Does anybody else subscribe to MS 365 (personal)?
They just put up prices by over 40%
Seems they want to use their customers to keep feeding their AI investments, because thats a whopper of an increase over one year.
I found the following article a bit interesting in the AI context:
I think thereās a very long way until AI can āunderstandā business problems to the degree to determine how to simplify it to a(n appropriate) model.
If you seed an LLM with a small amount of false claims, bad things happen. Context is medical data.
1 Like
Ignoring the politics of thisā¦
Iād think this is likely a poor use of AI. Iād think that your scoring system is going to be pretty arbitrary as to where youāre getting good value for money vs. poor value. Iād suspect that youād end up with a Pareto principle situation (assuming Iām not butchering the concept) where the easy projects that make up most of government spending appear to be winners (e.g. passports, ATC, customs and border control; somewhat optimized costs, clear outcomes) and the complex, messy things that government deals with (veterans health, working with First Nations, superfund sites, coastal land loss/erosion/sea level rise) are losers (e.g. high costs, unclear results).
Iām doubtful that thereās been any serious validation work to show this is an effective approach for optimizing spending or verifying that the mandate of the organization is being met, even with the cuts. This would be an example of making dumber people more dangerous e.g. the AI said I should do this, but doesnāt bother to check to see if the AI is capable of giving good answers to the question.
I find this news way to vague. Is the AI being used as a search? Is it writing code? Is it reading lots of documents? Is it providing summaries? Is it making judgements? What data is it looking at, what questions do they ask, and what output is it producing, and what are they doing with the output?
Without any examples or explanations or.. transparency.. itās impossible to judge.
1 Like
Btw, now that we have Elmo in charge, I want to add, the one thing AI is going to be really, really great at is surveillance.
Your data is out there. On facebook, and Google, and Amazon, and discord, and goactuary, and universities, and court systems, and so on, and can be cross referenced with screen names and emails and identified through anecdotes.
And generally speaking, it doesnāt really matter, because itās a lot of work to read everything that has ever been written. And now thatās just not true.
Well, the job losses due to AI integration are now picking up speed.
Here is a Singapore bank.
Iāve seen a number of announcements like this.
I wonder: is this projection based on measured (actual) performance of the new systems in pilot programs?
Or is it optimistic projections that are being used to project savings and make good news?
Iāve started using copilot to solve some of my coding questions in R. Itās definitely faster than looking things up on Google. It gets me to a more specific example faster and is capable of modifying the advice for a particular use case.
ETA: The more productive use case Iād like to see is if we can pair AI with a copy editor to help reduce translation costs.
That already exists?
I have been using it from time to time.
So, looking like Iām going to dip my feet into the AI pool. Hereās one example where AI is actually saving money: #rossvideo #rosslife #chatgpt #techsupport #ai #aifails #aisuccesses | David Ross | 49 comments
And I just got pitched by a lifeco their AI that theyāve seeded with all their underwriting data. So apparently I can ask questions about insurability and have it give me an answer.
One of my fishing with nerds buddyās has graduated and working as a project managerā¦but wants to code. heās been pitching me to do a project with him where he can code and build his resume, I do the design work. Iām interested because itāll let me maybe touch on some software stuff that I donāt do at work.
So, I pitched him on the idea of an AI for life insurance. Either like the life company, where it evaluates insurability for no medical insurance, or an overall general life insurance system.
Unfortunately, while I really consider myself a software person, and Ive been working in developing software in some aspect for almost 40 years, I donāt have a freakin clue how AI works. I started taking some courses online. Iām into the second one and still donāt understand lol.
I would say a combination of both things.
Optimistic projections (for the markets) of current operational improvements using AI.
Look at the latest news from Estonia below.
AI implementation is moving at breakneck speed now.
It may exist, but management(?) has been slow to make the leap to try it.
And the US is now coying China-based DeepSeek.
I figured this was going to happen as the large scale LLMs are still far too loss-making (and expensive to run) thus making them uncompetitive.
This changes the economic situation on the ground as it will allow smaller companies to run efficient AI models at far lower cost.
Good news really. Competition in action.
1 Like
Perhaps stating the obvious
1 Like
Probably could have posted this in political, but thought it was more appropriate hereā¦
Looks like generative AI for creating images still needs some workā¦
ETA: how many problems can you spot?
1 Like
I stopped counting at 10.
1 Like