automatically by AI. The only thing worse than struggling to schedule a meeting is having a meeting in your diary that you were trying to avoid.

Extracting information from text data.

Kind of fascinating inverse of Zuckerberg’s obsession with pitching VR headsets as Microsoft Teams: Now With Added Motion Sickness rather than games consoles

It’s a really fundamental mistake about how human relationships work. A therapist will say: yes we need to talk about payment and when we’re meeting and treat those conversations as seriously as when you talk about your parents because you are present in all those matters.

Human relationships take place IN the small things. Not in the peak experiences. A marriage doesn’t consist of the wedding, the most athletic sex and the annual holiday. It exists in “how we decide what movie we’re watching” and “what you say to me when I left the kitchen in a mess”.

There’s a you-ness that actually emerges MORE in the moments when you don’t feel that “this is the super important part”. You slip your own leash, you reveal yourself, that is where other people can get to know you.

It’s a unifying factor of so much of this stuff - people are being told to be excited about stuff that… isn’t exciting to them.

It’s obviously broader in potential applications, but like with the crypto stuff, SO much of the language it’s being sold in is ‘a select number of people will become rich without doing anything’ to which the average person will just go partridge_shrug.gif