Gold, Silver, Platinum, Palladium, Rhodium: Each Precious Metal Is Precious in Its Own Unique Way
Olegs Jemeljanovs, PhD, CFA·7 min
Alternative Investments
We Need A Lot More Than Data: How Startups Can Harness AI
Amit Garg
July 19, 2020·4 min read
You open your phone, take an eye exam, and find out your risk of Alzheimer’s 3-5 years ahead even though you have no symptoms.
A doctor takes data (CT scans, genomic tests, blood labs, demographics etc) as inputs, the software is trained specifically on the patient’s biomarkers, and gives a recommendation what drugs would work best.
Veterinarians find out what drugs approved for humans can be repurposed to crush a dog’s specific cancer. Eventually the data can be used for humans since cancer is a shared malady.
These are not pipe dreams but examples of real companies -- Neurotrack, Onc.ai, and FidoCure (full disclosure: invested through Tau Ventures) respectively. Personalized medicine, which has been science fiction for generations, is less fiction and more science these days. In many ways AI is the army and startups the generals of this revolution. This post is not about the promises and pitfalls of AI but really where the world is today, highlighting invariably the opportunities.
Moore’s Law established a trend around computational power that has held true for almost 50 years, we need a framework for how to think about the data explosion. At Tau Ventures we are skeptical around data as a service but in some domains, especially regulated ones, we see successful business models.
At a high level AI today is a powerful tool that can help humans make decisions especially if we constrain the problem space i.e., be very specific about what we are asking. One major type of AI is supervised learning which infers relationships based on inputs that include responses -- the better data you have the better models you have. Another major type of AI is unsupervised learning which infers relationships based on inputs without responses -- basically finding hidden patterns in data. But in both cases we may have an oversampling of certain types of data and missing not just edge cases but significant use cases. Think about AI classifiers that couldn’t recognize Obama’s face, which raises huge issues beyond the scope of this short article. At Tau Ventures we see the need for many startups to help us fill in the blanks, whether it is tools to adjust for biases, collect hidden data, or actually create / simulate data.
Originally published on “Data Driven Investor," am happy to syndicate on other platforms. I am the Managing Partner and Cofounder of Tau Ventures with 20 years in Silicon Valley across corporates, own startup, and VC funds. These are purposely short articles focused on practical insights (I call it gl;dr -- good length; did read). Many of my writings are at https://www.linkedin.com/in/amgarg/detail/recent-activity/posts and I would be stoked if they get people interested enough in a topic to explore in further depth. If this article had useful insights for you comment away and/or give a like on the article and on the Tau Ventures’ LinkedIn page, with due thanks for supporting our work. All opinions expressed here are my own.
1) We Have Data. A Lot Of It.
In Jan 2020 Waymo (Google aka Alpahbet’s self driving division) announced it had logged 20 million miles of driving. Consider that a year before they had 10 million and a year before that closer to 4 million -- the growth is exponential. Perhaps no other area exemplifies a pre-cambrian explosion of data more than medicine. A decade ago I remember hearing that half of what you learn in a medical school would be outdated by the time you graduated four years later. It’s obviously not true because core concepts will remain unchanged. But it is also still true given the amount of medical knowledge overall is not just increasing but actually accelerating.
Moore’s Law established a trend around computational power that has held true for almost 50 years, we need a framework for how to think about the data explosion. At Tau Ventures we are skeptical around data as a service but in some domains, especially regulated ones, we see successful business models.
2) But The Data Needs Massaging
Unstructured data is an unpolished diamond, unlabeled data can be masking fool’s gold. At Tau Ventures we believe in multiple winners -- the opportunity is growing so much it will invariably lead to many startups succeeding. Trying to make sense of climate data so you can plant crops and maximize yield? Trying to make sense of the movement of people within a store to optimize placement of items? Trying to make sense of the conditions in a factory to ensure less errors and higher quality? Once again these are all real startup cases and our view as investors is that the key to succeeding is choosing the right beachhead, getting to recurring contracts, and building vertical-focused solutions since we are still far from general purpose AI.
3) And We Will Need to Adjust The Data Too
At a high level AI today is a powerful tool that can help humans make decisions especially if we constrain the problem space i.e., be very specific about what we are asking. One major type of AI is supervised learning which infers relationships based on inputs that include responses -- the better data you have the better models you have. Another major type of AI is unsupervised learning which infers relationships based on inputs without responses -- basically finding hidden patterns in data. But in both cases we may have an oversampling of certain types of data and missing not just edge cases but significant use cases. Think about AI classifiers that couldn’t recognize Obama’s face, which raises huge issues beyond the scope of this short article. At Tau Ventures we see the need for many startups to help us fill in the blanks, whether it is tools to adjust for biases, collect hidden data, or actually create / simulate data.
Originally published on “Data Driven Investor," am happy to syndicate on other platforms. I am the Managing Partner and Cofounder of Tau Ventures with 20 years in Silicon Valley across corporates, own startup, and VC funds. These are purposely short articles focused on practical insights (I call it gl;dr -- good length; did read). Many of my writings are at https://www.linkedin.com/in/amgarg/detail/recent-activity/posts and I would be stoked if they get people interested enough in a topic to explore in further depth. If this article had useful insights for you comment away and/or give a like on the article and on the Tau Ventures’ LinkedIn page, with due thanks for supporting our work. All opinions expressed here are my own.
Amit Garg
I have been in Silicon Valley for 20 years -- at Samsung NEXT Ventures, running my own startup (as of May 2019 a series D that has raised $120M and valued at $450M), at Norwest Ventures, and doing product and analytics at Google. My academic training is BS in computer science and MS in biomedical informatics, both from Stanford, and MBA from Harvard. I speak natively 3 languages, live carbon-neutral, am a 70.3 Ironman finisher, and have built a hospital in rural India serving 100,000 people.