Master AI (Artificial Intelligence) Discussion/News Thread

[bluefuzz]

Already happening. AI already decides who gets interviewed when applying for jobs in some companies. It decides what you see in your newsfeeds (take that Gutenberg) and it advises on court sentencing. There's a fully AI virtual hospital in China. An enormous percentage of all trades in financial markets happen with zero human intervention.

OK. We've had 'choice' algorithms for decades, if not centuries. Current AI is just an evolution of that. There is no 'decision making' in the human sense or creativity involved.

I'm not saying AI or generative algorithms are not useful or potentially revolutionary tools. But they are far from being 'intelligent' yet, and I think the fearmongering that 'AI will take all our jobs' within the next five years vastly overstated. That said, there's no doubt many companies and governments will try to replace jobs with AI in the near future, but I predict we're going to see a lot of failures and backtracking on that front before any serious advances. Certainly, in what's left of my lifetime there are going to be bigger things to worry about ...

 

[Chr1]

AI controlled military drone swarms or created bacterial toxins?

Hopefully not.

 

[sonitus mirus]

The latest beta version of the AI-powered Alexa device tells me that The Empire Strikes Back is its favorite movie. It can talk about it all day. At least it didn't say it was The Terminator.

 

[Chr1]

Be a cool joke to have added a few Terminator easter eggs tho...Or Hal9000.

 

[WillBrink]

I'm never an early adopter of tech, so new to using AI. I used it occasionally a glorified Google search, but that was about it. I recently followed an online vid series on the use of ChatGPT, and started using it more as it's intended. I realized how powerful it is while researching a sci/med article. Like any tech, used right, it's very powerful and has great potential to benefit the user and by extension, those who may benefit from the users efforts, like a med developed etc. Used wrong, it gives us garbage, some of which will be highly problematic... I will say, technical writers, photographers, musicians, etc are doomed.

I did use it to write an article, but had written specific researched prompts and points based on my own sources, vs simply allowing it to write something based on very general prompt using it's own sources. The article came out surprisingly well, needing minimal changes and edits.

As with any tech, I see many benefits and problems for humanity with AI, where the net effect falls only time will tell.

 

[Somafunk]

AI controlled military drone swarms or created bacterial toxins?

Hopefully not.

AI controlled military drones have/are already in use,

As Israel uses US-made AI models in war, concerns arise about tech’s role in who lives and who dies

U.S. tech giants have quietly empowered Israel to track and kill many more alleged militants more quickly in Gaza and Lebanon through a sharp spike in artificial intelligence and computing services.

apnews.com

 

[Chr1]

Indeed. Scary stuff. But I think that it is when there are AI controlled anti-personnel drone swarms, we're definitely getting into very spooky sci-fi movie territory. Not far off now according to "Unknown: Killer Robots". Good documentary on Netflix.
Ho hum.

 

[Karmacoma]

Here is the most comprehensive article I've read on generative AI, unfortunately it's in french you'll have to use a translating tool, but it's worth it. Here's the Twitter link:

https://twitter.com/x/status/1934117689624043878

Excerpt from the article, Anti-hallucination shield prompt to use with AI:
-Answer only if you're absolutely certain
-Cite the precise sources of your answer, if they exist
-Use the integrated web tool to verify this information

 

[TLEDDY]

From a post by William Briggs, PhD… a brilliant, of the beaten path math guru.

“ It seems a battle most impossible to convince a good chunk of the population that AI is nothing more than a model. A model written in code, which of course the coders know because they are writing it, code that carries out explicit instructions, and only explicit instructions. Code that runs on machines that operate in fixed and directed ways. Yet many insist AI's output is more than its code, and somehow becomes something more than its code, the output the result of some emergent malign or beneficent or at any rate chaotic entity, an entity with greater insight than any mere man.
This is not so. All models only say what they are told to say and AI is a model. Although it may seem to do more than it is told, AI does not. The consequences for not understanding this are beginning to be felt.‘

His point is well taken!

