[ExI] Why AI agents invent their own language if you let them chat.

[spike at rainier66.com]

-----Original Message-----
From: extropy-chat <extropy-chat-bounces at lists.extropy.org> On Behalf Of Adrian Tymes via extropy-chat

On Wed, Oct 7, 2026 at 6:22 AM John Clark <johnkclark at gmail.com> wrote:

>> No. Appeals to authority are NOT always fallacies.

>...Yes they are.

>...Authorities may be a useful heuristic, such as if you can't repeat the experiment...  Adrian


Have you noticed how AIs are being treated as arbiters of truth?

Skip the next paragraph if you are in a hurry.  

Do perform that experiment I suggested a few weeks ago: find a topic about which you know more than anyone, about which you have opined online in the past.  We all have them.  In my case it is applying a technique for superposition of probability distributions to predict the time interval in which the record for the largest known prime number will be discovered, and how the model leads to a paradox if viewed analogous to a Reimann sum.  Over time, the elements on the time axis get narrower and the sum includes more elements.  If we look at it as analogous to a Reimann sum, the "integral" should get more accurate.  Instead it gets less accurate.  Ernst Mayer debated this topic online 30 years ago.  Earnst was right, from the mathematician's point of view.  I made the "money" from the engineer's point of view.  

OK then, if you AI Mersenne primes, AI gives great informative answers, very educational at a top level.  But if you drill down to the level of detail in the paragraph above, you get down to words written by Ernst and me.  I soon realized I am talking to myself.  If you keep asking questions at that level or further down into the weeds, you get nonsense, because no one ever took up that field of research.  

Conclusion: AI can write a masters thesis now, but cannot write a PhD thesis for you.  Yet.

spike