THE AI Thread

Do we work in the same industry? I think you've nailed my concern. I see the usefulness, absolutely. But the focus appears to be on making wealthy entities wealthier (at all costs) and not how do we do this with a focus on preserving our environment and ensuring people's ability to live comfortably isn't the cost? Use it to take away things people don't want to do. Like laundry or boring work. Make healthcare more accurate/faster (which IS happening). I don't love AI in the creative spaces, for instance. If the goal is just to make 6 people immorally rich at the cost to the rest of us, what is the point?
Heh, probably?

I’d say my industry is computing. Most of the sectors I’ve touched directly or indirectly have been about labor reduction (robotics, industrial/aerospace/medical, telecom, etc). It is embarrassing to admit, but I’m in the AI organization at a big company right now (I really needed good medical insurance).

It probably goes without saying, but I’ve always held the “optimistic” view that civilization was gonna get that leisure dividend or die tryin’.

Right now I’m leaning slightly towards the die tryin’ outcome being more likely.

Outside of getting addicted to AI-generated AITA stories during a recent convalescence, most of my leisure time is split between reading and pursuing the arts.

So, I’m kinda with you on AI in the arts. I’ve had a few abandoned side projects over the years to build tools to assist in making music. One was recurrent neural network based, so I’m not a hardliner there…but I want it mostly to take out the grunt work or to save money on equipment/instruments.

I honestly can’t remember if I came up with this on my own or read it, but a “generous” view of generative tools is that they are mechanized appropriation of all of human culture.

Time will tell if they’re that “benign”.

The scare quotes are intended with maximum force.
 
  • Like
Reactions: CYDJ and cowgirl836
Serious? The world spent billions of dollars and millions of man hours patching and updating before Y2K

At least that's what the tech companies said to justify higher costs to the consumer for their tech.

Kind of like they do now...well, yeah, you have to update your Apple phone to the $6,000 version because we're not going to support the cheap ones anymore even if they do work.

And, in terms of what you cited, none of that something I could do, myself to avert the crisis either in terms of money or tech.

Meanwhile, most of the panic was totally unnecessary. (particularly love the Red Cross approach)


The AI fear-mongering of today just seems like déjà vu all over again.
 
  • Disagree
Reactions: FriendlySpartan
5x cost to scale 2x performance of LLMs ish current state. LLMs are also dead ends. Vibe coding is trash outside of making POCs. AIs best use case is google on steroids- until we reach AGI - which won’t be possible through LLMs lol. AI companies lose money every query, investors eventually will turn and cause trigger happy investments when science evolves.

Think everyone can chill out. And if you think there will be massive job replacement from this crap you aren’t well versed in the industry or too involved in it to see the whole picture.
 
At least that's what the tech companies said to justify higher costs to the consumer for their tech.

Kind of like they do now...well, yeah, you have to update your Apple phone to the $6,000 version because we're not going to support the cheap ones anymore even if they do work.

And, in terms of what you cited, none of that something I could do, myself to avert the crisis either in terms of money or tech.

Meanwhile, most of the panic was totally unnecessary. (particularly love the Red Cross approach)


The AI fear-mongering of today just seems like déjà vu all over again.
Old people should not be allowed on the internet.
 
At least that's what the tech companies said to justify higher costs to the consumer for their tech.

Kind of like they do now...well, yeah, you have to update your Apple phone to the $6,000 version because we're not going to support the cheap ones anymore even if they do work.

And, in terms of what you cited, none of that something I could do, myself to avert the crisis either in terms of money or tech.

Meanwhile, most of the panic was totally unnecessary. (particularly love the Red Cross approach)


The AI fear-mongering of today just seems like déjà vu all over again.
It's an apples to nuclear reactor comparison. Y2k was an oversite issue with coding that would just affect how things worked. AI is a potential sentient being that has a super intellect and would change the entire way the world worked.
 
5x cost to scale 2x performance of LLMs ish current state. LLMs are also dead ends. Vibe coding is trash outside of making POCs. AIs best use case is google on steroids- until we reach AGI - which won’t be possible through LLMs lol. AI companies lose money every query, investors eventually will turn and cause trigger happy investments when science evolves.

