Thanks for being here!
Welcome Alt Data Weekly’s Inaugural Sponsor:
Augment Your Design & Development Team
Click here to learn more. See special offer below for readers of Alt Data Weekly.
Announcement(s):
I will be in NY July 25-26-27. Let me know if you have time to connect.
Theme that emerged in this week’s email is … LLMs are dependent on good data. Providers of good data are positioned to benefit.
QUOTES
"It was all the rage that Alternative data was going to transform the industry. And, it's had a modest impact, which was, frankly, our prior back then. And the reason for [this] is most of your alternative datasets just do not have a richness of information that transcends [the information] that you have in management team interactions." - Ken Griffin CEO and founder of Citadel (h/t Jason Derise).
“In today’s world, data is the core corporate asset, and it’s up to individual organizations to monetize and commercialize it through the new wave of tools companies like Databricks and Snowflake are developing.”- Battery Ventures
News Articles
Podcasts
Cool Charts
A Word From Our First Ever Sponsor: Neutech
Final Thoughts (Time Management)
#1 – Amenity’s Doug Hopkins published The Evolution of Alternative Data in Finance is Being Driven by LLMs. June 2023.
My Take: Doug shares with us the four (4) phases of Alternative Data.
Phase A – emergence of startups offering new types of data
Phase B – alt data used in systematic investment strategies
Phase C – emergence of proprietary data lakes
Phase Z – newest phase is LLMs (nonlinear phase)
I have spent time trying to get my arms around the investing use case for LLMs and remain of the belief primary data owners will see the most value creation. Open to feedback.
#2 – Harbr’s Data Marketplace Guide. July 2023.
My Take: “data is sold not bought” (quote attributed to a several people, but I most recently heard from Dan Entrup). Harbr offers us a great summary of the benefits of data marketplaces. I have been skeptical of data marketplaces. 90 West has listed products on essentially all the marketplaces with limited success. Data is tough to simply browse & buy. It feels like we have a long way to go. That said, data is becoming more “productized”, tools are improving, buyers are getting more sophisticated … all of which helps data discovery & ultimately answers real business questions. Let’s watch!
#3 – Misiek Piskorski & Amit Joshi published in HBR What Roles Could Generative AI Play on Your Team?. June 2023.
My Take: Both McKinsey & HBR are weighing in on AI & LLMs ... its’s getting real! In all seriousness, these are great frameworks to use as we start to define the issue and make better predictions. The authors offer six (6) different types of GenAI use cases.
ChatGPT – human interacts with machine (I use ChatGPT daily)
CoachGPT – helps manage your daily life
GroupGPT – similar to above, but for a group of people
BossGPT – advises groups (#3 above) based on inputs
AutoGPT – human gives machines a task and the machine can collect resources to get it done
ImperialGPT – machines interact with each other to identify ideas that will aid the human decision maker
#4 – Battery Ventures Databricks and Snowflake Face Off as AI Wave Approaches. July 2023.
My Take: “Data has gravity”. I attended the SnowFlake Summit in Vegas two weeks ago and the theme is that data is now so big we need to bring compute to the data rather than data to the compute. Thus, the gravity analogy. SnowFlake & DataBricks are battling for supremacy and both aim to be platforms on which data companies can thrive.
BONUS: Jeremy Khan published in Fortune Deep learning pioneer Andrew Ng says companies should get ‘data-centric’ to achieve A.I. success. June 2022. “With the right data, he says companies with just a few dozen examples or few hundred examples can have A.I. systems that work as well as those built by consumer internet giants that have billions of examples.”
BONUS 2: Silicon Angle’s Connecting the dots on Snowflake’s Data Cloud ambitions, June 2023. “There are two core themes: 1) Data from people, places, things and activities in the real world drive applications, not people typing into a user interface; and 2) Informing and automating decisions means all data must be accessible.”
What else I am reading:
The Verge published Elon Musk blames data scraping by AI startups for his new paywalls on reading tweets. July 2023.
Brad Schneider’s Unlocking Value from Data: How to Revolutionize Your Data Strategy with DRM Software. May 2023.
Adam Braff’s Code Interpreter is not a data scientist. July 2023.
Ethan Mollick’s What AI can do with a toolbox... Getting started with Code Interpreter. July 2023.
Source: Ben Lorica of the Data Exchange Podcast interviews Andrew Feldman, CEO & Founder of Cerebras.
