I feel like I live on a different planet to many on HN.. Any time I've dabbled with the current roster of LLMs for work tasks (I'm a game programmer). They are utterly useless, complete waste of time. Definitely not something that seems promising and warrants more time invested.
I always point this out too. They had a war of independence, the main goal of which was so the elite slave owning class wouldn't have to pay tax. Yet they call it a revolution.
They have never had an actual revolution akin to French revolution and the July revolution.
When you look at how molecules like RNA work, and krebs cycle, and the billions of cells we are composed of, and so on, it always strikes me as astronomically lucky that we function at all. Like how can this assemblage of Rube Goldberg machines function for more than 1 seconds without catastrophically falling apart?
I think multicellular creatures on earth are just so complex they are basically ineffable.. We can understand certain general principles and statistical trends, but the entire system holistically is incomprehensible for a human level intelligence.
Kind of analogous to ML, we absolutely understand how each neuron works, we built them! But we often dont really understand how the resulting model works.
What's happened in software / computing in the last 20 years that's good? Imo it could be argued that overall the user experience has gotten worse. Dead internet theory, enshittification.
* The web is pretty much dead. Time Berners Lee's ideals certainly are.
* Computing is dominated by completely evil megacorps.
* They are making a concerted effort to make people as tech-illiterate as possible and also make universal computing illegal.
* Theres been years where GPU's were being price gouged, 1st by crypto bros, then NFT bros, now LLM bros.
* Cant even buy RAM now.
* They put e-fuses into hardware now, comes right out of the factory as ready made e-waste that cant be repurposed.
* The biggest platforms, Android and iOS, are walled garden, locked down, corporate nightmare worlds. And there is practically no alternative.
* Social media is making people depressed and also very easy to manipulate en-masse by anyone willing to pay.
* Moore's law stopped and software bloat overtook performance gains.
* VR might have been cool but it was pre-enshittified in its nascent stages. Freakin' facebook bought Oculus before they had released a single headset.
Yeah that is good, there are some good things, but I think on average the computer user experience is worse.. In the 80's and 90's with the rise of microcomputers and the net and the web and Linux, things were so utopian, things just seemed to keep getting cooler and computing was empowering people.
Apparently one reason cars from the 90's last longer than new ones (which almost always fail in some way immediately after warranty expires) is the advancement and increased usage of computer modelling / simulation.
In the 90's they had pretty much mastered car manufacturing and made parts which they were certain would outlast the warranty, erring on the side of caution they mostly ended up making parts that lasted much longer than the warranty.
Now, with computer modelling and simulations, they can accurately design a part to be as cheap as possible to make while being just durable enough to last for the duration of the warranty.
D4A did a good video on it.. https://youtu.be/SeMZGICNSMg?si=sideQIwNBr9s9QW6
Even in the LLM dept: LLMs are the most general AI systems to date, and the performance only ever goes up.
Whether adding +5% per model release is enough to get a broadly superhuman system remains to be seen. But my take is that there's no such thing as "not working on AGI" in the frontier labs. Everything that's being put into modern frontier systems is AGI groundwork, one way or the other.
Claims of AGI imply that LLM's have intelligence. They don't, they are fancy probability machines. They don't THINK the way we do, they just do 200 matrix multiplications until their training data is massaged into what you need. They don't dream, they don't remember what you tell them. Even if you write one sentence, 'attention' means they will ignore half of what you say and key in on the wrong thing. This just happened to me today on a frontier model.
I'm not saying that AGI is impossible, but the focus on LLM's is probably not the right approach. I don't think we will ever make it until we understand the human mind better.
Do you think your brain doesn't do a type of gradient descent, trying to fit its little predictive algorithms to its senses? Do you think you aren't a fancy probability machine with overinflated self-esteem?
An average LLM of today has better reading comprehension than an average human, and the gap only grows release to release.
"Understand the human mind" turned out to be a distractor. The bitter lesson won: you can take a "good enough" AI architecture, burn a shitton of data into it with an unholy amount of training compute, and get halfway to AGI - no "understand the brain" required. LLMs are so fried in imitation learning on human-generated data they even inherit humanlike failure modes.
I mean I can write a non-llm program that "has better reading comprehension than an average human" depending on what you think reading comprehension means. Today I went to ask an LLM some very simple questions, stuff you can google and "do these lines have X word in it" and it failed to answer pretty spectacularly several times in a row, so I'm just not feeling the LLMs are superior intellgence today.
reply