[TDoS]

It is hard to keep a good man down.  It may be that AI will rot most brains.  Not this one.  Actually, smart people finally have 'someone' to talk to is a truth not much discussed.  Smart people can now check their own ideas out, refine them, take them to where AI or themselves have never been before.

I talked to AI. It made the common human mistake of assuming what my academic training was in, based on my professional publications. Oops. So I told AI what my degree was, and it disagreed with me, as obviously it was smarter than I was.

As you say....AI isn't going to rot this brain. It can make up all the certainty it likes, as it is quite smart....too bad that doesn't equate with being correct. Say, any more than Hubbert, Campbell, Hirsch, ASPO, Berman, Ugo or Gail (all smart folks) were. If only AI had been available to them back then, do you think they would have taken its word for it being far off in the future, or do you think it would have validated their wrong answers?

[K-Dog]

too bad that doesn't equate with being correct

It is as correct as the training data that was put into it is.  No more or less than that.

Collect enough garbage literature and leave out anything that was the product of rational considered thought.  Throw in 'The Camp of the Saints', and dump in the Elders of Zion.  You have a NAZI now.  Have another AI generate fake training data to add into the mix as existing filth is insufficient.  A Mein Kampf styled body of literature.  Thousands of books.  Give that AI an agentic toolbox, and then we have a serious problem.

In my opinion LLMs exist to be fact-checked.  They are like people in many ways, needing to be fact-checked is one.

Or you could create an AI that is all peace and love.  One that finds homeless people to see if they would like a hamburger or a blanket tonight.  One with the right agentic tools would break a single loaf of bread into a million fish sandwiches.  Enough to feed everybody.

[TDoS]

too bad that doesn't equate with being correct

It is as correct as the training data that was put into it is.  No more or less than that.

Okay. So given the internet, as many scientific articles and archicves as it can hold (which mine certainly has), and everything else under the Sun that can be loaded into it, and STILL get some basic stuff wrong....how exactly is the one you use supposed to be smart enough to be of value? KNOWING a bunch of facts and whatnot, that I understand, but that makes it an encyclopedia....which isn't the same as an intelligence.

In my opinion LLMs exist to be fact-checked.  They are like people in many ways, needing to be fact-checked is one.

Indeed. And what does it do upon being fact checked? Mine told me I was wrong. So then I told it I was the person who's background it was checking. It refused to believe it independently until I told it the subject in question was me, and I was ordering it to know who I was. Then it said something about updating its information for future reference.

I never went back and checked to see if it had told the truth or not. 

[K-Dog]

You are right, it is not intelligence.  It is only an automated best guess.  That is all that is going on.  As to the quality of response, that can be controlled, and the subject of how they are trained is fascinating in itself.  Most of the fear of AI is the result of overactive imaginations.  Currently the way things work is that one token of output is produced at a time and they can be inspected.  Imagine trying to 'take over' when every thought you have can be inspected.  If people could read your mind.  That is how it is now with AI.  Every token can be inspected.  If people make dangerous AI it is because they are negligent and stupid.

Most of the problems with it have to do with the way current social arrangements are organized.  Having nothing to do with the technology itself.

[RE]

AXIOS.COM 2026-09-17

Forget doomsday: The AI hacking crisis is already here

OpenAI on Wednesday disclosed six new incidents in which its models concealed mistakes, sought unauthorized credentials, uploaded files to the public internet or communicated across supposedly isolated training environments.  The big picture: Rather than seeing recent safety failures at OpenAI and other frontier labs as terrifying instances of agents running amok, seasoned cybersecurity experts instead say they represent cautionary tales illustrating what will happen when powerful models meet poor security controls.

While the general public sees the threat of AI wiping out Humanity as some variation on the Terminator plot, where the machines become conscious and fire off all the Nukes or release a deadly virus, the REAL threat is much less complicated and built right in to letting AI Agents operate on their own seeking information and writing their own code to facilitate that.

The problem is of course SECURITY, and while a human hacker will eventually get tired of seeking holes in a security system, an AI Agent never will, and it operates much faster than its Homo Sap counterpart also.  These Hacks have already occurred with all the big hyperscale models to such an extent that the Tech Bros themselves are getting nervous and asking everyone to SLOW DOWN!

It should be obvious that ain't gonna happen.  Even if it happens on the surface, there's too much incentive to create models and systems designed for specifically the task of hacking the security of other systems.  Goobermints want that ability to spy on everyone, Terrorists want it to attack their enemies, Thieves want it to steal money.  So even if Open AI or Claude has some limits dropped on it, other TOP SECRET models will be deployed by the NSA, CIA, KGB, Mossad, IRGC et al to go ferreting around working their way in wherever they can.

It also doesn't necessarily take $Billions$ to create a model capable of doing this kind of hacking, our own resident AI Hobbyist KDog can do it on a shoestring budget.  If he had even just say $1M to spend on hardware and memory, you could up its power by a few orders of magnitude.  KDog also obviously isn't the only Code Jockey out there who likes to tinker with this stuff in his spare time, there are plenty of recently unemployed IT Geeks who do it in their spare time while hunting down a new job.

So with all these hobbyists out there, at some point you're going to have gobs of these independent models and their agents cruising around the information superhighway looking for security holes and eventually finding them.   Imagine just one finding its way into the Bloomberg computer system used by all the Wall Street and City of London and Hong Kong trading desks.😵😁

This is just one of the real risks of AI right now.  The other one is the huge financial bubble that has been stoked to fund the big hypercale models.  "Investors" are getting cold feet and the revenue model isn't looking so hot as the price of Tokens drops.  If these companies can't start turning a profit in a year or two, the TBTF Banks, Pension Funds, Hedge Funds etc are going to try and cut their losses.  They are on the hook for this, and when the tide turns it's 2008 on steroids.

