Primary ProcureTech Concern: (Gen-)AI Integration/Impact

The non-stop hype coming straight from the A.S.S.H.O.L.E. is continuing to cause market confusion and utter chaos.

Why?

Gen-AI is on the concerns list because it’s the tech-du-jour. Five years ago it was (advanced) (predictive) analytics. Ten years ago it was the fluffy magic cloud. Fifteen years ago it was SaaS. Twenty years ago it was the World Wide Web. And so on.

But not one of these technologies, all sold as the panacea that would solve all your woes, solved your problems because all of the promised capabilities were just silicon snake oil, and Gen-AI is no different. The hype cycle may be slowly coming to an end, but it will quickly be replaced by Some-BS-World-Model-Adjacent-Agentic-AGI that will be sold as the AI that finally solves all your problems but, in reality, still won’t be anything close (but, if narrowly applied in the right domains where the client has sufficient data might actually work quite well … but won’t do anything reliably in general and the failure rate will still be 80%+, which is the average tech failure rate for the last 25 years … and SI knows, because the doctor has been following tech failure for over 25 years).

Not only is Gen-AI no different than the previously over-hyped tech-du-jour offerings of the last two decades, but with a failure rate of 94%+ (McKinsey, and 95%, MIT), it’s arguably the worst yet! And, as per our predictions, it’s not going to get much better. If the failure rate gets as low as 90% this year, it will be the closest thing to a tech miracle that we can conceivably get. Like every other tech before, Gen-AI will only solve a relatively small set of problems.

Just like

  • The Web only solves remote connectivity
  • SaaS only allows solutions to be built in the cloud
  • Analytics only provides insight where you have the right, sufficient, data and the right algorithms to get useful insights
  • Gen-AI is just a next-gen probabilistic deep neural net that often does
    • better semantic processing
    • better search
    • better summarization
    • better potential pattern identification (but only if you can learn how to prompt it to do so and only if you have it trained on the right data subsets, not the entire web which is now more than half AI slop)

    but does so at the additional expense of

    • hallucinations
    • intentional falsehoods
    • thoughtless reinforcement
    • cognitive atrophy
    • etc. etc. etc.

As a result of this, as far as I’m concerned, the AI bubble can’t burst fast enough! It’s all hype, buzzwords, and hallucinatory bullcr@p. And, frankly, any (claims of) agentic AI built on it are fraudulent. (After all, we’ve already seen what happens when you let AI run your vending machine. The last thing you want is it buying for you!)

Especially when, on top of hallucinations, we have plenty of examples of:

We’ve said many times that LLMs are not helpful and ChatGPT (in particular) is not your friend, that if you have a headache you definitely shouldn’t take an aspirin or query an LLM, and that, frankly, you’d be better off with a drunken plagiarist intern because that’s the best case result from an LLM. Most are worse.

Frankly, it’s time to stop falling for the artificial intimidation, fight back against AI Slop, and remember cutting edge tech is NOT defined by the C-Suite or the incessant marketing from the A.S.S.H.O.L.E. that is targeting the C-Suite on a daily basis!

Impact Potential

Huge! Companies will continue to waste millions individually and collectively hundreds of billions on the next generation tech that, with a probability of 90%+, will generate a (huge) loss.

Major Challenges/Risks

The major challenge is not with the tech, it’s helping companies realize that they’re probably not ready for the tech. The reason that tech failure rate has averaged 80%+ over the last twenty years is that consultancies keep promoting, vendors keep selling, and companies keep buying advanced leading edge tech they are not ready for. The reality is that unless you are in the top 10% of buyers of tech, already on the latest tech, and have sufficiently mastered that tech, you are not ready for Gen-AI (which should not have left the research lab when it did and, in all honesty, should still be in the research lab since it still only works in a small number of well defined scenarios and is so bad that every year a couple of AI founders turn away from AI because of it — with Yann Lecun walking away from Meta and LLMs and reverting to world models, that can be thought of as next generation (Semantic) Web 3.0 models augmented with [deterministic and dependable] automated reasoning and, hopefully, very little dependence on hallucinatory probabilistic models [beyond what’s needed to semantically parse an input].)

