sheffield

vp eng photographer, time sculptor.
slowly self-destructing, just like my tweets.

[sheffield @sheffieldworks]

saw the odyssey last night, and my primary thought is that almost no one that comes into my twitter feed knows a damn thing about movies

[kirsten lum @kirsten_lum_]

It can take months to perform the preparation steps that underlay the mind-boggling numbers quoted in this post.

https://t.co/j8vuL8wNKx has reduced that to a week and change, and getting shorter every day.

[kirsten lum @kirsten_lum_]

So if semantic objects aren't schema objects like tables and columns, what are they?

In short, they are units that link a concept in the real world to data in the database when those concepts need to be modeled as logic rather than as a column.

You heard that right... 1/n

[kirsten lum @kirsten_lum_]

Schema vs Semantic Layer vs Context Layer

The semantic layer isn’t just a layer of semantic information on top of your schema (database objects).

It’s also new layer of objects that do not neatly “fit” within the rigid confines of data structures.

Instead, they align to reality, to the nouns and their characteristics, and how they interact via verbs.

The objects are organized by an ontology, and are operationalized in the semantic layer.

And all of these are extensively described for humans and machines in the context layer.

Data is in the schema. Semantic objects are defined in an ontology and operationalized via a semantic layer that binds semantic objects to schema objects. And all of these things, their characteristics and proper usage, are described in the context layer.

That is why our tool is called https://t.co/TBEWa1IYzH. It’s where the schema meets the semantic.

[kirsten lum @kirsten_lum_]

Bad news: AI in data doesn't work without context.

More bad news: Context does not exist because humans can't stand to constantly update it, so the context layer must be automated.

Good news: It already is. https://t.co/mOiNtxsnCi

[sheffield @sheffieldworks]

the super hard relational benchmark: a benchmark that ai can't solve on its own.
we put everything difficult about enterprise data into a single benchmark, and we're building what we hope is the most effective way to automate getting the context layer enterprise ai needs.

[kirsten lum @kirsten_lum_]

Enterprise data warehouses are where AI performance claims go to die. Existing benchmarks hand models the metadata they need, like foreign keys, column descriptions, evidence hints, and test little more than the ability to write correct SQL.

[sheffield @sheffieldworks]

every shopify store somehow knows the real reason that testosterone levels have tanked

[sheffield @sheffieldworks]

the curse of intellectuality is the unconscious belief that, because someone explains something simply, it's a simple conclusion to reach on your own

[ThePierian @pieriangallery]

See you Thursday. @sheffieldworks

[sheffield @sheffieldworks]

and @elonmusk won’t pay alimony

[The Babylon Bee @TheBabylonBee]

Judge Determines Trump Will Get The Nation On Weekdays, While Musk Gets Every Other Weekend And Holidays https://t.co/TimbPnV2Lc

[sheffield @sheffieldworks]

middle aged woman
walking through the airport
wearing sweats.
her t-shirt says, “champagne is my spirit animal”

[sheffield @sheffieldworks]

the mountain
melts in the sun
exposing its rust

[sheffield @sheffieldworks]

only indifference is free

thomas mann

[sheffield @sheffieldworks]

100% genius of course

[sheffield @sheffieldworks]

petition to rename https://t.co/Ltue5WOJ1I to https://t.co/1wIPCRijg9

[sheffield @sheffieldworks]

petition to rename https://t.co/Ltue5WOJ1I to https://t.co/1wIPCRijg9

[kepano @kepano]

technology always finds the next reproducible thing

art always finds the next irreproducible thing

[sheffield @sheffieldworks]

this is what i've been working on for the last year

[sheffield @sheffieldworks]

it's frustrating how many people "succeed" in life by simply copying what others are doing

that's not how variance works.
you're aiming low, you just don't understand it.