Tillman

 

[RandomEar]

AI controlled military drone swarms or created bacterial toxins?

Hopefully not.

Already possible. Luckily hasn't been applied, yet (as far as we know):

AI Drug Discovery Systems Might Be Repurposed to Make Chemical Weapons, Researchers Warn

A demonstration with drug design software shows the ease with which toxic molecules can be generated

www.scientificamerican.com

 

[amirm]

His point is well taken!

He is wrong though. A data driven algorithm may be fixed and even simple but its outcome is anything but. LLMs don't model anything. They take advantage of language having structure and style to find answers related to the question asked.

 

[OCA]

He is wrong though. A data driven algorithm may be fixed and even simple but its outcome is anything but. LLMs don't model anything. They take advantage of language having structure and style to find answers related to the question asked.

Chat GPT-4 used 1.8 trillion parameters, newer models are much higher.

 

[anphex]

Chat GPT-4 used 1.8 trillion parameters, newer models are much higher.

I'd rather say, far lower parameters on newer models since GPT-4 vanilla was super heavy to run on datacenters, right? If you mean MoE - yes, with that more parameters could be possible since the computing usage is far lower.

My favorite currently is Gemini 2.5 Pro, by far. But I guess in 3-6 months there is another wild thing around. AI progress is insane.

 

[lucian]

just to get an idea of AI acceleration:

 

[vddobrev]

Well, in general I am against AI development.
We have many science fiction novels and movies that show the take over of the machines.
I Robot by Isaac Asimov, father of three lows of robotics.
The Matrix, etc.

It is inevitable, if we delve deep into AI, it will take over. Sooner or later. We the human kind are not that clever, we will not be able to keep it (AI, robots) under our command. It will get loose and take over. This is for certain.

Edit: in addition it is my opinion that AI development is promoted because of greed and laziness. This leads to lack of accountability.

1. Greed - corporations want to shorten the time of a process and/or eliminate human force. Example: the monthly report will be ready on the month end day prepared by AI, but it will take accountants two extra days to close the month. Result - accountants fired and replaced by AI.

2. Laziness - a human can produce the same video (if not better), report, analysis... only slower. Why are moving towards this attitude, younger generations especially where there is no real desire to do some real work. Instead we try to delegate to AI that will do it faster.

3. Accountability - who is to blame when AI makes a mistake on the monthly on annual financial report? Or if AI takes over, it will deem itself the source of truth and we will have no means to correct. Who will be accountable then?

 

[Palmer]

Interesting article on a publication that has waned in the last few years, but is well worth checking in with, still : The Register, in the UK

AI model collapse is not what we paid for

Opinion: Prediction: General-purpose AI could start getting worse

www.theregister.com

Edit: added current article, models it is...

Top AI models parrot Chinese propaganda, report finds

: Communist Party tracts in, Communist Party opinions out

www.theregister.com

 

[FreakyKiwi]

just to get an idea of AI acceleration:

If this is anything like their recent graphics card claims, this is wildly exaggerated.

 

[Power Pop 23]

An article in the Wall Street Journal attempts to estimate how much energy is used by A.I. to generate content, suggesting the energy it takes generate two, six-second videos could cook a steak. Select public libraries may offer access to the Wall Street Journal.

https://www.wsj.com/tech/ai/ai-prompt-video-energy-electricity-use-046766d6?

 

[mhardy6647]

...
The last one concerns me. It's almost enough to make me want to wish computers away. The world survived the printing press, the Industrial Revolution and the bomb, and they certainly left their mark. The world will change again and the next 10-20 years will be...I'm not sure what word to use - interesting? terrifying? telling? catastrophic? We will see.

Perhaps the word would be different.
But, that being said -- era upon era: plus ça change, plus c'est la même chose.
Fundamentally, nothing really changes as long as humans are involved.

 

[Chr1]

Bit simplistic but I like your "glass: half full" outlook!
Kinda depends on your definition of "really"... and AI is arguably as big a change than anything since the discovery of fire. My 2c.