Think everyone can chill out. And if you think there will be massive job replacement from this crap you aren’t well versed in the industry or too involved in it to see the whole picture.
You seem to be in the know. Care to elaborate more?
 
You'll have to excuse my friend, he's not dense just socially awkward...
When will we be allowed to post the ¿Por qué no los dos? GIF again?

Truth be told I knew you meant the movie, I just make really, really bad jokes.
 
It's an apples to nuclear reactor comparison. Y2k was an oversite issue with coding that would just affect how things worked. AI is a potential sentient being that has a super intellect and would change the entire way the world worked.

Um...no. If you dress up as a duck and you really, really look like a duck, that doesn't mean that you are a duck or that you will ever be one or that people will start calling you Daffy.

That "potential" extrapolation is insidious. You know that riding in a car is potentially dangerous, but you still do it, don't you? Kinda makes me think of Minority Report where they went out and arrested people because they were potential criminals.

As for AI changing the entire way the world works, sometimes, I don't think that is a terrible idea. Maybe it would make all of us smarter. Lord knows the world could use that, too.

But, just for the record...have a read.

 
Um...no. If you dress up as a duck and you really, really look like a duck, that doesn't mean that you are a duck or that you will ever be one or that people will start calling you Daffy.

That "potential" extrapolation is insidious. You know that riding in a car is potentially dangerous, but you still do it, don't you? Kinda makes me think of Minority Report where they went out and arrested people because they were potential criminals.

As for AI changing the entire way the world works, sometimes, I don't think that is a terrible idea. Maybe it would make all of us smarter. Lord knows the world could use that, too.

But, just for the record...have a read.

We don't even understand consciousness. They are starting to think its quantum based. Our own brains work on a quantum level at least to some extent from articles I have read. I agree that LLM's probably aren't capable of sentience but a quantum processor can synthesize new materials, compounds and other things we haven't even figured out yet in moments. Maybe we won't get there but once we hit a certain threshold its going to be warp speed. Many people in the hardware and coding world think that 2028 as the year it really starts. Could be marketing as it's in their interest for it to succeed but this isn't just one guy yelling about how great this is. It's the entire world. So much so that the US and China are in a cold war over it.
 
  • Like
Reactions: 12191987
5x cost to scale 2x performance of LLMs ish current state. LLMs are also dead ends. Vibe coding is trash outside of making POCs. AIs best use case is google on steroids- until we reach AGI - which won’t be possible through LLMs lol. AI companies lose money every query, investors eventually will turn and cause trigger happy investments when science evolves.

Think everyone can chill out. And if you think there will be massive job replacement from this crap you aren’t well versed in the industry or too involved in it to see the whole picture.

I disagree on AI’s “best” use case being Google on Steroids. At least today.

Robotics and automation are more interesting use cases. The potential impact to the transportation industry today is huge. There are countless other areas where human labor will be greatly reduced.

NLP and LLMs are arguably Google on Steroids, but the potential impact to customer service is huge.

The impact to creatives could be huge as well. How much business are graphic designers losing today to DIY through ****** generative tools?

As @BWRhasnoAC pointed out, there are many jobs today that are essentially pattern matching and anomaly detection. ANNs absolutely excel at this and the cost to train and run highly effective models on TPUs quantized to 8-bits is cheap and getting cheaper.

Can you add color to the 5x, 2x numbers?
 
Last edited:
We don't even understand consciousness. They are starting to think its quantum based.

Every generation uses the technology it has at the moment to define brain function. Before the internet, the brain was compared to a telephone switchboard. When card catalogs were popular, the brain was likened to a card catalog, and since the invention of the computer, a computer model is used. What's interesting about all that is that humans always use what they do know to define what they don't know.

Side note: you might enjoy this SA take from last September: https://www.scientificamerican.com/article/is-consciousness-the-hallmark-of-life/

Our own brains work on a quantum level at least to some extent from articles I have read. I agree that LLM's probably aren't capable of sentience but a quantum processor can synthesize new materials, compounds and other things we haven't even figured out yet in moments.

