Building AI requires a model of the MIND. The critical question is what kind of mind we talk about.
Here is a practical, technical approach, refined over decades under real and delivered projects. (For nerds like me: "mission statement", videos, GitHub repo.)
Here is a practical, technical approach, refined over decades under real and delivered projects. (For nerds like me: "mission statement", videos, GitHub repo.)
𝐌 odel𝐈 ntellect𝐍 arrative𝐃 ialog
However, there is another alternative, which has nothing to do with solving real problems. That is efficient economic and social engineering, the one that wins in a "mature consumer society" (a.k.a. Idiocracy). If you think of actual people reading these words, or that history seems to repeat itself, that is not a coincidence.
𝐌 oney𝐈 diotism𝐍 arcissism𝐃 ogmatism
I guess you would like this one, Axel C. 😉
Brad Hutchings
This post might be the perfect antidote to one just above on my timeline noting that just a week after Sam declared a "code red", which I'd never before heard of as a real thing in a tech company in 35+ years of doing this, OpenAI shipped ChatGPT 5.2, which leads all known and imagined benchmarks again. So thank you for this.
Lorand Kedves
Brad Hutchings Those benchmarks... 🤣 🤣 🤣 Chanting around a black box for a "machine god" to emerge, instead of doing proper research and engineering based on existing but forgotten white box models. Just as predicted BTW...
I think my question that I sent to these overhyped kids a decade ago is still legit, and they have no better answer than the silence I got back then:
I would ask you a silly question: what is your definition of "intelligence"? No need to give links to AI levels or algorithms, I have been on the field for 20 years. I mean "intelligence", without the artificial part, "A" is the second question after defining "I". At least to me :-)
Also, to clarify those words for the record 🙂
Money - the circular financing scheme (same old bubble, has nothing to do with AI)
Idiotism - by Neil Postman https://youtu.be/YtjjFmCxc8s?t=547
Narcissism - no judgement, this is expected and enforced by the cult members
Dogmatism - that is the real issue https://youtu.be/8pTEmbeENF4?t=1791
It's just unfortunate that this second "mind" destroys my field of interest, so I sometimes vent a bit of steam here. 🤷♂️
Thanks for the response 🙏
Lorand Kedves
Brad Hutchings Unless those people are dead and forgotten and were far ahead of you, and you see how and why they failed. I tried to publish an article about some of them as part of my CS PhD (before the "chatbot revolution") It was of course rejected but here it is in case you are interested: The Science of Being Wrong.
It's hard to "keep the faith"...
Nice to meet you! 🙂 🖖
Brad Hutchings
Lorand Kedves I feel a Bon Jovi song battle brewing. But it is late here. So... Have a nice day! 🤣
Simon Gant
id suggest a mind thats done the work......IYKYK
Lorand Kedves
Simon Gant By my model, that would mean a self-aware, -testing and -improving knowledge graph that can (and would) "know anything".
I explained the "secret sauce" to the OpenAI folks when they were not millionaire demigods but the puppets of Musk desperate to come up with something more interesting than a NN that can play games... 🤣
I don't know any attempts tested in delivered systems better than my ones, so "if you know, I want to know"... 🙏
That has nothing to do with this LLM / GenAI freak show.
Veteran engineers and researchers (like myself with 25+ years, a whole life in this arena) don't play with dreaming machines (ref Andrej Karpathy et al) or stochastic parrots. OK, ok, our company built one in the 90's, "Prody", a precursor of the infamous Clippy... 🤭 Don't blame me, I built decision trees, hypercubes and an agentic runtime.
But that should not be new...
for anyone who knows the warnings of the pioneers, like a certain A. M. Turing... I highlighted the apparently forgotten parts and work on the real challenge he proposed. Define the terms "machine" and "thinking" before making a statement that contains them, aiming at the goal set by Douglas Engelbart.
Augmenting Human Intellect.
H-LAM/T... IYKYK 😉
I explained the "secret sauce" to the OpenAI folks when they were not millionaire demigods but the puppets of Musk desperate to come up with something more interesting than a NN that can play games... 🤣
I don't know any attempts tested in delivered systems better than my ones, so "if you know, I want to know"... 🙏
That has nothing to do with this LLM / GenAI freak show.
Veteran engineers and researchers (like myself with 25+ years, a whole life in this arena) don't play with dreaming machines (ref Andrej Karpathy et al) or stochastic parrots. OK, ok, our company built one in the 90's, "Prody", a precursor of the infamous Clippy... 🤭 Don't blame me, I built decision trees, hypercubes and an agentic runtime.
But that should not be new...
for anyone who knows the warnings of the pioneers, like a certain A. M. Turing... I highlighted the apparently forgotten parts and work on the real challenge he proposed. Define the terms "machine" and "thinking" before making a statement that contains them, aiming at the goal set by Douglas Engelbart.
Augmenting Human Intellect.
H-LAM/T... IYKYK 😉
Francis Y.
Lorand Kedves what 'white box' models do you propose that can anywhere near match the effectivness of a transformer?
Lorand Kedves
Francis Y. There is a total mismatch here.
I talk about transparent and reliable information systems (old school architecture / engineering spiced with machine learning and stretched towards Chomsky's generative grammar, a top-down approach). Not those that are "good enough to fool me on areas where I am not an expert", sorry for a less enthusiastic definition of chatbot / genai (that starts with Hinton's ignoring informatics, rejecting Chomsky and going bottom-up: build a brain simulator and if it is big enough, it will become "intelligent", without even defining this word. Nobel prize science indeed...)
The MIND model addresses the question of what thinking is, in a way aerodynamics addresses the question of flying (Turing challenge one: what is 'thinking'). This is the way you can build a jet plane that somewhat resembles to, yet totally different from a bird (Turing challenge two: what is a 'machine').
So the question is rather: what 'a transformer' has to do with a white box model of thinking? I don't have answer to this one, do you?
But I think that they go in different directions, and if that's correct, the question of 'comparing effectiveness' is meaningless.
