Careful what you ask for Maiden, AI is getting pretty advanced and ass kissing may soon be a feature...

Congressional AI hearings coming after DeSantis used AI to create pics of Trump hugging Fauci and used it in an ad.

So Congress doesn't know about photoshop and editing software?

They are like 400 years old on average, I am surprised most know what a computer or indoor plumbing is.

Will put it here too since not everyone reads the politics thread. This is peak fucking AI use. Absolutely hilarious shit.  Just had a long rant from Trump that he would have built a wall around the infinity stones to stop Thanos but Biden's immigration policy just let him take them.

Was funny they claimed last week AI could not make jokes, then this comes out.  So want to know what the fuck they used to train these as this is great shit.

The Obama Biden Trump AI squabbles are the best.

Well that way fast. Channel already banned. Sucks as that was some of the funniest shit I heard in a long, long time.

TrumporBiden2024 AI stream is back on twitch and growing in popularity. Not sure what the ban was, but came back fast and is now even more insane.  

Some highlights from the Twitch Trump or Biden 2024 debate thing. It is so stupid, but so fucking hilarious.

Dude how could Biden not win in a landslide if he just opened with that 10 seconds LMAO.

The Biden AI is insanely savage. The Trump one was is ok, but the Biden one says the most insane, unhinged shit.

First neuralink implant expected in the next six months. This is gonna be interesting as the science for the AI link that Musk is promising still does not exist.

Elon claims he will be one of the first people getting one, but I seriously doubt even he is that stupid.

This is an interesting report from Morgan Stanley:




Having been born after the industrial revolution but growing up watching manufacturing automation getting better and cheaper through the price of robotics etc. it is amazing to be ringside as this happens all over again for information workers as well as a range of service industries.

As an example, the new version of Photoshop can just remove backgrounds and do it near perfectly, I did this in the past for work and the masking etc was a huge job, now it is the click of a button.
Who would have thought we would witness the end of models in print and television, now everything can be computer generated and considering how good the result is now at the very emergence of the technology it is only going to rapidly improve as the hardware and software does.

www.visualcapitalist.com/sp/ranking-industries-by-their-potential-for-ai-automation/

So basically the more WFH it can be?

Studies like this one presume all problems can be solved with surface level knowledge. And people are finding right now that AI either simply cannot answers that require deep knowledge or expertise to sold, or it confidently gives a wrong answer falling to a local minima, a known problem for AI.  Teaching computers conceptual reconstruction that comes with expertise is proving very problematic.   

Also that list is really weird as AI lawnmowers are already a commercial thing, but this presumes that no one will lose their job to one. Which part of the error of this thinking, it presumes AI will not be paired with robotics.

Also there is a bit of a conflict of interest as MSCI sells administrative support tools, which ironically is the industry they found at most risk for disruption.

Also if AI can ever beat just in time instruction, it means it learned metacognition and now is self-aware. People better pray it does not consume 25% of teaching jobs, as that means the terminators are already here.

Jul 3, 2023 2:21:17 GMT @ness said:

So basically the more WFH it can be?

According to that we are all going into construction or service jobs.  

The most interesting finding from AI research on expertise is that when AI gives an answer, even if it explains why it gave the answer, novices will not be aware of flaws in the answer as they lack the conceptual knowledge that expertise brings to determine why the answer is wrong. So anyone using AI to replace a expert worker, is taking massive risks.  

The expertise hurdle has been tackled for 40 years now and still has not been cracked as AI cannot understand why a solution is correct at one level and wrong at another. Maybe with quantum logic it will be cracked, but not with what we got now. It picks one paradigm and solves for it.  But even it is cracked, an expert will needed to make sure it solved for the correct paradigm or pick the correct paradigm of solution.

Jul 3, 2023 2:41:38 GMT c said:

Studies like this one presume all problems can be solved with surface level knowledge. And people are finding right now that AI either simply cannot answers that require deep knowledge or expertise to sold, or it confidently gives a wrong answer falling to a local minima, a known problem for AI.  Teaching computers conceptual reconstruction that comes with expertise is proving very problematic.   

Also that list is really weird as AI lawnmowers are already a commercial thing, but this presumes that no one will lose their job to one. Which part of the error of this thinking, it presumes AI will not be paired with robotics.

Also there is a bit of a conflict of interest as MSCI sells administrative support tools, which ironically is the industry they found at most risk for disruption.

