Welcome… to… the machine 
Pink Floyd was quite prescient with that one.
1 Like
Well, I don’t play a mean guitar, but I do like a steak bar.
2 Likes
A short talk on some of the ways they are using to ID fake images.
1 Like
Loved it. The main message is: There are very good scientific methods to tell generated images apart from real ones. So we do know which ones are fake and we can separate slop from truth. We‘re not helpless.
1 Like
I am reading a book at the moment and this passage leapt out at me, especially as it relates to this topic:
“… no algorithm was innocent of its designers. It couldn’t be. There was an originating purpose for an algorithm, however distant in its past - a reason some human made it, even if it had evolved and folded in on itself and transformed.”
- From. A Memory Called Empire, by Arkady Martine
And that is why I love Sci-Fi. It tends to be prophetic.
6 Likes
Meanwhile on the “is this a good idea” side of things, the FT published an article about potential overspending of 1,5tn $ on cloud compute capacity until 2030 if “AI” doesn’t start turning a profit.
I hope you Murricans are ready to get your grid power and water supply rationed until the bubble bursts.
4 Likes
Isn‘t the AI craze everywhere? Why reduce this to the USA?
4 Likes
Dotcom all over again… 
1 Like
I hope our electronic replacement species turns out better than us, as I’m beginning to think biological lifeforms have too many inherent (and fatal) flaws, making us unsuitable for long-term survival.
2 Likes
Anyone else finding Gemini a lot more useful and reliable than ChatGPT 4? I want to say more mindful?
I‘ve been using Perplexity a lot due to this so never got into the habit of using OpenAi or Google web offerings on a regular basis.
But now that OpenAi has released a small open source model for home use (fits in 16gb), I wanted to try it and play around. LM Studio has a low entry barrier and works with AMD GPUs via Vulcan on Windows.
Because AFAIK 70% of those datacenters are supposed to be constructed on US soil.
3 Likes
And because in much of America, home owners have little to no say in what gets built next to them.
3 Likes
1 Like
I like this guy
It can’t play tic-tac-toe worth a hoot either.
Wheels
The strange thing for me is: This is supposed to be the easy part. Just recognize it is math, and pipe it into Wolfram Alpha (or an equivalent math engine).
1 Like
It’s truly mindboggling how idiotic this is.
But, AI isn’t being lead by particularly clever individuals. It’s just tech bros with limited experience in IT and trustfund springboards that launched them.
I legitimately want to see Altman, Musk, and everyone else involved tarred and feathered.
3 Likes
Deleted my previous post, because I didn’t mean to reply to Aginor directly, but rather to the thread.
I’ve solved untold algebraic equations with Copilot in an effort to help tutor my daughter in a subject that she has a difficult time focusing on in the classroom, and ones that I’ve long forgotten how to solve. As an older parent trying to prevent her from failing a class, Copilot has been amazing. It walks you through its problem solving line by line, rather than spitting out just the answer. That helps me remember and her learn the process. She can’t use it on exams, so she must be absorbing some of the problem solving processes.
And it does this is seconds, a huge benefit to exhausted, working parents who often aren’t asked for help until 9:30 at night.
So that fact that ChatGPT can’t solve a specific problem is an interesting talking point, but no more. And something like Notebook LM is freaking amazing. Give it a few resources and ask it to create a layman’s study guide about any complex subject, like quantum physics, or astrophysics.
That doesn’t mean that we should ignore the environmental impact of such tools. Rather, discover ways to harness the tech while working to responsibly source its energy needs.
4 Likes