Episode Details

Should you strip suspected AI-generated text out of a training set? The position here is no, not blind. Slop detectors are classifiers that flag one to five percent of clean human writing as fake, and they misfire on dense expert prose and standardized legal boilerplate, so teams in financial services and legal have deleted the backbone of their corpus and watched accuracy fall. A recent experiment found filtering degraded performance more than the contamination removed.

You leave with a Monday plan: quarantine and hand-sample flagged items rather than deleting, tag provenance at ingest instead of at training time, and read the unknown-provenance estimates from dbt Labs and MIT Sloan Management Review with the vendor incentive in mind.

Key takeaways

• Run any slop detector on a batch of known-good human data first, and if it flags more than a few percent as fake, the detector is the problem.
• Quarantine flagged records instead of deleting them, then hand-review a sample of about a hundred to see whether the tool is catching slop or catching your experts.
• Stamp every source at the point of ingest with where it came from and whether a human or a machine produced it, rather than reconstructing provenance at training time.
• Treat filtering as a governance decision about unknown-provenance data, not a question of which detector to buy.
• Discount the dbt Labs figure of over sixty percent unknown-provenance data as a vendor number and cross-check it against independent work such as MIT Sloan Management Review.

Chapters
0:00 Why a filtered dataset made accuracy drop
0:50 How often detectors flag human writing
1:36 Contract summarization corpus gutted by filtering
2:26 Data provenance as a governance problem
3:28 Quarantine flagged data instead of deleting
4:00 Test the detector on known-good data

Go deeper, free lessons

• Data quality dimensions: why 'good enough' destroys trust
• Shift-left data quality: embedding governance in the engineering pipeline
• Data lineage & metadata management: knowing where your data was born
• Data observability: detect problems before your users
• LLMOps & evaluation

Full article and transcript: https://www.mba-training.com/blog/ai-slop-training-data-quality-cdo

MBA Training, mba-training.com