- xpltBecause it's awkwardly close to the letters AGI, maybe
- > You open your laptop and begin.
FizzBuzz in an on-site interview on your personal laptop?
- I especially love the part of "instant responsiveness" where a UI changes on interaction an then just shows placeholders while actually fetching data for 2-5 seconds.
Otherwise, snappy interfaces are kinda nice.
- /irony ?
Honestly curious -- does your mileage vary that much on LinkedIn? It appears as though the feed is mostly a rolling dumpster fire + couple invites every week.
- > I’m assuming these days AWS is just fleecing the big companies who feel like there’s only one option.
My employer is a customer at a scale that AWS calls us strategic industry or something. Some years ago, the young engineer in me was all hyped about the range of serverless, managed services being very accessible in one place. Piecing together those resources felt very unixy. I bought into the coolness so much, I could hardly believe such tech was actually used in my industry.
I could've known better all along. Any of our solutions would work just as fine in an open-source stack. And in the bigger picture -I moved to a platform-focused role- it doesn't even matter that much. Two-thirds of our consumption is 24x7 EC2 on-demand. For sure, many projects in the company are wasting a lot by simply holding it wrong. But even if not, we're also left wondering
> Why would anyone pay the crazy high AWS prices?
and decided to do something about that.
- This. I've been in a similar situation working as a data scientist for 2.5 years and went back to software engineering a couple of years ago.
I'm not trying to say that data science and analytics are necessarily bad environments; I just came to realize that I had different expectations for my work than my organization.
While the skills of a software engineer — e.g. quickly comprehending technical references, operating APIs, ability to type more than three lines of code straight — are highly valuable for a data analyst to be productive, I had to realize that my organization did not appreciate the craftsmanship of producing code as much as I still do.
Over time, I had to witness my analyses end up on slides or in Excel workbooks, only to be looked at once. They'd done their job and weren't needed anymore. I was effectively "prompt completing" analytic requests from management to understand the organization/business/whatever better — always with the same result for "my work".
Providing an organization with the intelligence to understand their business is for sure not a bad motivation and can be fulfilling. It's just not a good fit if — instead of the analytics — you consider the software you create for the analytics as your work.
- @nineteen999 makes a valid point, but I don't think this is a case of protectionism.
From what I can tell, it's more about RH's ludicrous licensing policy and the economic practicality of using something "application binary compatible." (•_• )
- I'm really not sure how to feel about the fact that, meanwhile in Germany, RH has made its way onto the non-preferred vendor list of at least one large enterprise in the automotive industry.
- Well played
- On point.
Just because the L2-norm yields the same rankings as cosine similarity for the particular case of normalized embeddings when retrieving relevant documents doesn't mean that any other L-norm or commonly used measure in the field of (un)supervised learning or information retrieval presents itself as a viable alternative for the problem at hand — which, by the way, had to be guessed too.
Looking at the "history" of this development (e.g. bag-of-words model, curse of dimensionality etc.) provides a solid explanation for why we've ended up using embeddings and cosine similarity for retrieval.
Though, I'm curious to see any advancements and new approaches in this area. This might sound snarky, but I still commend the author for doing what I wasn't able to by now: writing down their view of the world and putting it out for public scrutiny.