
Why Accuracy Isn't Enough: A New Quality Model for Real-World ML Components
AI components, whether traditional machine learning or generative AI, are embedded in many present-day systems. However, despite advances in automation, infrastructure, and tooling for developing AI components, many do not leave the prototype stage or reach production because they fail to meet overall system quality expectations. In our latest podcast from the Carnegie Mellon University Software Engineering Institute, Rachel Brower-Sinning and Robert Edman, both machine learning scientists in the SEI's Tactical Edge and AI-Enabled Systems Initiative, sit down with Grace Lewis, the initiative's lead and an SEI principal researcher, to discuss a quality model for machine learning components to support their prop...
Opens the player: the full transcript, every word clickable, with translation and flashcards.
Is this episode right for you?
B2 · Upper-intermediate
Comfortable if you already follow TV news
- 71% of the words are B1 or easier.
- 193 words a minute: natural, fairly fast speech.
- 130 harder words (B2 and up) in the whole episode.
Try it: tap any word
Real sentences from the start of this episode. Every word is marked with its level.
is this concept of a system.
, .
So why shouldn't we treat it like software, right?
Tap a word to see its meaning in your language.
Do you know these words?
8 harder words from this episode. Tap one to see what it means, then mark it.
Mark a few words and we'll tell you how this episode fits your level.
Is this episode right for you?
B2 · Upper-intermediate
Comfortable if you already follow TV news
- 71% of the words are B1 or easier.
- 193 words a minute: natural, fairly fast speech.
- 130 harder words (B2 and up) in the whole episode.
Transcript
The opening of the episode. The whole transcript is in the player.
harder words (B2+)
Going back to what I think differentiates our work from other work that I see out there, is this concept of a system. So if you think about what a machine learning component is, it is software. So why shouldn't we treat it like software, right?
And so when you say the requirements that come from the system, I think that is very important because when you do system decomposition, it so happens that at some point you're going to say, okay, this should be a machine learning component.
And so the requirements for that component are being derived from the system. And that is where it's important to be able to say, say, okay, what can I test at the ML component level? Today, AI components, whether traditional machine learning
That was the opening. The rest of the episode (23 min) is in the player: every word clickable, with translation, saved words and flashcards.
Continue in the playerAutomatically generated from the audio. May contain errors. Audio belongs to its publisher and is played from their feed. Rights holders can request removal — copyright & takedown policy