Mediocrity

Several important things to unpack here.

1. credits to Microsoft for doing serious work on what otherwise is just a mediocre concept of computer intelligence called LLMs.

2. credits to Microsoft for stating today what to me is obvious since back when “AI” was still called big data: «There is absolutely no guarantee that a LLM’s output accurately reflects the source data you’re interested in». Which is reason enough to call a lunatic to anyone using these systems.

3. the reason why it is obvious that LLMs are not reliable, the same as any other statistical approach (including another engineer’s favourite masturbation called “baesian reasoning”), is simple: these systems translate raw real data into a finite, comparatively small set of parameters, therefore they lose information which cannot ever be reconstructed to the original.

4. however this admission comes at least a decade and billions of dollars AFTER these actors have been promoting the fake statistical emulation of intelligence called LLMs; they should be demanded to do better based on the resources they have.

5. As a society we cannot afford to waste such vast resources to deliver billion dollars garbage during 10 years and afterwards start pretending that it is better to clean the garbage instead of doing better scientific research from scratch.

6. so those engineers that have been cheering the next big thing in “AI” should all be labelled as delusional at best and psychotic at worst; totally unfit to perform any duty in exact sciences

7. TWO YEARS AGO I demonstrated how you can analyse large quantities of text in a scientific way, in a deterministic fashion, not probabilistic. This was the basis model for the first proper, scientific reasoning system and it shows how my approach is ahead of the best engineering since years ago.

Now go back to playing with your decade old toys. You are already outdated and each passing day you are thrown further back into the dustbin of History. And I am making sure that you will stay that way.