Beyond Theory: AI’s Pragmatic Side*lwufI_EpOopuFLwc

Original Source Here

Beyond Theory: AI’s Pragmatic Side

There’s a tendency to discuss artificial intelligence with a certain degree of abstraction.

Maybe it’s the complex math powering the latest massive models, or the equally thorny debates around safety. Perhaps it’s still a challenge for many of us to bridge the mental gap between a concept that comes with an aura of futurist optimism (and an occasional overtone of dread) and a technology that’s already here, changing the way we work, create, and live.

Our latest selection of recommended reads focuses squarely on the latter. These articles are all about practical use cases, the nitty-gritty aspects of AI’s applications, and other related topics that will appeal to the tinkerers among you. They don’t require advanced expertise, only a healthy dose of curiosity. Enjoy!

  • How to unlock the power of DALL·E. We’ve all seen the stunning, extremely shareable images created by OpenAI’s DALL·E and its newest iteration, DALL·E 2. Producing these images, however, isn’t a straightforward process just yet, and requires some distinctly human finesse. Iulia Turc’s post explores the emerging subfield of prompt engineering, or prompt design: coming up with the text strings that generate the visual artifacts you’re after.
  • Making art by bringing together two powerful AIs. Alberto Romero’s latest work approaches a similar topic from a different angle. Instead of crafting prompts for an image-generation model on his own, he delegated some of the work to GPT-3, and let the model provide descriptions of famous landscape paintings that he then fed into Midjourney’s image-generation engine. The results are fascinating, and raise a whole set of questions around creativity, authorship, and reproducible artistic style.
  • Could an AI ever apply for a job on your behalf? Writing one customized cover letter after another is easily one of the least fun aspects of job hunting. As a thought experiment of sorts, Amber Teng decided to leverage GPT-3 and Python to build a prototype of a cover-letter generator; it might not land you a FAANG job interview just yet, but it nonetheless opens up possibilities and interesting conversations around hiring, identity, and current HR practices.
Photo by Hayley Clues on Unsplash
  • The legal ramifications of working with AI. Sure, with great power comes great responsibility, but risks and potential harms are not far behind, either. The European Union’s landmark Artificial Intelligence (AI) Act requires practitioners to navigate an increasingly complicated legal, political, and ethical terrain, and Ayush Patel’s overview is a helpful place to start learning about current constraints and best practices.
  • A hands-on approach to text generation. BLOOM, a new large language model released by Big Science, has made a big splash in recent weeks thanks to its accessibility, open source status, and size. Danie Theron gave it a try, and came back to report on the practical steps it took to get BLOOM going. (Danie also shares some helpful code snippets in case you’d like to try it out yourself.)

We hope you enjoyed this week’s highlights! If you’d like to support the work we do to bring you the best in data science, machine learning, and AI, consider becoming a Medium member.

Until the next Variable,

TDS Editors


Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot

%d bloggers like this: