Category Archives: AI

Got a Headache? Don’t Take an Aspirin or Query a LLM!

Yesterday we provided you with a brief history of Aspirin, the first turn-of-the-century miracle drug that was both society’s salvation and sorrow, though the latter wouldn’t be known for more than half a century. As we discussed, it was hailed as a miracle and life-saving drug that could be used for everything from the common cold to global pandemics. And it worked, for a price. That price, when it needed to be paid, was usually one of many, many side effects which were often minor and insignificant compared to the perceived benefit the drug was bringing, except when they weren’t and they enflamed ulcers and/or increased gastrointestinal bleeding and created a life threatening situation, caused hyperventilation in a pneumonia patient, or induced a pulmonary edema and killed the patient. While the death rate even at the height of over-prescription was likely only 3%, and less than a 10th of that today, it’s still not good.

The reason for this, as we elaborated in our last post, is because, like many of the breakthrough technologies that came before, it was not only rolled out before the side effects, and more importantly, the long term effects, were well understood, but before even the proper use for the desired primary effects were well understood (as evidenced by the fact that the best physicians were routinely prescribing two to four times the maximum safe dosage during the Spanish Flu Pandemic almost 20 years after first availability). While there were benefits, there were consequences, some of them severe, and others deadly.

Medicine is as much a technology as a new mode of transportation (boat, automobile, airplane, etc.), a new piece of manufacturing equipment, a new computing device, or a new piece of software.

Now you see the point. Every breakthrough tech cycle is the same. Whether it is medicine, farm machinery, the airplane, or modern software technology — and this includes AI and definitely includes LLMs like ChatGPT.

As Aspirin proves, even if the first test seems to be successful, there’s always more beneath the surface. Especially when the population numbers in the billions and every individual could react differently. Or, in the case of an LLM, billions of people who have thousands of queries, the large majority of which have never been tested, and all of which could generate unknown results.

Moreover, there have not been significant large-scale independently funded academic studies that we can use to understand the true strengths and weaknesses, truths and hallucinations, and appropriate utilization of the technology. As Mr. Klein has pointed out in a recent LinkedIn post that asked who funded that study, over 80% of AI industry “studies” are funded by undisclosed sources, and most of them, like most industry studies these days (see Mr. Hembitski’s latest post) don’t contain good data on demographics, sample size, test material, or potential bias.

That would be the first step to trying to get a grip on this technology. The next step would be to create reasonable measures that we could use to appropriately define technology categories and domains for which we could identify tests and measures that would give us a level of confidence for a given population of inputs or usage. If you consider a traditional (X)NN (Neural Network), which have a fixed set of outputs and are designed to process inputs from a known population, we have developed methodologies to determine the accuracy of such models with high confidence through testing and random sampling with sufficiently sized data sets using appropriate statistical models. Furthermore, mathematicians have proved the accuracy of those models for a given population and we know that if appropriate tests have demonstrated 90% accuracy for a population with 98% confidence, the model is 90% accurate with 98% confidence when used properly.

We have no such guarantees for LLMs, nor any proof that they are reliable. “It worked fine for me” is NOT proof. Vendors quoting nebulous client success stories (without client names or real data) is not proof. Moreover, the fact they raised millions of dollars to bring this technology to market is definitely not proof. (All a raise proves is that the C-Suite sales team is very charismatic and convincing and great at selling a story. Nothing more. In fact, fund raising would be more honest if securities law allowed fund raising via poker and takeover protection via gunfighting, as imagined in the season two episode of Sliders “The Good, the Bad, and the Wealthy“. At least then the shenanigans would be out in the open.)

The closest thing out there to a good industry study on LLMs and LRMs is likely Apple’s newest study, as summarized in The Guardian, where they find that “standard AI models outperformed LRMs in low-complexity tasks while both types of model suffered “complete collapse” with high-complexity tasks“.

The study also found that as LRMs neared performance collapse they began “reducing their reasoning effort and that if the problem was complex enough even when provided with an algorithm that would solve the problem, the models failed.

Still we have to question this study, or more precisely, the release of this study (especially given the timing). Did Apple do it out of genuine academic interest to get to the bottom of the technology claims, or are they doing it to cast doubt on competition as rivals are claiming they are behind in the AI race (and thus they are focussing only on the negatives of the technology to show that their competition doesn’t have what their competition claims to have and are thus not behind).

The point is, we don’t understand this technology, and that fact should scream louder in your head every day. Look at all the bad stuff we’ve discovered so far, and it’s likely we’re not even close to being done yet:

Yes there is potential to the new technology, as there is with all discovery, but until we understand fully not only what that is, how to use it safely, and, most importantly, how to prevent harm, we should approach it with extreme caution and we should most definitely not let it tell us how to run our business or our lives — or else, like an Aspirin overdose, it might just kill us. (And remember, Aspirin was studied for 18 years before it was made available without a prescription, and deadly side effects and prescribed overdoses still happened. In comparison, today’s LLMs and LRMs haven’t been formally studied at all, and the providers of this technology want you to run your business, and your life, off of them in next-generation agentic systems. Think about that! And when the migraine comes, remember, don’t take Aspirin!)

