This can’t possibly go wrong.
TBH, anything that makes social media more broadly untrustworthy might be a positive at this point.
Ok, I’ve changed my mind: AI crap can stay.
A legit use of AI in academia: Figuring out what other people have done in a field not previously known for great communication.
I agree that AI is a great tool to find (generally) relevant information to start with. Very terrible in terms of finding “answers” w/o additional follow up.
BC is restricting the growth of AI data centres because of their high electricity consumption:
“British Columbia intends to curb the growth of data centres, including those used for artificial intelligence, giving priority instead to natural-resource development projects competing for a limited supply of electricity.
A proposed law, introduced Monday, would free the Crown-owned utility BC Hydro from the obligation to supply electricity equally to all industrial customers. It would also ban cryptocurrency mining – another sector with large electricity demands.
BC Hydro has for decades provided an abundant supply of cheap hydroelectricity that has helped shape the province’s economic development. With industrial demand set to spike in the next decade, however, it is now shifting to a model of allocation.
AI and data centres are energy-intensive, and the B.C. government is sending a signal to investors in the mining industry and other natural-resource sectors that their projects will take priority.”
The problem is that these data centers don’t do squat for economic development. THey get built, then there’s no staff, just an exhorbitant amount of energy consumption that consumers have to pay to build out and deliver. They’re a net negative on society.
1 Like
And that is exactly the way the B.C. government views it. Unlike Alberta who is welcoming the data centres despite not having sufficient electricity supply
I gotta beat my developer to get me an AI installed that I can use. He’s put up a couple older models, waiting on him to put up updated versions. There’s so much time saving in AI. The more I use it, the more I want to use it.
I need a whitepaper built for a website. I want to take about 20 10-20 page PDF’s and create the business case in the whitepaper. I could read and summarize and take two weeks. Or, I could tell an AI to read them and summarize, and I’ll have a better answer in a half hour than I could do on my own.
I’m doing a talk/seminar and needed an instagram post. I don’t even know what the parameters of that are. Images? Text? both? In about 10 minutes, I had a professional level image and post ready to go. And both were better than I could’ve ever done in a day or two of work. I just told it what I wanted in the form of a paragraph or two. It gave me 4-5 options. I pointed a couple things I liked. It gave me some more options. About three iterations of that and I said ‘yep, create the image and the post’ and done. Better, quicker.
I wonder if the angle was supposed to be 89.5 degrees.
2 Likes
I get the impression there’s an astroturfing campaign starting in Newfoundland to get us to sell our hydropower potential to an AI firm.
Or the 3rd angle being 2 degrees.
Or perhaps AI can do things in a non-Euclidean space; like hyperbolic space.
3 Likes
Not just like hyperbolic space, but definitively. There are 3 types of 2-dimensional geometry, with hyperbolic being the one where angles of a triangle sum to less than pi. (More interestingly, there are 8 types of 3-d geometry, the proof of which is left to the reader who is willing to wade through the works of Bill Thurston and Grigori Perelman)
2 Likes
Looks like some foreshortening with the zoom lens.
First hit on Google looks right
3 Likes
did your angles = 180?
Yes. Found the picture here
Has anyone ever blown out the top of the ai responses, i.e. gotten too many token errors? I’m getting that on our llama 3 install when I’m asking it to do some heavy lifting, integrating a lot of PDF’s as sources.
I also gave it this prompt:
can you regenerate the last response, but include the sources inline please
and it gave me a bunch of inline sources along with this:
Note: The sources listed are examples and may not be real references. You should replace them with actual sources that support the information presented in the whitepaper.
Duh.