So DeepSeek has this very cool feature that displays what it is “thinking” before it gives you its answer. It’s quite neat in that you can see its “thought” process, but it also has the added benefit of revealing whatever bias it might have developed with its training data.
In this case, I asked it if we might be living in a “slow motion World War 3” with the Maiden Coup in Ukraine being the opening shots. The mf thought that I might “buy Russian propaganda” because I called it a coup rather than a revolution.
So although DeepSeek is Chinese, it was still very clearly trained on a lot of mainstream / 
information.
Removed by mod
The machine learning models which came about before LLMs were often smaller in scope but much more competent. E.g. image recognition models, something newer broad “multimodal” models struggle with; theorem provers and other symbolic AI applications, another area LLMs struggle with.
The modern crop of LLMs are juiced up autocorrect. They are finding the statistically most likely next token and spitting it out based on training data. They don’t create novel thoughts or logic, just regurgitate from their slurry of training data. The human brain does not work anything like this. LLMs are not modeled on any organic system, just on what some ML/AI researchers assumed was the structure of a brain. When we “hallucinate logic” it’s part of a process of envisioning abstract representations of our world and reasoning through different outcomes; when an LLM hallucinates it is just creating what its training dictates is a likely answer.
This doesn’t mean ML doesn’t have a broad variety of applications but LLMs have gotta be one of the weakest in terms of actually shifting paradigms. Source: software engineer who works with neural nets with academic background in computational math and statistical analysis
Depends on what you want to accomplish and how much resources you want to expend.
I have not discarded LLMs. I know some people use them to great effect, but one must be deeply skeptical of their use as oracles.
If you use them for their intended purpose, they can be useful, just as autocorrect is useful. I have used LLMs to great effect for helping me cut down on my word count for certain assignments, or as a psudeo-google search for coding assistance.
I am well aware that the other approaches cannot do these things. They tend to suck at language processing. However, AI architectures using explicitly coded rules have the advantage over LLMs that they are not so prone to hallucinating, which makes them safer and more useful for certain other tasks.
Not to mention that LLMs themselves were largely unviable until the creation of the attention mechanism and humanity throwing ungodly amounts of resources at them (hundreds of billions of dollars of investment).
I am aware that human brains also hallucinate logic. That’s why I don’t place must weight on random anecdotes when talking about politics or science.
What historical revisionism? The only thing my comment mentions is that inference engines did not receive as much hype or funding as LLMs, which is true. And how is anything I have stated “luddism”?
This doesn’t mean anything when all the AI companies are hemorrhaging money at an epic scale. At least the LISP bros can say that they never built the monument to the irrationality of capitalism that is the AI stock bubble.
Or maybe they did with the dot com bubble. Idk much about that period.