Also if AI can ever beat just in time instruction, it means it learned metacognition and now is self-aware. People better pray it does not consume 25% of teaching jobs, as that means the terminators are already here.

I looked at the AI lawn Mowers, they are pretty average and only work in very specific scenarios
We have an AI vacuum cleaner and I don't see that this is putting anyone out of a job.

Many jobs are too manually different to be automated, like the hamburger making example earlier in the thread.
Could someone buy an AI mower to do their own lawn? yes, for sure, but it is never going to be set and forget unless your lawn is a perfect small rectangle.
In many applications robotics are rejected because of the variation in the tasks required to perform.
I have actually seen factories where robots are used and a percentage of the goods have to be manually restacked before the robot is able to deal with them... Which undermines the point. It also means you end up not reducing your headcount as the robots have to be monitored.

But if we presume AI advances over the next 20 years, then robotics should as well.

And robotics show the pitfall of AI, AI will work until it does not. Except with AI people using it may not even notice it gives a wrong solution to a problem until it is far too late.

People been using AI to play the stock market for 40 years and all failed so far to beat humans even without a black swan event. When AI can replace most of these jobs, it should also be able to play the market perfectly sans black swan events.

Jul 3, 2023 3:19:29 GMT c said:

But if we presume AI advances over the next 20 years, then robotics should as well.

And robotics show the pitfall of AI, AI will work until it does not. Except with AI people using it may not even notice it gives a wrong solution to a problem until it is far too late.

People been using AI to play the stock market for 40 years and all failed so far to beat humans even without a black swan event. When AI can replace most of these jobs, it should also be able to play the market perfectly sans black swan events.

A couple of points:

Firstly, it is all not that simple.
You have seen the Boston Dynamic examples, the data processing required for those moves is astronomical.
We are not talking about a small increase in processing power being required, but an huge leap forward.
Real time processing of data isn't really viable yet on any platform.
Even ChatGPT, the models are largely crunched off line to be able to produce answers and the hardware required is mind blowing.

This is why in the past it has been easier to automate a process with machinery over robots.
Easier to produce a vacuum cleaner with AI built in, but again, it is a 2-axis solution.

If you want a lawn mower they tend to cut the grass and leave it there, if you want to gather and bin the clippings that is a whole other thing.
What about leaves dropping on the lawn, what about tree branches, what about rocks and kids toys.
That's where tasks go from simple to complicated.


Secondly, nobody consistently beats the SPX and most people who try return to the means.
It is a rational system, but people are irrational.

For AI to be able to replace jobs it would also need a huge leap forward. If AI cannot mow a lawn, how would we expect it to manage a multibillion dollar corporation in real time, a task that chart claims should be easy as AI can replace a lot of jobs in management and administrative support.

AI will eventually beat the SPX. People are rationally irrational.  Prospect theory and fuzzy logic can account for most of the irrationality.  It is the tuning and dodging of local minima that is the problem.  But I doubt most models will limit themselves to the SPX.

Jul 3, 2023 4:20:58 GMT c said:

For AI to be able to replace jobs it would also need a huge leap forward. If AI cannot mow a lawn, how would we expect it to manage a multibillion dollar corporation in real time, a task that chart claims should be easy as AI can replace a lot of jobs in management and administrative support.

AI will eventually beat the SPX. People are rationally irrational.  Prospect theory and fuzzy logic can account for most of the irrationality.  It is the tuning and dodging of local minima that is the problem.  But I doubt most models will limit themselves to the SPX.

This is why they are so specific in the industries they say are rife for disruption.
The jobs at risk first are those that are most repeatable.
Many sales, admin and procurement jobs are repeatable, some obviously are not.

One of the game changers with recent AI developments is the use of multiple AI agents that interact with each other.
They are still clunky and prone to timing out, but the possibility is there to create an almost infinite number of single task AI agents interacting.
This is way more simple than making a single AI agent that does everything.

It also closely mirrors what you will see in Industrial Automation.
The assembly line is still the most efficient way to do things even with robotics, each automation cell does a specific task and then passes the job onto the next cell.
The more complex each cell becomes the greater the likelihood of failure and the harder to diagnose what went wrong.


