“Do your own research!!!”
1 Like
This report used AI
1 Like
Half of statistics are made up 70% of the time.
1 Like
1 Like
School system officials and an Omnilert spokesperson said the technology worked as intended.
In this case, Omnilert’s monitoring team reviewed an image of “what appeared to be a firearm” on the person at Kenwood Monday night, said Blake Mitchell, a spokesperson for Omnilert.
He said the image closely resembled a gun. However, Allen, who was handcuffed and searched by police, said an officer showed him the picture and it looked like a bag of chips to him.
Really, I’m less bothered by the grotesque failure than by the fact that we are casually drifting into a hardcore surveillance state.
Big brother is now watching you, recording you, and assessing you in real time.
1 Like
You have 20 seconds to comply
5 Likes
RFK has hacked the AI system? Taking MAHA a bit far.
1 Like
Don’t bring donuts, FFS.
Waiting to see videos of someone yelling “it’s low sodium!!”
I have been one of the Beta Testers for my company involving Copilot.
While its not super useful yet at the excel level (given the complexity of the work that we do) I think that the new direct Copilot function in excel will make a more positive change as we can start using it directly in embedded functions.
Was explaining this to our CFO (he is an accountant) as the C-Suite is now increasingly focusing on the cost/benefit of paying for AI functionality (this year they are being ruthless with costs).
There was a lot of FOMO initially about AI use (throw it at every problem and see what sticks) but now it feels like they are shifting towards evidencing more practical (and positive value add) uses of the technology.
1 Like
It is unclear to me how to measure the benefit of many of these more “ad hoc” ai uses.
In many cases you probably need some kind of randomized blinded trials to do it right, but that is not practical.
2 Likes
I thought cooling data centres in space was actually a real PITA?
Reminder human’s are still very involved in AI
https://www.cbc.ca/news/business/ai-economy-human-labour-data-annotation-fine-tuning-jobs-9.6967918
This was my thought as well. Here’s Google’s paper, they don’t talk much at all about cooling, mostly focused on radiation shielding and launch costs.
Reading it I think the headline is overstating the extent that this is a plan vs a random farfetched idea to play with.
This has been like pulling teeth. I wanted some compliance advice from the upstream agency’s compliance department centered around the use of AI. I wanted to know what type of infrasctructure etc would be suitable.
So I emailed them first to say sort of 'we both agree you can’t use a public facing AI, even if you ‘anonymize it’, so based on that, what are my requirements?
They said "use public AI’s, just make sure you anonymize your data’.
I said, uh, no. You can’t anonymize it. Here’s an example, two prompts:
- Free life insurance continuing education credits
- Best birthday gifts for my 60yo spouse.
And…out of 40 million people those two prompts uniquely identify me. Because they have my IP address and therefore my location, and in the venn diagram, of people in my 'hood that have a 60yo spouse and are a licensed life insurance agent, the intersection is just me. One person.
Then I throw out this prompt:
- best life insurance company for someone with factor V lieden.
And now, by the same logic, I’ve probably exposed my client’s personal medical and insurance information to a company that specializes in reconstructing data like this. And that’s not legal at all.
So here’s what they came back with:
Which is a pretty weak way of saying don’t use it. They didn’t address that it clearly breaks privacy and health record regs, just waffled and said theres no AI specific. But, the problem isn’t AI specific. It’s no different than posting this stuff on facebook.
Anyway, we’re a bit further ahead.
1 Like