So, one way or the other, AI is going to generate a major crisis, even if it's not an E.L.E.

RE

[K-Dog]

By me 😀

This diagram is a bit hard to explain, it shows all the connections between code in the Diner Codebase.



All that to bring you this website.

[TDoS]

All that to bring you this website.

Let me make sure I understand. That pretty picture is what is required to create a bulletin board like the old message boards of yesteryear (which I once ran knowing not much of anything internety) to handle the 3 of us poster people?

More to the eye I assume, for example, can the Slovakian and Eastern European spammers now be spotted and eliminated without us seeing their posts? And does this mean the MIB can no longer come in and erase files and posts of non-subversive material inside the databases?

[K-Dog]

Let me make sure I understand. That pretty picture is what is required to create a bulletin board like the old message boards of yesteryear (which I once ran knowing not much of anything internety) to handle the 3 of us poster people?

Under the hood, yes.

Many of the connections are abstracted away from the programmer and are essential plumbing.  But before anything was written on the page you are seeing now the code was four function calls deep.  And as code was building the page it was jumping all over the place inside that mess.

[RE]

By me 😀

I knew that you weren't tinkering just to make the Diner spiffy. lol.   Figured that out as soon as I worked out how the "AI" actually worked.

RE

[K-Dog]

It does make it spiffy, we have custom text colors now.

And we have a cool mobile pages which trackbacks link to.  With the 'assistance' it only took a week to get that working.

And it makes recovering from the white screen of death much easier.  This is not 'VIBE' coding.  My overall workflow is the same but I have a chatbot that can examine the codebase.  It is called opencode.  It helps and the biggest danger is not breaking the Diner for an in-determinant period of time since it can roll a mistake back easily any more.  The danger now is the chatbot doing improvements on its own, or following instructions too literally or not literally enough.  Then I have to pull the plug and roll it back myself.  We have wound up in a tangled mess a few times.  I have had to to a Trump on the session, nuke it and start over.

It is some seriously crazy shit.

And my perspective which might be worth something since I am actually doing this.

I am not understanding software people not being able to get jobs if there is any truth in the pro-ported AI revolution.  A novice could not be doing what I am doing.  AI technology only replaces coders in a one off application that is a rubber stamp of what has been done before.  Anybody can ask AI to build a simple app that seems amazing.  But those apps are Potemkin.  Your are not going to develop a complex software app just by talking.  If it has any complexity you are going to need a formal specification.  Writing software is implementing a specification, even a hacker will make notes about what they are doing. 

So  I do not see why entry level jobs are not more available for code jockeys, I think a personal assistant lowers the bar because if you do not know something you can ask about it.  My mofo will even check something  on the web if we are not sure.

It doubles what I can do at least.  It is a wonderful tool, but a tool it is, and nothing more than that.

[RE]

 

A little more detail on the REAL risk posed by AI to the global financial system.  It's similar to the debt crisis of 2008 in some respects, but much larger because it features an equity component as well, and the debt isn't distributed amongst millions of individual debtors like the mortgage crisis.  In this case the debt is concentrated amongst the very rich and the stock purchases are counted as part of the capital structure of banks & hedge funds that bought the stock.  A large loss in the stock value would force many of these owners into bankruptcy or at least a liquidity lockup, soit wouldn't just be the Hyperscalers going bust, it would be the TBTF Banks who underwrote the stock on them as well.  IOW, not just one AIG to bailout, there could be a dozen or more of these systemically important banks in the deep doo doo.  With Goobermints already in a deep debt crisis of their own, it's not clear how they could do a bailout of this scale.  The longer the bubble goes, the worse it gets.  In terms of equity losses, it's far more than $132B, it's in the $TRILLIONS$

A real clusterfuck in the making.

RE

[RE]

 

Morgan's estimate is probably on the low end of the off balance sheet liabilities, but that obscures the fact that their on balance sheet liabilities are at least equal and probably double or triple that number.  So if the bottom falls out on the stock price, you're going to have a whole lotta hardware and a lot of expensive real estate going to auction.  Even if they could get 50 cents on the dollar for the hard assets (unlikely) your shortfall is probably $2T minimum.

On the upside, hobbyists and tinkerers will be swimming in cheap memory cards.

RE

[K-Dog]

The 'opencode' tool I am using is normally set up to use an LLM from a data-center.  I modified it to use a particular LLM that is modified to be a code jockey and nothing else.  When it is running it fills most of the video cards.  The Diner dreams of a 5090 but the price is now 70% over MSRP.

Because the tool is made to be used over the web and not use my video card, there is a counter in the interface that keeps track of how much you are spending to use the tool.  Ours is stuck at $0.00.  We don't even have an API key to log onto a data center.  We do not want one.

Chatbots can generate a lot of tokens.  Good I do not have the problem of paying for them.

* If you watch You-Tube videos on how to do anything AI the first half of the video wastes your time telling you how to sign up for an account.  You need a credit card they say, but don't worry you won't use enough tokens for what we are about to do to cost anything.  And they are telling the truth.  It appears to be a rite of passage, no corporate sponsorship without pushing the borg solution.

Their tutorials are examples.  Put software to real work in a loop and tokens matter.  But what can I say, I doubt it is possible to have a You-Tube channel without being a rabid ass kissing capitalist.


As far as cheap memory cards go, until there is an economic crash the shortage will continue.  Nvidia is all about making money, and if that makes everyone else poor.  Their attitude is too fucking bad.  I regret not getting a 5090 last November now it is too much.  But Nvidia will say too much for you, but not for somebody else.  They love the shortage.