Why we stopped using the mathematics that works
gfrm.in
gfrm.in
> In 2012, Alex Krizhevsky submitted a deep convolutional neural network to the ImageNet Large Scale Visual Recognition Challenge. It won by 9.8 percentage points over the nearest competitor.
Maybe there’s another definition of “works” that’s implicit and I’m not getting, but I’m struggling to picture a definition relevant to the history-of-deep-learning narrative they are trying to explain.
I'm not closed to it. You can check my comment history for frequent references to next-generation AIs that aren't architected like LLMs. But they're going to have to produce an AI of some sort that is better than the current ones, not hypothesize that it may be possible. We've got about 50 years of hypothesis about how wonderful such techniques may be and, by the new standards of 2026, precious few demonstrations of it.
Quoting from the article:
"Within five years, deep learning had consumed machine learning almost entirely. Not because the methods it displaced had stopped working, but because the money, the talent, and the prestige had moved elsewhere."
That one jumped right out at me because there's a slight-of-hand there. A more correct quote would be "Not because the methods it displaced had stopped working as well as they ever have, ..." Without that phrase, the implication that other techniques were doing just as well as our transformer-based LLMs is slipped in there, but it's manifestly false when brought up to conscious examination. Of course they haven't, unless they're in the form of some probably-beyond-top-secret AI in some government lab somewhere. Decades have been poured into them and they have not produced high-quality AIs.
Anyone who wants to produce that next-gen leap had probably better have some clear eyes about what the competition is.
I agree.
The MYCIN system was rather good at medical diagnostics and like other systems of the time had an ad-hoc procedure to deal with uncertainty which is essential in medical diagnosis.
The problem is that is not enough to say "predicate A has a 80% of being true" but rather if you have predicate A and B you have to consider the probability of all four of (AB, (not A) B, A (not B), (not A) (not B)) and if it is N predicates you have to consider joint probabilities over 2^N possible situations and that's a lot.
For any particular situation the values are correlated and you don't really need to consider all those contingencies but a general-purpose reasoning system with logic has to be able to handle the worst case. It seems that deep learning systems take shortcuts that work much of the time but may well hit the wall on how accurate they can be because of that.
https://en.wikipedia.org/wiki/Drools
is pretty good as is the Jena rules engine but none of these have ways of dealing with uncertainty which are necessary if you're going to be working with language and having to decide which of 10,000 possible parses is right for a sentence. People used to talk as if 10,000 rules was a lot but handling 2 million well-organized rules with Drools is no problem at all today.
I think the problems of knowledge base construction are overstated and that a lack of tools are the problem. Or rather, the Cyc experience shows that rules are not enough, that is, after Lenat died it got out that Cyc didn't just have a big pile of facts and rules and a general reasoning procedure but it had a large database of algorithms to solve specific problems. That is, in principle you can solve anything with an SMT solver but if you actually try it you'll find you can code up a special-purpose algorithm to do common tasks before the SMT solver really gets warmed up.
Part of the production rules puzzle is that there never was a COBOL of business rules rather you got different systems which took different answers to various tricky problems like how to control the order of execution when it matters, how to represent negation, etc.
Without commenting on the merit of the claims, the problem with this statement is that in many cases there is no universal "technical superiority", only tradeoffs. E.g. Betamax was technically superior in picture quality while VHS was technically superior in recording time, and more people preferred the latter technical superiority. When people say that the techinically superior approach lost in favour of convenience, what really happened is that their own personal technical preferences were in the minority. More people preferred an alternative that wasn't just "good enough" but technically better, only on a different axis.
Even if we suppose the author is right that his preferred approach yields better outputs, he acknowledges that constructing good inputs is harder. That's not technical superiority; it's a different tradeoff.
It's possible a majority of pofeople would have been marginally happier with betamax than vhs. Even in that case, vhs can still win because a minority of people had a strong, stubborn preference for it, even if a majority of people had a weak preference for betamax.
If 1,000,000 people are willing to pay $5 more for video quality but 800,000 people are willing to pay $8 more for longer recording, which wins out?
Not to mention savings on the producer side are relevant too, not just consumer side.
I'm not saying the above is necessarily the case. Just pointing out that markets aren't majoritarian, they're utilitarian.
(Both got more recording times through Long Play techniques a.k.a. quality degradation and through actually longer magnetic tape in the cassette, but at least in the beginning it was clear-cut).
I was talking somebody through Bayesian updates the other day. The problem is that if you mess up any part of it, in any way, then the result can be completely garbage. Meanwhile, if you throw some neural network at the problem, it can much better handle noise.
> Deep learning’s convenience advantage is the same phenomenon at larger scale. Why specify a prior when you can train on a million examples? Why model uncertainty when you can just make the network bigger? The answers to these questions are good answers, but they require you to care about things the market doesn’t always reward.
The answer seems simple to me - sometimes getting an answer is not enough, and you need to understand how an answer was reached. In the age of hallucinations, one can appreciate approaches where hallucinations are impossible.
In particular, please show me a worked example of a decision tree meta learning. Because its trivial to show this for DNNs.
Even for computer science, take a look at Turing Award from 1966 [0], we will see how short sighted we are if we only follow the trend. Time will tell and smart people will find new path.
I don't know if I entirely agree with the article, but it has some food for thought.
QWERTY has many variants, and every single geopolitical institution have their own odious anti-ergonomic layout, it seems. So this case is somehow different to my mind. As a French native, I use Bépo.
You'll then find lots of "Blah blah blah: blah blah blah." Ten of the sentences in this article are of that form.
Then there's of course "it's not x but y". It avoids that exact construction, but is still plentiful in the article.
> What happened next was not a reasoned evaluation of competing paradigms. It was a gold rush.
> Not because the methods it displaced had stopped working, but because the money, the talent, and the prestige had moved elsewhere.
They also got very tiresome long before fading away.
> I’ve spent the last few months building agents that maintain actual beliefs and update them from evidence — first a Bayesian learner that teaches itself which foods are safe, then an evolutionary system that discovers its own cognitive architecture. Looking at what the industry calls “agents” has been clarifying.
> What would it take for an AI system to genuinely deserve the word “agent”?
> At minimum, an agent has beliefs — not hunches, not vibes, but quantifiable representations of what it thinks is true and how certain it is. An agent has goals — not a prompt that says “be helpful,” but an objective function it’s trying to maximise. And an agent decides — not by asking a language model what to do next, but by evaluating its options against its goals in light of its beliefs.
> By this standard, the systems we’re calling “AI agents” are none of these things.
LangChain… Now that’s a name I haven’t heard in a long, long time..
Anyway, that’s a cool idea. But also his blog posts include phrases like “That’s not intelligence, it’s just <x> with vibes.” Urg. Slop of the worst sort.
But, like I said, I like the idea of keeping a running tally of what tool uses are useful in which circumstances, and consulting the oracle for recommended uses. I feel slightly icky digging into the code though; there’s a type of (usually brilliant) engineer that assumes when they see success that it’s a) wrong, and b) because everybody’s stupid, and sadly, some of that tone comes through the claude sonnet 4.0 writing used to put this blog together.
You know people actually write like that. The LLMs learned it from somewhere.
Join the crowd dude. It's still true, no matter how inconvenient it is.
If that really isn't an option, then yes ML/AI isn't for you in this case.
The 10,000 hours thing is encouraging because the amount of effort you put in as far more important than your natural ability.
... Until you get to the point where everyone is already working as hard as humanly possible, at which point natural ability becomes the sorting function again.
There is nothing particular that suggests this is infinitely scalable.