The only place you should be using Gen-AI is where a non Gen-AI solution doesn’t exist, the task is well defined, and you can build a custom in-house model that works reasonably well in the majority of situations and that can be implemented with guard-rails. But that’s something you can only do if you have a high TQ (Technical Quotient) and have mastered last generation tech. Right now, you should be tripling down on E-MDMA and Advanced Analytics as this tech has improved to the point where it can allow you to optimize processes, spending, schedules, and anything else you can think of with high accuracy and low cost with basic analytics skills as so much comes pre-packaged and the visualizations and drill-downs are much more intuitive than they were a decade ago. Plus, these firms have figured out how to use multiple forms of AI to classify your data with high accuracy and minimize the work required by you to fix errors and reclassify to your preferred schemas. It’s literally drag and drop as compared to the complex rule-building that used to be required. In addition, you should be looking for the mature A-RPA (Advanced Robotic Process Automation) solutions that are highly customizeable and capable of “self-learning” such that the parameters that trigger exceptions will adjust over time based upon user acceptance or rejection of recommended actions and the platform will automatically encode new processing rules based upon the users’ actions on an exception. Much better than Artificial Iiocy that decides everything based on hallucinations.

THE FINAL WORD

If you haven’t mastered all of the tech that existed before Gen-AI, including classical machine learning AI that has been studied, optimized, and proven to work for over a decade, you’re not ready for Gen-AI, should treat it like the drug it is (as it does more damage to your cognitive abilities than many illegal drugs), and JUST SAY NO!

Primary ProcureTech Concern: Weakness & Volatility in Emerging Markets / Trade Wars

Emerging markets are your future markets, and often the source of critical raw materials.

Why?

Given that a lot of outsourcing has been redirected to these “low cost” markets over the past two to three decades, any rapid increase in volatility becomes a significant concern, especially if the markets are not strong enough to weather the storm. A major event could wipe out an entire subset of the supply base literally overnight, greatly increasing supply shortages and increasing the market complexity. Or at least make it unsustainable, such as a 145% tariff on China which is the source of over $500 Billion dollars in imports into the USA.

Impact Potential

The impact of a “low cost” market becoming unavailable, or at least unsustainable, is moderate to severe, especially if all of your outsourced eggs are in the same country basket. One lesson that some companies haven’t learned yet is that dual sourcing is not reducing risk if the two sources of supply are in the same country (or the same small geographic region — because if you have two factories located 100 miles from each other on two sides of a border, guess what, one natural disaster can wipe them both out).

If your primary source of affordable supply is wiped out overnight, it could take months to identify a new source of supply and quarters to secure the supply and get your supply chain flowing properly.

Major Challenges/Risks

Foreign Market Predictions
It’s hard to predict what’s going to happen in a foreign market that you aren’t in everyday. You can follow economist predictions, follow currency trends, try to get a grip on the trade relations between that country and your home country, and so on, but it’s not easy. If you can predict early enough, you can take action. But if an administration, without warning, decides to drop 100%+ tariffs on your source of supply, you’re in trouble.

Alternate Sources of Supply
Sometimes there’s few sources of supply for a given material, part, or product outside of a given country that has a similar total cost of acquisition, especially if you aren’t sourcing at full volume. Identifying alternate sources of supply that you can switch to quickly can be quite a challenge.

New Market Identification
If the emerging market also happens to be one of your primary emerging sales markets, the hit from volatility can be quite significant if the volatility results in rapid inflation, job loss, or both and your sales start to drop rapidly.

Final Words

Given the globalization of today’s supply chains, where a product can depend on materials and parts from dozens of countries, weakness and volatility in emerging markets is a significant concern. And we have yet another (fourth) reason you need an economist!

Primary ProcureTech Concern: Tightening Credit Conditions

The world runs on money, regardless of what form it comes in. Gold, cash, or credit. Credit is particularly important because it helps an organization bridge between cash cycles.

Why?

If economic downturns or inflationary pressures arise quickly, then credit will also tighten.

Impact Potential

If the organization, or its suppliers, needs credit to produce and distribute the goods for sale, the lack of interim credit could lead to reduced inventories and sales and even bankruptcies.

