Soshu Sadashige?

[eternal_newbie]

I was browsing through some old blog posts on the Tozando website's blog, and found this article on the Soshu-den:

 

https://japanesesword.net/blogs/news/the-soshu-den-exploring-the-pinnacle-of-Japanese-swordsmithing

 

Strangely, the article mentions on several occasions a legendary and historically significant Soshu smith - in fact a whole line of them - named "Sadashige", ranking him alongside Shintogo Kunimitsu, Masamune and Norishige.
 

Quote

4.3.1 Sadashige's Contribution to Soshu-den

Several smiths bore the name Sadashige throughout Japanese sword history. In the context of "Legendary Soshu-den Swordsmiths" and "Masterworks," this often refers to smiths who either were direct students of major Soshu masters like Sadamune (Masamune's leading student) or worked within the Soshu tradition, embodying its core principles. A Sadashige, for instance, is recorded as a student of Sadamune, continuing the direct lineage and artistic legacy. Their blades, while perhaps less widely known than Masamune's, still represent the pinnacle of Soshu craftsmanship, characterized by:

 

• Hamon: Active and brilliant *nie-deki* *hamon*, often featuring *notare* and *midare* patterns, sometimes with a more subdued yet equally artistic expression compared to Masamune's boldest works.
• Jigane: Well-forged *itame* or *mokume* *jigane*, rich in *ji-nie*, demonstrating the deep understanding of steel quality and forging techniques inherent to the Soshu school.
• Sugata: Robust and well-balanced forms, reflecting the practicality and aesthetic sensibilities of the Soshu-den.
   

These smiths ensured the continuation and subtle evolution of the Soshu-den style, producing blades that are highly valued for their adherence to the school's high standards.

 

Is this a translation issue where AI (or human error) accidentally substituted "Sadashige" for "Sa" or "Nobukuni"? Or is there a whole chapter of Soshu-den that I've somehow missed in my years of study?

[PNSSHOGUN]

Most of these articles appear to be AI generated, wouldn't put much faith in any of the details offered. 

[eternal_newbie]

Just now, PNSSHOGUN said:

Most of these articles appear to be AI generated, wouldn't put much faith in any of the details offered. 

 

I figured that might be the case, just wanted a second (human) opinion to be certain. Hallucinating an entire Soshu grandmaster and his inheritors out of thin air does sound like something ChatGPT would do.

[nulldevice]

Hallucinations are a fascinating part of generative AI and one that can quickly lead down paths of nonsense if one is not careful. In a world of specialized terminology and weird habits (nihonto is very guilty of this), this becomes even more common. The resources from which ChatGPT or whatever model you were using no doubt are referencing other AI generated articles rife with their own hallucinations.

 

We're starting to see AI-generated "educational" articles pop up with references to nihonto on non-nihonto (typically repro sword sales) sites with a mish-mash of correct and incorrect information. The problem is is that nobody will check these for historical accuracy and the next batch of AI-generated articles will use these as references and sources of truth when making the next article thus further tainting the pool. I remember in High School when teachers would always tell us "Wikipedia isn't a legitimate source, you must go deeper to the source material, journals, etc. and conduct your research there". The same is only more true today in the internet of generative-AI, only a hundred-fold more so. Due diligence is key and in fact, the old books might be making a comeback as sources of wisdom upon which we can study and learn. 

[Robert S]

13 minutes ago, nulldevice said:

Hallucinations are a fascinating part of generative AI and one that can quickly lead down paths of nonsense if one is not careful. In a world of specialized terminology and weird habits (nihonto is very guilty of this), this becomes even more common. The resources from which ChatGPT or whatever model you were using no doubt are referencing other AI generated articles rife with their own hallucinations.

 

We're starting to see AI-generated "educational" articles pop up with references to nihonto on non-nihonto (typically repro sword sales) sites with a mish-mash of correct and incorrect information. The problem is is that nobody will check these for historical accuracy and the next batch of AI-generated articles will use these as references and sources of truth when making the next article thus further tainting the pool. I remember in High School when teachers would always tell us "Wikipedia isn't a legitimate source, you must go deeper to the source material, journals, etc. and conduct your research there". The same is only more true today in the internet of generative-AI, only a hundred-fold more so. Due diligence is key and in fact, the old books might be making a comeback as sources of wisdom upon which we can study and learn. 

It's actually a really interesting point that the only reliable sources may soon be hard copy books published before the advent of AI.  I don't remember the title of the book, but there was a science fiction book which considered this possibility - it was called the infocalypse.  Here we are living it!

[nulldevice]

7 minutes ago, Robert S said:

It's actually a really interesting point that the only reliable sources may soon be hard copy books published before the advent of AI.  I don't remember the title of the book, but there was a science fiction book which considered this possibility - it was called the infocalypse.  Here we are living it!

There are still useful and incredible uses of AI, even within niche hobbies. That being said, most content now days that is being AI-generated isn't done with educational or research purposes in mind. Its a way to blast the internet with content to drive up engagement, SEO ratings, and other general bloat while adding net negative to the information we have and wasting an enormous amount of energy, resources, and space to fill the web with useless nonsense. 

 

On the other hand, there are plenty of situations where the pattern recognition, pseudo-logical processing (the best kind because we can still give it rules to follow but it has "intuition" which allows much more complex pattern matching than one could write in a standard program), and the ability to be constrained allow it to be a phenomenal tool in the right hands with the right intents. 

[Robert S]

2 minutes ago, nulldevice said:

There are still useful and incredible uses of AI, even within niche hobbies. That being said, most content now days that is being AI-generated isn't done with educational or research purposes in mind. Its a way to blast the internet with content to drive up engagement, SEO ratings, and other general bloat while adding net negative to the information we have and wasting an enormous amount of energy, resources, and space to fill the web with useless nonsense. 

 

On the other hand, there are plenty of situations where the pattern recognition, pseudo-logical processing (the best kind because we can still give it rules to follow but it has "intuition" which allows much more complex pattern matching than one could write in a standard program), and the ability to be constrained allow it to be a phenomenal tool in the right hands with the right intents. 

I agree that currently AI can be useful, because currently AI are trained on and referencing almost entirely human generated information.  But this begins to change as you get an acceleration of AIs learning from AI generated information, including hallucinations.    Without safeguards this has the potential to create an exponential propagation of unreliable information, to the point where it becomes unknowable whether or not a specific piece of digital information is reliable.  Then new systems of truthing become important - either that the information can be traced back to a pre infocalypse source, or that there is a highly developed reputation checking element attached to every piece of information.