Oct 8, 2026 · 6 min · 4 segments
What are small models in AI, and why should business owners be interested in using small models? Jay Smith, Co-Founder and AI Engineer at 10x Velocity, which specializes in small models, shared his…
eh.
One of our recent podcast guests works a lot with AI and governments, and so pretty strict data privacy standards that have to go by.
Apparently one of the, I mean, you could call it a loophole, just sort of like technicality of AI privacy policies is altering your data to anonymize it can actually be fairly miniscule in terms of what the alterations actually are, like changing just a letter within a data set can satisfy legally the anonymization standard.
So for business owners out there, even if a major AI company is saying, hey, we're anonymizing your data, we're altering it, it might not be that dramatically altered.
So it looks for first names, last names, social security numbers, credit cards, birthdays, addresses, custom fields that you create.
And then they go to the model with these placeholders and then it comes back and it does what's called rehydration where it puts the name back in or the birthday back in.
But again, if you completely redact everything and send it to the model, the answer you get is not going to be very high quality.
eh.
One of our recent podcast guests works a lot with AI and governments, and so pretty strict data privacy standards that have to go by.
Apparently one of the, I mean, you could call it a loophole, just sort of like technicality of AI privacy policies is altering your data to anonymize it can actually be fairly miniscule in terms of what the alterations actually are, like changing just a letter within a data set can satisfy legally the anonymization standard.
So for business owners out there, even if a major AI company is saying, hey, we're anonymizing your data, we're altering it, it might not be that dramatically altered.
So it looks for first names, last names, social security numbers, credit cards, birthdays, addresses, custom fields that you create.
And then they go to the model with these placeholders and then it comes back and it does what's called rehydration where it puts the name back in or the birthday back in.
But again, if you completely redact everything and send it to the model, the answer you get is not going to be very high quality.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.