Major Challenges/Risks

Economic Market Prediction:
Predicting whether the economy is going to grow, stay flat, or recess (or depress) is the first challenge, as that’s a leading indicator of credit markets.

Credit Market Prediction:
Based on the projected economic changes, predicting the base and prime rate changes, availability of credit, and the future cost to your organization and your primary suppliers.

Alternative Credit Sources:
If your primary sources are projected to become considerably more expensive or restrict credit access, can you identify alternate sources? Moreover, how much will those cost, how long to establish the relationships, and how reliable will they be?

Alternative Credit Arrangements:
If right now you are just using loans or lines of credit, maybe you need to consider early payment discounts, invoice factoring, or alternative supply chain based credit arrangements.

Final Words

Credit conditions depend heavily on economic conditions, so this is yet another reason you need a good economist.

The Squirrels Have Us Right Where They Want Us!

Over the last couple of years we’ve chronicled multiple instances of squirrel sabotage and how squirrel sabotage is spreading north, the rise of the terror squirrels that have organized their own rigorous training camps, and how they are targeting us when we are at our weakest.

We’ve done this while all the major news sources have not only stayed quiet, but published articles about how cutesy the squirrels are and how a significant number of Americans are now maintaining their sanity by watching squirrel videos.

And, even worse, there is a growing number of Instagram and Tik Tok Influencers who are feeding, befriending, and even housing their own packs! This is EXACTLY where the squirrels want us! That way, when they’re ready to take back the continent, we won’t suspect a thing.

They know their time is close. A few well placed copies of Mein Kampf. Some well timed sabotage during protests and law enforcement operations. Increased stress and angst through well timed power outages. Once a revolution starts, and everyone needs to be armed, their time will be close. Then they just have to wait for mini single shot firearms to start being mass produced (as every lady will want to conceal one or two on her person, just in case), at which point, once there are millions to be stolen, they’ll organize their operation to clean out entire warehouses overnight (since they are small enough to get in and out without anyone knowing).

And we won’t suspect a thing because they’ve been the critters helping to keep us sane with their cutesy acts and subliminal messaging. (The whole point of Squirrel with a Gun was to show us how insane the idea of a squirrel with a full size gun was and ensure we never suspected them of being capable of mass violence. However, tiny derringers come in around 4″ in length and 4″ in height, with the tiniest being about 3.7″ and 2.4″ (like the NAA-22S). Small enough for a squirrel, big enough to take out even the most hardened human (when they sneak up and fire a shot at our temples in close range). And since there are at least as many squirrels as there are of us …

There is NO Infinite Compression – The Latest DeepSeek Paper is BullCr@p!

Every decade or so, some idiots who never studied Huffman coding or Information Theory believe they have cracked the problem of infinite compression, and this linked paper is just the latest example of this lunacy. I really hope this was a joke paper authored by AI because it’s all bullcr@p!

On average, a text token in a LLM should require 20 bits or less (as 17 bits support a 129,000 word vocabulary) while a vision token can be 16,384 bits (based on 1024 dimensional continuous vectors) — because it takes a lot of bits to represent pixelation of a square in a 2-D image! This says you can store about 820 text tokens in the same space it takes to store one vision token. Or, you can store the entire text (lossless) in 48K, versus the 4M it would take to store the 250 vision tokens (using very lossy compression) that are required in the paper. Looks like a LOT of people can’t do basic math if this is being praised as revolutionary!

Moreover, the raw text, which maintains the full context if the tokens are kept in order, is not only fully lossless, but can be compressed using a modified Lempel-Ziv algorithm to take up an average of less than 2 bits per character (and achieve up to an 80% compression rate). Given that the average length of a word in average text is 5 characters, and a space is one character, 2500 words would be 15,000 characters, storable in 30,000 bits or a mere 4K! In other words, this paper is trying to pass off a ONE THOUSAND FOLD increase in space requirements as space saving! Pure lunacy!

In other words, if someone is claiming something too good to be true, it is! Don’t fall for it or the sure to follow claims that DeepSeek OCR is revolutionary because of this. (Since every document is different, you can’t imagine the true loss with a 90% vision token reduction!)