To understand automation it is a case of looking at the number of steps.
Mowing the lawn is an example of a simple task for a person that is easily done.
When you start mapping out a process it becomes complicated, how wet is the lawn; has anything been left on the lawn, where are we going to put the clippings, is the bin full etc.
You would likely need one robot to mow, another to empty the hopper, another to rake etc.

I find it at work, part of a site induction is that we are meant to have JSEA that steps through the process of the entire job, but it is impossible because of the number of variables and the cascading change that makes for a project.

Some desk jobs are as simple as receive an inquiry, check if we have current pricing, if not chase up an offer from the supplier, prepare the offer, send it through... These are the types of jobs that really could have been automated already, AI just saves someone the steps of writing out all the IIF statements. 

Play a shit ton of modded Minecraft. Modded Minecraft becomes all about logistics and automation.  Lots of fun with logic gates and production lines.  Almost like factorio but even more chaotic if you build things organically and expand as needed.

If all you need are IF statements, then there is already a replacement for your job. We could replace those simple inquiries and sales jobs with a computer program, but it was found companies wanted a person to deal with. If your site did not have a person, they would find a different one that did. For these jobs to go away, execs would have to give up this aspect of control. And at least this generation has proven they are not ready for that yet. Something goes wrong, they want to reach a person not just a have a computer give them the facts.

Crazy weekend in the world of AI with the board of OpenAI sacking the founder Sam Altman

This one isn't done and there's rumblings he will be back and the board out.
Microsoft will have some anguish having bought in because of Altman.
Lesson is never give up control or buy where you don't have it

He will come back with a better contract in the end. Do not need the board on your side if you got the investors on your side.  Looks like the deal was bring him back, or we go with him to his new company.

///

OpenAI REALLY needs to milk what they have now, as their type of AI is not really novel, and easily copied.  Moreso when the general methods are published.  Will be a race to the bottom fast as cheaper and cheaper versions are released to the wild.  At least for transformer models.  Given the speed of the evolution of methods, transformer learning is unlikely to stay at the top for long.  Moreso now they are commercial, as need to spend more and more time on user features and cannot do large pivots anymore.

I'm going to lighten the mood here, some of these AI covers are insane. Ethically is it OK? Maybe not. I mean, some people have curated entire " new " albums for artists and released them for lols. What's stopping someone releasing a track as their own and using a star's voice and passing it off as a legitimate track as opposed to a bit of internet fun? 

How long before a star sells their " likeness " instead of cutting a verse? 

This will not last very long. Major studios been buying rights to image now, so they can also make their own AI music in the style of bands. Meanwhile they are suing creators and the creators of the tools for copyright infringement. Uses traditional studio math for damages claiming these songs cost them millions each. Due to consolidation, the studios def have the leverage to push congress to push in laws to help them crush AI as well.

Microsoft signed Sam Altman. Microsoft has been taking their top guys as they jump ship, and will likely buy OpenAI once it hits bankruptcy next year or so after all of the lawsuits around training data and copyright infringement. The rise and fall of OpenAI gonna make one hell of a story in the future.

so

um

uh

there is now a fully created AI song (from writer to singer to piano) and Anna Indiana, in her debut single "Betrayed by this Town" is warning us that humans are beyond redemption and Skynet is coming for us

Sitting at my favorite café
Sipping my tea, it's Saturday
Thinking 'bout all he's done to everyone
This town is full of broken dreams
Shattered hopes and silent screams
Somebody please help me

[Chorus]
Betrayed by this town
Let's tear it all down
We're all just destined to fall, I've lost it all
Betrayed by this town
Let's tear it all down
We're all just destined to fall, we've lost it all

[Verse 2]
Alone in the streets, alone in my thoughts
Thinking of all our favorite spots
I thought someday things might turn around
But I was lost and never found

[Chorus]
Betrayed by this town
Let's tear it all down
We're all just destined to fall, I've lost it all
Betrayed by this town
Let's tear it all down
We're all just destined to fall, we've lost it all

[Verse 3]
Faces painted with smiles, lies are told
A façade of unity, our vitality sold
So I sit here in silence just wondering how
To rewrite the tales this town won't allow

[Chorus]
Betrayed by this town
Let's tear it all down
We're all just destined to fall, I've lost it all
Betrayed by this town
Let's tear it all down
We're all just destined to fall, we've lost it all
I've lost it all
We've lost it all

Wait did youtube copystrike youtube o.O