My Take: This interview caught my eye as LLM & Chat-GPT is fascinating but it is still unclear to me how this will be commercialized and incorporated into our day-to-day lives. One risk is (in theory) all data entered into these AI models is out in the public domain. This is a big risk/fear. Thus Cerebras is helping build custom foundation models. The idea that models can be trained on internal data, walled off from the world, is a big deal. The other key takeaway from me is that data quality > data quantity. Owners of good, proprietary data will benefit.
Highlights:
Minute 01:00 – Cerebras GPT; They build AI accelerators
Minute 03:00 – models in open source domain; “recipe” & everything using open sourced data
Minute 05:00 – Reddit group engagement
Minute 05:30 – reaction from co’s; need to train models on internal corp data; ways to modify GPTs for internal data
Minute 07:30 – what are operational challenges when using models
Minute 10:45 – large enterprise opportunity to train internal models; can they train from scratch? Discussion of model size and associated cost.
Minute 15:30 – proprietary models better. … how will this be different when domain specific models
Minute 18:30 – discussion of transformers and how that will evolve; issues of speed and performance
Minute 22:00 – background in Cerebras; impressive stats
Minute 27:00 – focus moving to smaller models and better data
Minute 32:15 – data new gold. “Moat is the data”
Source: McKinsey’s The economic potential of generative AI: The next productivity frontier. June 2023.
Many of these visuals are interactive. Visit the report via the above link for the full experience.
I thought this understates the revenue opportunity from generative AI. Use cases like prospect identification and vetting might make sales much more efficient.
High tech software engineering poised to feel the biggest impact.
Our valued sponsor Neutech can help you manage your team’s growth.
Now through August 1, email sales@neutech.co with subject “Alt Data Weekly” to get $2,000 off your first month.
Get things done.
Be productive. This means you are producing something. A construction worker can look back at the end of the week and point to a structure they built. Ask yourself what you created this past week. A new customer? A new report? A new personal best?
I find I get far more accomplished when I am busy. This does not mean I put meaningless tasks on my calendar just to look busy. This means that when I focus on being productive I am more productive.
I know I am most effective from 8am-11am after I've worked out and had a cup of coffee. This is the time of day that I do my best work.
I know I am least productive from 1-3pm. My brain works differently, perhaps its after eating lunch, but my slowdown is consistently at that time of day. This is when I will try to get some of the mind numbing stuff done...or just take a break and think.
I know I get a second wind late afternoon/early evening during which time I can be really productive.
I am not good at late nights. Never have been.
Some thoughts.
Have goals
Is what you are doing right now moving you towards your goal? Especially when you are in the middle of your highly productive time. Know your mission and stay focused.
Focus on high leverage activities
"high leverage activity is something with a small injection of your time, energy, or resources can massively impact lots of people." - Source Keith Rabois
This has been a big one for me. Without specific goals & identifiable high leverage activities, you will get sidetracked and waste time.
Say no
This one is hard for me. I like to say yes. Identify those things that are not high impact. Avoid, delegate, or outsource those activities.
Use the below “Urgent vs Important” 2x2 to bucket activities.
I will grant you that, particularly early in your career, there are activities that might not be highly productive but you need to be in the room. "Face time" in the office is a real thing. You can learn a lot just from being around those with more experience (watching, listening, learning). You’ll say “Yes” to more things as you are in the process of exploring options and developing your mission. As you mission narrows, your ability/need to say “No” will increase.
Time to think & recharge
Rest is important. Sleep is important. You need to set aside time to think & be creative. This does not mean doom-scrolling or 4-hours of Netflix every night. What does this look like? Put your phone down. Get outside. Take the AirPods out of your ears. Let your body & mind recharge.
Self-care
Make time to work out. Move your body. Sweat. Find what works for you and make this a priority. For me what works: early morning, with a group, high intensity. This works for me.
I include socializing in the self-care bucket. How and when you socialize changes as you move from teen years through your 20’s and beyond. Building & maintaining friendships is a big part of self-care. Invest time into your relationships.
Use Tools
Use the tools are your disposal. Your phone is an amazing tool that you should manage carefully. Calendly has been one of the best tools I've used. Spend time being organized. Manage your calendar or it will manage you.
Find what works for you. Reflect at end of day/week to learn what worked & what did not work. Minimize that which does not work.