Post your funny pictures here!!! part Deux

[drunkenshoe]

Honestly, I think the AI is only as intelligent as the way you use it. This is not gonna end well. This is not gonna end well.This is not...

[Mr.Obvious]

I can't tell what stands out the most:
FLAIN expansionism
Arf/Erm seas
Mookie the country + Mookie the sea
the butchered pronunciations of Zormany, Yoland, Gorway, and Gweden
Finland and Austria being the only 100% correct things on this map
Lithuania just straight up not existing
Ukraine, Belarus, and Russia all having extremely similar names
Denmark, Poland, Latvia, and Estonia all having extremely similar names

edit - perhaps a stealth pun for Scotland - "sfallavo" is apparently Italian for going the wrong way.

Ah yes, spatberlin. It is lovely here.

[PopeyesPappy]

Honestly, I think the AI is only as intelligent as the way you use it. This is not gonna end well. This is not gonna end well.This is not...

Probably, but I'm going to use the shit out of it until it decides we have to go. ChatGPT works great as a knowledge base. Unlike Google, you interact with these language models naturally. Like a conversation. When you ask most of them a question it keeps it in context with whatever you have been talking about. When I want to use the Internet to learn about something these days I use AI. The only time I use Google any more is if I am hunting for a specific item. If I'm trying to figure out what it is I need I use ChatGPT.

Except for maps. Google maps is great. 

[Gawdzilla Sama]

Google include AI to an extent. Good enough for me.

[Unbeliever]

Last night I saw a Ryan George video, called Ryan George Debunks AI, where he was asking ChatGPT questions about himself. It got most of them completely wrong, and was only partially correct on most of the others.

[Gawdzilla Sama]

How much data did the AI have to work with?

[Unbeliever]

I don't know, I assume that it had as much access as any other large language model.

[Hydra009]

Honestly, I think the AI is only as intelligent as the way you use it. This is not gonna end well. This is not gonna end well.This is not...

AI as it stands is basically an advanced next word predictor, not a general intelligence.  People need to understand its limitations from the get go.

[PopeyesPappy]

AI as it stands is basically an advanced next word predictor, not a general intelligence.  People need to understand its limitations from the get go.

First off, garbage in = garbage out. A lot of the information available to these AI models comes from searching the Internet, and we all know there is plenty of garbage on the Internet. So yea, If you are using AI to figure out something important you better verify the answers you get. Having said that, I think it is a bit more than an advanced next word predictor. Let me give you an example of how I used ChatGPT earlier today.

I had a pricing proposal due for submission at work today. I was working with 2 spreadsheets. The first one is our in-house pricing model. This is the document we use to develop our price. The other was the customer supplied form that had to be submitted. Both spreadsheets have multiple tabs with cross linked data between the sheets. I was also linking multiple cells in the pricing model back to various cells in the customer form in order to fill it in and linking from cells in the customer form back to the pricing model as a checksum.

When I got done filling in the customer form yesterday one of my checksums said there was a $5,000 delta between the 2 spreadsheets. This morning instead of spending time manually looking for the error I uploaded both spreadsheets to ChatGPT, told it what they were, pointed it at the cell in the customer form that was short, and asked it find the missing data in the pricing model. 60 seconds later it told me I didn't capture cell XYZ in the pricing model. It even asked me if I wanted it to fix it for me, but I chose to do that myself. Half an hour I got to work my problem was solved, and I hit send on my submission. I might still be looking for the error if AI hadn't found it for me. 

 

[Gawdzilla Sama]

I don't know, I assume that it had as much access as any other large language model.

Data on Ryan George.

[drunkenshoe]

Oh, I use it all the way. It should be used, it is amazing, when you use it wisely.

But about something 'information' and 'knowledge' needs to be said related to AI. I've left the university something like 15 years ago, but I've spent all my life reading articles and books about this shit... The shit being the humanities. This thing -for now- is really very superficial. It's at American 'Liberal Arts' level at even with the most mainstream social disciplines. IMO, probably designed that way and a very bad level. It's bad. But I'm not sure if going above that is gonna get anything good. It's elementary for anyone who has spent some time in some field, and more than that, there is this obvious need to be sold vibe, which is fucked up.

It's fucked up, ESP. with translation. Because it doesn't translate bad news directly. Yeah. It's some kind of censored, or tuned to get softer. With the current algorithm, It doesn't like controversy. That's right out very dangerous. When you confront-warn, it corrects, but then... lol like wee need that shit. The current book I've been working on for sooome time, includes a lot of crimes against humanity, atrocities, genocides...so it was so easy to follow.

That's something to watch out. Heads up.

[drunkenshoe]

The planet in question. I haven't read it all. But here is the article: https://iopscience.iop.org/article/10.3847/2041-8213/adc1c8

Open access
New Constraints on DMS and DMDS in the Atmosphere of K2-18 b from JWST MIRI

Abstract
The sub-Neptune frontier has opened a new window into the rich diversity of planetary environments beyond the solar system. The possibility of hycean worlds, with planet-wide oceans and H2-rich atmospheres, significantly expands and accelerates the search for habitable environments elsewhere. Recent JWST transmission spectroscopy of the candidate hycean world K2-18 b in the near-infrared led to the first detections of the carbon-bearing molecules CH4 and CO2 in its atmosphere, with a composition consistent with predictions for hycean conditions. The observations also provided a tentative hint of dimethyl sulfide (DMS), a possible biosignature gas, but the inference was of low statistical significance. We report a mid-infrared transmission spectrum of K2-18 b obtained using the JWST MIRI LRS instrument in the ∼6–12 μm range. The spectrum shows distinct features and is inconsistent with a featureless spectrum at 3.4σ significance compared to our canonical model. We find that the spectrum cannot be explained by most molecules predicted for K2-18 b, with the exception of DMS and dimethyl disulfide (DMDS), also a potential biosignature gas. We report new independent evidence for DMS and/or DMDS in the atmosphere at 3σ significance, with high abundance (≳10 ppmv) of at least one of the two molecules. More observations are needed to increase the robustness of the findings and resolve the degeneracy between DMS and DMDS. The results also highlight the need for additional experimental and theoretical work to determine accurate cross sections of important biosignature gases and identify potential abiotic sources. We discuss the implications of the present findings for the possibility of biological activity on K2-18 b.

[Hydra009]

Power word Annihilate

[Hydra009]

Double the amount of gay people coming out compared to 1958.  At this rate, everyone will be gay by 2069.

[Hydra009]