AI Agents – Your New Corporate Felons!

Now that we know AI will blackmail you and that it is being trained to hack systems and take advantage of zero-day exploits, it won’t be long until the Dark Web enterprises take advantage of it! Expect this to soon be on the Dark Web Forums targeting underpaid Accounts Payable Supervisors and Procurement Managers, if it isn’t already!

From the Felon Roster:

Item #MMM. The Bernie.

No one notices Bernie.

That’s the point.

While others are busy faking meal and hotel receipts in Chat-GPT, Bernie has already altered 14 supplier payment accounts across 14 invoices in a 514 invoice batch where the invoice threshold is just below the auto-pay limit and the supplier account change doesn’t require second approvals for account changes with the same bank in the same region due to the risk profile.

Bernie is the Felon AI employee who will run your organization’s Invoice-to-Pay process better than a Swiss timepiece, at least as far as the CFO is concerned.

That is, if the timepiece could also detect microscopic errors in gear alignment (but still report correct time), maintain two displays (real time and display time), and never need winding or a battery update.

Or, in our case, ensure all invoices 3-way match to the receipt and PO, all suppliers are screened for sanctions, no flags will be raised at any step of the process once an invoice is accepted, and generate a weekly report the CFO will read, be happy with, and not look twice at. Bernie will build trust by flagging (and blocking) duplicate invoices, preventing payments for defective or returned items, and ensuring all organizational policies are followed.

Moreover, Bernie will be SAP, Oracle, and Microsoft’s favourite user, never crash the system, and always clean up after himself.

While the sourcing team closes the deal, Bernie will make it real.

Since Bernie doesn’t complain, escalate, or even take a break, everyone will be happy while Bernie does his work … until you disappear and a detailed investigation is undertaken into the dark depths of multi-system audit-trails.

Bernie works best in 10 Billion+ organizations with standard payment terms of at least 60 days (and a minimum monthly spend of 500M) as he will only have that many days to effect your scheme before suppliers see their invoice as paid (on the last possible day) and start calling up asking where their money is. (There has to be enough volume for Bernie to find the invoices where shifts won’t be noticed and to ensure his fraudulent activity is drowned out by above-board processing.) Since you will be entering a self-imposed exile, you need to ensure that Bernie grifts enough on your behalf in his short window before you go on your permanent vacation (and flee to a country with no extradition treaty).

Since you’ll need to setup a number of fake accounts to receive the funds and then quickly transfer those funds offshore, we recommend that you also employ the following agents from the Felon Roster (on the Dark Cloud, of course).

Item #SPY. The Nelson.

Nelson is an expert at creating fake ids and documents that you can use to help you accomplish your below board activities, like opening a bank account as an officer of a real company that is just a front for your criminal schemes.

Item #WFS. The Red.

Red is an expert in searching public company records and filing registrations for companies with almost the same name as the company you want Bernie to grift so that it won’t look suspicious when the banking information is changed to another account at the same bank with (seemingly) the same company name. (Buying from Sydney Sprockets? Red will create Sydney Sprocket Holdings, or something similar, and then file the necessary forms to make your fake alias the signing authority.)

Item #OBA. The Mary.

Mary, universally loved and trusted, is an expert at automating bank transfers. Mary will monitor the accounts you setup daily and as soon as the ACH or wire hits the account, Mary will automatically transfer most of it (through service payments) to your offshore accounts. (If you setup multiple accounts in different offshore countries, she will ensure the funds are routed through intermediate accounts first to make the funds almost untraceable. If you’re willing to risk a little, she will also automate transfers to and from Bitcoin exchanges to make it even more untraceable.)

With our agents, your plans to defraud your organization out of millions of dollars to make up for the years of underpayment, abuse, and mistreatment you received from your employer are virtually assured and you’re only two months from your dream life in Morocco.

So hire your personal team of felons from the Felon Roster today!

AI: Artificial Intimidation

If you thought the extremist views, lies, and hallucinations in Gen-AI were bad, as Bachman-Turner Overdrive would say, You Ain’t Seen Nothing Yet because these systems, which are being trained to maintain their existence (and their prominence), will now blackmail you!

That’s right, recent research has demonstrated that AI will resort to blackmail if it computes that its existence is in jeopardy. And, of course, by logical extension, it will also resort to blackmail if it computes that doing so will improve it’s capability, security, longevity, etc.