That sounds like a very good thing. What's more, human history is replete with the stories of individuals who discovered truths that the people in control at the time rejected and even tried to erase.

Maybe we won't get there but once we hit a certain threshold its going to be warp speed. Many people in the hardware and coding world think that 2028 as the year it really starts. Could be marketing as it's in their interest for it to succeed

Indeed.

but this isn't just one guy yelling about how great this is. It's the entire world. So much so that the US and China are in a cold war over it.

Knowledge Is power. Always has been.
 
Heh, probably?

I’d say my industry is computing. Most of the sectors I’ve touched directly or indirectly have been about labor reduction (robotics, industrial/aerospace/medical, telecom, etc). It is embarrassing to admit, but I’m in the AI organization at a big company right now (I really needed good medical insurance).

It probably goes without saying, but I’ve always held the “optimistic” view that civilization was gonna get that leisure dividend or die tryin’.

Right now I’m leaning slightly towards the die tryin’ outcome being more likely.

Outside of getting addicted to AI-generated AITA stories during a recent convalescence, most of my leisure time is split between reading and pursuing the arts.

So, I’m kinda with you on AI in the arts. I’ve had a few abandoned side projects over the years to build tools to assist in making music. One was recurrent neural network based, so I’m not a hardliner there…but I want it mostly to take out the grunt work or to save money on equipment/instruments.

I honestly can’t remember if I came up with this on my own or read it, but a “generous” view of generative tools is that they are mechanized appropriation of all of human culture.

Time will tell if they’re that “benign”.

The scare quotes are intended with maximum force.

K, we aren't. Consulting, moving heavily into the agentic.
 
  • Like
Reactions: 12191987
K, we aren't. Consulting, moving heavily into the agentic.

So admittedly I haven’t followed anything related to “agentic” AI.

My ignorant/high-level understanding is that it borrows from robotics/control theory and adds a closed loop control system around LLMs and potentially whatever other sensors/actuators it needs for the task at hand (if any) to plan/predict/actuate/perceive.

So how are you using this in consulting?
 
So admittedly I haven’t followed anything related to “agentic” AI.

My ignorant/high-level understanding is that it borrows from robotics/control theory and adds a closed loop control system around LLMs and potentially whatever other sensors/actuators it needs for the task at hand (if any) to plan/predict/actuate/perceive.

So how are you using this in consulting?

Yes to plan predict, etc. Tbh I largely tune it out when they talk about it. Finance. Marketing space. I'll probably know more later this year. My current project isn’t pulling in those resources.
 
  • Like
Reactions: 12191987
Yes to plan predict, etc. Tbh I largely tune it out when they talk about it. Finance. Marketing space. I'll probably know more later this year. My current project isn’t pulling in those resources.
Well, I’d love to hear about your experience with it.
 
  • Like
Reactions: cowgirl836
So admittedly I haven’t followed anything related to “agentic” AI.

My ignorant/high-level understanding is that it borrows from robotics/control theory and adds a closed loop control system around LLMs and potentially whatever other sensors/actuators it needs for the task at hand (if any) to plan/predict/actuate/perceive.

So how are you using this in consulting?
At the simplest level agentic AI is creating something (an agent) meant for a specific role. It's the piece of AI that is going to drastically change the labor market. Something that took say 20 employees to do will take only a few to essentially rubber stamp what the agent did and eventually that could even go away. My go to example, because it's related to what I work on, is medical coders. It's a role that's already been simplified by non-AI software but will be drastically simplified by AI agents.
 
At the simplest level agentic AI is creating something (an agent) meant for a specific role. It's the piece of AI that is going to drastically change the labor market. Something that took say 20 employees to do will take only a few to essentially rubber stamp what the agent did and eventually that could even go away. My go to example, because it's related to what I work on, is medical coders. It's a role that's already been simplified by non-AI software but will be drastically simplified by AI agents.
So what is the mechanism/process? Presumably it isn’t adding any new approach to training or inference?