• Marct
  • Posts
  • Marct 🌱 / Living with AI, OpenAI Partnerships, ChatGPT+more data!

Marct 🌱 / Living with AI, OpenAI Partnerships, ChatGPT+more data!

Author

Alex Ker
February 23, 2023

Heya! We are switching up the format of Marct.

First, some opinions and observations I have on the AI opportunities landscape (ranging from a few sentences to few paragraphs each week).

Next, the internships/jobs at the frontier of AI you know and love.

Lastly, a brief collection of links to the top reads, projects, and research that grabbed my attention for learning. I’m dedicated to bringing the highest quality information, so I promise I’ll try to not make it overwhelming and cluttered! Only the pieces that stood out to me and I think are a must read.

Bonus: If you want to chat about anything AI, feel grab a time with me here

Alex's Take

One key role that companies are hiring these days are prompt engineers, people who can intelligently query or prompt generative AI like DALL-E, ChatGPT to come up with useful (or funny, even entertaining) responses.

For NLP specifically, you need to be good at “chain-prompting”, or providing the appropriate context to help you get what you are looking for. Anthropic is even hiring a “prompt engineer and librarian” full-time to do this: https://jobs.lever.co/Anthropic/e3cde481-d446-460f-b576-93cab67bd1ed. My prediction is teachers will be the best prompt engineers for the three reasons (see Twitter thread below). Maybe writers will translate their articulateness into prompting? What are your hot takes on prompt engineering?

On a separate note, “wrapper” ChatGPT startups seem to be thriving in the market lately—the critical question is how defensible the business is. For me, I need to see the following things in a ChatGPT-powered startup to believe there is at least a chance of surviving in the long run:

  1. Differentiated distribution or marketing channels +

  2. Proprietary data or ways to fine-tune models, which brings us to LangChain, the startup I’m most excited about this week. They help engineers build complex products by supercharging large language models with data that is not available on the internet. Using one query’s result as context for another query is powerful. Look at this demo where a Snowflake database is used jointly with Wikipedia to answer questions that depend on multiple datasources.

Jobs (hand-selected creme de la creme)

Abacus - Software Engineer 1 - Machine Learning: AI-Assisted Data Science and MLOps at Scale

AK’s thoughts: Autonomous cloud AI to handle end-to-end ML/Deep Learning at enterprise scale while remaining customizable; build specialized workflows, e.g. churn prediction, personalization, forecasting, NLP, and anomaly detection; opportunity to translate research and adopt neural networks for custom use-cases, to learn breadth of ML techniques and own custom solutions.

AK’s thoughts: Protect the world’s data with Rubrik in the InfoSec org; data governance (e.g. data mapping, implementing data retention policies, data lineages) is as important for data-driven organizations as building ML pipelines, and often looked at too late or when something goes awry; learn leading data governance strategies + implement practices at whatever company you go to next operating with data.

AK’s thoughts: Securing customer information is a top priority amongst companies working with vast amounts of data; use AI to prevent fraud and abuse; create techniques for pulling sensitive information in a privacy-preserving way; identity management will be critical as models learn from customers to derive insights—Persona is poised to fit this piece of the puzzle.

AK’s thoughts: Experienced founding team who created Uber’s Michelangelo platform and the most popular open-source feature store; best models depend on robust features and ways to continuously update them accurately and consistently, so could become a must-have for all ML products in the future; develop deep expertise in a specific (arguably the most important) part of the ML pipeline, as features contribute equally if not more than model architecture in performance.

AK’s thoughts: Democratizing computer vision and an API/infra to increase efficiency for first-time developers; role based on building training, search and development + delivering solutions on enterprise contracts + open source; excellent startup ex YC, winner of Pioneer.app competition—where founders must be fast iterators identified through the wisdom of the crowd.

Good Reads

Bain x OpenAI - OpenAI exploring enterprise partnerships that brings value to verticals, which will threaten all startups building on top of GPT3.

AI Platforms, Markets, & Open Source - Elad Gil’s latest piece on AI.

Cool Projects

Linus Lee Is Living With AI - Best read this week, on how to actually integrate AI tooling in a thoughtful way into our everyday lives, including reading research papers.

https://www.markiewagner.com/summ - Transcript Search and Summarization.

Research

If you like what you’re reading, please forward this to a friend who might enjoy this content.

Anyways, thanks for reading!

Let me know what you liked/didn’t, what you want to see, and if you do end up applying to something here—by emailing me at [email protected] :)

Cheers,

A.K.