Artificial Intelligence

[Tero]

i have used AI to diagnose toddlers, and it was actually right. Toddlers do not get sprained ankles. I try not to ask things about me. At least not things I would like to talk to a human about. But just skirting around the topic, I asked about magnesium citrate. It does seem to have a connection to anxiety. I had a bottle I had quit using. I aske about sleep this time:

Yes, evidence suggests magnesium citrate can aid sleep by relaxing muscles, reducing anxiety, and calming the nervous system. Studies show it may improve sleep quality, reduce the time it takes to fall asleep, and boost melatonin. It is particularly effective for those with a magnesium deficiency, which is linked to poor sleep.

link it gave
https://www.verywellhealth.com/magnesiu ... ep-7501514

not exactly Pub Med

The thing I have always failed to get info on is the actual lack of magnesium. There we got PubMed
Magnesium deficiency is extremely common, with approximately 45% to 50% of the U.S. population failing to meet the recommended daily intake. Globally, an estimated 2.4 billion people, or about 31% of the world's population, do not consume enough magnesium.
PubMed Central (PMC) (.gov)
+4

https://pmc.ncbi.nlm.nih.gov/articles/PMC6163803

[Brian Peacock]

So now you're using it as a quack doctor, a paediatrician, and a therapist that helps you remember the old days.

If Finns had friends they wouldn't need chatGPT!

[pErvinalia]

[Tero]

So now you're using it as a quack doctor, a paediatrician, and a therapist that helps you remember the old days.

If Finns had friends they wouldn't need chatGPT!

I have a Finn who has kind of hovered in the friend sphere, who may even have heard of me. I know his brother and knew his dad. He was living alone and would not come out for days. He was stuck on some web site, posting all day. Conspiracies may have been part of it. Finally the family organized an intervention and took his laptop away. He has to have lunch with his son once a week. He is a translator among other things so he types well. He was not happy posting on a phone. He was normal after a month. Here is Mats in his 20s. He was a record producer into the 2000s.

mats.jpg (10.88 KiB) Viewed 252 times

Finns are a bit funny about meeting people as well. I can't just call him up and go talk to him. His brother would need to be there. I would actually interview him but he is such a minor figure in Finnish music that it would mostly be stories of his band mates. He was in the band when a sleepy driver went in the ditch and the band's van ended up on its roof.

[Tero]

Big party in Finland. There are a few days they come out of the cave:

https://yle.fi/a/74-20223306
these are graduating students, so they are a bit more outgoing. You are not allowed to turn into the stereotype Finn until 30.

[Tero]

I asked about upgrades. Apparently all the time:

Key Aspects of Google AI Updates:Frequent Improvements: Google constantly updates its AI ecosystem, spanning areas like video generation, coding, and email management.Gemini Integration: Gemini is deeply integrated into Google devices, with updates aimed at making AI more proactive.Faster Functionality: Updates in early 2025 have already focused on making Gemini more capable, allowing for more complex reasoning and better handling of complex queries.

[pErvinalia]

So I have my DOF calculator app created. Claude Design created an HTML version. Was a slow process. I hit the limit at one point and had to wait three days for my limit to recharge. Went close to hitting my limit on the second run, so thankfully it got there, as my Claude subscription ends tomorrow. I got it to create a handoff document for Claude Code to translate it into an iOS app. Took about 5 minutes to create the iOS app. I can't test it yet, as I will need to wait till my new Mac arrives. But I assume it will have errors, going by the process with my clipboard app. Will try to correct the errors with non-subscription Claude, but if it can't handle it, I'll have to get another month subscription. These fucking apps better make me some money. This has turned out to be an expensive folly. :/

[Tero]

What's the process to sell apps to iPhone and Google store?

[Tero]

The language part of AI is quite ahead of all the stuff, in that is just has a huge number of examples to make sentences. And it needs to do it in a bunch of languages. I was thinking of the time I had leaerned prespostitons in Swedish. I then had to learn them in English and there was no one to one match. In Finnish we use cases. The case ending for "to you" or "for you" is the same, sinuLLE. But we do not use "for" for all the same things English does. "For many years" comes out as "in the time of many years" in Finnish. Or basically during. But we do not have during as such. We use a clumsy "in the passing of" phrase. We have words that look like prepositions but we do not put those in front of the noun, we put it behind. "Pöydän päällä" is "the table's top of."

All these thing come out perfectly going Finnish to English. I do not use English to Finnish much, strictly for one word at a time.

[pErvinalia]

What's the process to sell apps to iPhone and Google store?

Well you've got to pay apple to be a developer. AUD150/year. Then apple takes a cut of each sale. Google is free to submit apps. Not sure what cut they take. Most android apps are free. Both perform checks on the app before allowing it to be listed.

[pErvinalia]

Seth might have convinced me!

[not that it's never possible, but that there's good reasons why it might not be possible]

[JimC]

"Never" is a big ask. Current types of AI, no, I can agree that true consciousness is beyond them, but in the future, with an unforeseen twist?

[pErvinalia]

An interesting point was why are we more likely to attribute consciousness to LLMs VS something like Alpha Fold? They're both neural nets. Obviously we feel language use is highly correlated with consciousness. But is it really necessary? If it is, is there a difference between the LLM prediction machine and our own language use? We both pick the most appropriate next word based on our experience (training data for LLMs). Although, we understand meanings, something which LLMs arguably don't. And then you can throw in Chomsky's universal grammar and the idea that the brain has evolved to inherently understand language. But then we get back to the thing about inputs and outputs being the same while not knowing what's going on in the "inside".

[aufbahrung]

LLM's are part of the mechanism of humanesque thought that exceeds human thinking. More likely to be fifty years than five years away from a proper human like thinking machine. Not to say it won't be achieved. But like fusion, these fifty years problems tend to be always fifty years away from their solution. Maybe one day - nobody predicted the current situation so the future is a gamble....and maybe one day after we've all passed on

[Svartalf]

"Never" is a big ask. Current types of AI, no, I can agree that true consciousness is beyond them, but in the future, with an unforeseen twist?

the problem is not whether a program can gain actual consciousness and sentience, but whether it can act like it possessed those traits. and dammit, it you don't know what buttons to push to force them to show their limitations, they are so close to passing Turing's test.