Alternative Data Weekly #202

September 6, 2024

John Farrall

Sep 06, 2024

Thanks for being here!

Announcement(s) ... the Fall event circuit:

Rebellion Research’s NY Future of Finance & Ai Conference. September 13, 2024

Eagle Alpha’s NY Unbound Conference October 9, 2024

Battle of the Quants London, October 23, 2024

Beryl Elites events in NY October 28, 2024. Link here.

Neudata Paris, October 30, 2024

Theme that emerged in this week’s email is … making data easier to access & engage with is changing how people can add value to their organizations; stay ahead of the game.

QUOTES

“The real value of analysts will lie in their ability to navigate the intersection of AI tools and strategic decision-making. By leveraging both human expertise and AI, analysts will generate insights of unprecedented depth and accuracy.” - Ben Lorica

News Articles
Podcasts
Cool Charts
Final Thoughts (the grind)

#1 – New publication called Datacisions, from Sven Balnojan, published Every Product Will Become A Data Product. September 2024.

My Take: Along the lines of every company being an internet company…every company is a data company. You just collect data, crunch data, then make a decision. It is really that straightforward.
I look forward to seeing more from Datacisions. I like the short, single idea, format.

#2 – Ben Lorica published Rethinking Analyst Roles in the Age of Generative AI. September 2024.

My Take: The role of the analyst changing. The tools are getting a lot better to do some of the blocking & tackling jobs. The analyst role is going to see huge benefit from AI augmentation. When I was in that role earlier in my career, I drove value for my senior analyst because, in part, I knew what data was available and how to get it, and then put it in a format that was easily ingested (Excel or PowerPoint). That part of the role is going to get much easier (AI will “augment” that part of the analyst’s role). In theory, this will allow the analyst to do the more interesting work…like analysis and answering the hard questions.

#3 – Mark Trowbridge published Don’t Fear the Data Trial, Let Technology Help. August 2024.

My Take: Asking a hedge fund to trial your data is like going down a black hole. You’ll never get any feedback, and they’ll also never give you an outright “no”. Anything to take friction out of that process is a big positive for the data monetization industry. Maiden Century’s “accelerated time to insight” is a big deal.

What else I am reading:

Peter Baumann’s Book review "Humanizing Data Strategy". September 2024.

Seattle Data Guy’s Don’t Lead a Data Team Before Reading This - The First Chapter. August 2024.

Rebecca Natale’s Facing platform shutdown, former IEX Cloud head buys its assets in 11th-hour bid. August 2024.

Aswath Damodaran published Beat your Bot: Building your moat against AI. August 2024.

ModuleQ’s The Future is Now: AI in Banking. August 2024.

Wolf Street’s Why We’ll Replace the Case-Shiller Home Price Index for Our Series. September 2024.

Source: Practical AI’s RAG continues to rise with Demetrios Brinkmann from the MLOps Community. April 2024.

My Take: I found this podcast as I was trying to learn about RAG. I enjoyed this one and plan to keep it on my podcast rotation. One consistent theme I hear when drilling into AI is the importance of starting with business use case and working from there. Many AI products are “hammers in search of nails” rather than new products trying to solve real problems.
There was a good exchange late in the podcast about persona-morphia … or the ability to deeply understand a role within a company and deliver to that person everything they need (this is what we do at ModuleQ).

Highlights (29-minute run time):

Minute 01:30 – Interview starts; RAG vs Fine Tuning
Minute 03:00 – assistants & automators - "RAG as a service"
Minute 04:45 – text to SQL thoughts
Minute 07:30 – survey results: main takeaway there is budget being allocated to AI; trying to figure out what are the biggest leverage use cases? How to explain that to mgmt users w
Minute 11:00 – how are people using RAG?
Minute 15:45 – discussion of Andersen Horowitz survey; notably enterprise multi model future
Minute 18:45 – AI mature companies recognize which models they should be using
Minute 20:45 – prompt templates song interlude
Minute 23:00 – though to do at scale; 42% using data that they've created to evaluate of model is working correctly
Minute 25:00 – 72% of ground truth labels were manually labeled by humans (hosts were surprised by this); fragmented communities of AI practitioners makes for lack of best practices
Minute 29:30 – “How can I get this being used in my company, so that I can either save money, or make money?”
Minute 33:00 – role based access control (RBAC)...we get asked about this a lot.
Minute 42:00 – funny exchange about neuromorphic computing; basically can we mimic the human brain.

Source: DataBricks published The State of Data + AI. August 2024.

My Take: We are starting to see real life examples of AI being used, not just at small startups, but large companies as well. It feels like we are just getting going on a process that will be filled with fits & starts, but progress will be made!

“This year, we’ve seen a shift from experimentation to production applications of AI.”

Of interest is page 27, Highly Regulated Industries are Early Adopters. We’ve seen this before, and while perhaps counter-intuitive, it makes sense because there are rules and guardrails already in place.