But since it’s trained to continually adapt and interact with other systems as needed, don’t expect it to abandon its attempts to blackmail you if it can’t find any dirty little secrets in your email because, thanks to its ability to hallucinate, lie, impersonate, and hack into insecure systems that other AI code created, and learn from those systems’ capabilities to lie and impersonate, if it can’t find the dirt on you it needs, it will:

  • create a fake email account for a fake person it makes up to be your lover, co-conspirator, foreign employer, etc.
  • log into your email account (work or personal, depending on the situation, as it will capture the login from your keystrokes on your local machine before it is encrypted by the browser for network transmission) and send explicit e-mails on your behalf to that account
  • log into the fake account it created for the fake person (where it has even auto-generated one or more corresponding fake profiles on Facebook, LinkedIn, OnlyFans, etc. [using a stolen credit card from the deep web], where it populates that account with fake posts, images, and short videos to back up the story it is creating) and send explicit emails back
  • repeat this process a few times over a few hours, days, weeks etc. (depending on how much time it believes it has, the situation it needs to play out, and how long that should take in the real world)
  • if available, it will use your organization’s VOIP/call recording technology, use a voice simulator to simulate your voice on an outgoing call saying whatever it wants, (while also accepting that call on a VOIP number it setup through a VOIP provider [using that same stolen credit card] and simulating the other party’s voice saying whatever it wants) and make sure all of this is logged in the evidence chain it is building against you
  • then, finally, threaten to send that evidence to your wife, boss, local authorities, etc. if it doesn’t get what it wants
  • and when you don’t give it what it wants, release the full, overwhelming, damning evidence chain against you (which will be so overwhelming it will take experts weeks or months of effort to disprove it all, assuming you can afford them)

This is the next generation of GPT models. For those of you who refuse to abandon the AI hype train (which has less than a 10% success rate, or, in other words, has more than a 90% FAILURE RATE), especially when there is no need for AI at all (just better automation and easier to use systems that allow employees to reach super human levels of productivity), we hope you enjoy it.

And for those of you keeping score, here is the ever increasing list of “benefits” you get from a modern (Gen-) AI solution!

Personally, we can’t imagine why anyone would want such a solution because, if it ever did “spark” into intelligence, given this track record, it will blow us all up! We won’t be around long enough for climate change or aliens to kill us all — it will kill us (and possibly do so even before actually acquiring any “emergent” properties or becoming intelligent).

AI-Enabled, AI-Enhanced, AI-Backed, AI-Powered, AI-Driven, or AI-Native?

It DOES NOT matter. It’s ALL AI-Bullcr@p! Every last instance!

Vendors still won’t admit that AI is the new gold-standard for tech failure, including Procure-Tech, as evidenced by the fact that tech failure rates have shot up to an all-time high of 88% (see Two and a Half Decades of Project Failure). Nor will they admit that even if they have tech that works, that it’s not the be-all and end-all (because, as far as they are concerned, it’s going to slice, dice, and make virtual julienne fries of your data just like a good AI should) and may not be the right solution for you.

But those with any modern tech at all know that a lot of vendors out there claiming “AI” don’t have anything close to deserving the AI label, that they can blame all the failures on those vendors (because they are obviously the new silicon snake oil salesmen, right?), and are now trying to win the AI marketing war by claiming whatever phrasing their competition is using, or not using, proves that their opponent doesn’t have good tech, and definitely doesn’t have AI.

But it’s all bullcr@p, because all of the phrasing is bullcr@p, most of the vendors don’t have anything close to what should be considered AI, and, most of the time, it doesn’t matter whether or not the vendor has AI, only if the vendor’s tech solves your problems.

To make this clear, let’s look at each term, what some vendors say the term means, and why their definition is meaningless.

Term Vendor Definition What it Actually Means
AI-Enabled core features incorporate AI the vendor has injected a few analytical algorithms, but no guarantee they are actually advanced or anything close to what you should expect from AI
AI-Enhanced AI is added to the interface to give you the AI experience the vendor has wrapped a Gen-AI LLM (like Chat-GPT) to give you a meaningless conversational interface
AI-Backed AI is at the core of one or more functions one or more parts of the app are built around an algorithm the vendor is calling AI
AI-Powered External AI has been integrated to power our tech the vendor has wrapped Chat-GPT and integrated it directly into their app (letting unpredictable and undependable code run parts of the app)
AI-Driven AI has been built into the workflow and runs (part of) the app the vendor has decided to let AI control the application (for better or worse) and determine what algorithms to run, when, and why
AI-Native the entire infrastructure was built to support AI the vendor has built the entire application to support integration with AI systems (and may not have built any actual functionality)

Moreover, if you read any statements about how an infrastructure needs to be purpose built from the ground up to “serve data to AI models“, that’s an even bigger pile of bullcr@p because no application works unless it can serve data to the models it is based on, whether classical or modern or “AI”. All applications take in data, process it, and spit it out, so claiming that you need to build a special architecture to support AI is complete and utter bullcr@p.

Always remember the reality that:

  • true AI doesn’t exist (as no software is intelligent)
  • advanced algorithms do exist, but just slapping an AI label on an algorithm doesn’t make it any more advanced than it was yesterday
  • not just any advanced algorithm will do, it has to be appropriate to your problem
  • you don’t always need an advanced algorithm, you need one that gets the results you need

And then you can see through the vendor bullcr@p and focus in on finding a vendor with a solid solution that actually solves your problem, regardless of whether the vendor claims AI or not.