Numerai and Polychain
USV recently made two investments in hedge fund managers. You might gather from that fact that we are moving away from venture capital. But that would be an incorrect assumption. What we are doing is finding new ways to invest in technologies that we think are interesting. These would be technologies like artificial intelligence, encryption, and blockchain.
Last week it was leaked, and then announced, that USV had led a round of financing for Polychain Capital, a new kind of fund that invests in digital tokens. We have discussed these digital tokens here at AVC and elsewhere. We do plan to directly invest in blockchain based protocols via tokens, but we won’t be able to invest in all of them in a venture capital model, so we are also betting that a fund vehicle focusing on these assets will make a good investment.
Yesterday, my partner Andy wrote a blog post on the USV blog about Numerai, a new kind of hedge fund that allows machine learning experts all over the world to crowdsource public market investments by building models and sharing the selections that come out of their models. The data they use to power these models is encrypted and so are their models. It is the “blind leading the blind” according to this piece in Wired. I like that description as the focus moves away from things like fundamental and technical stock analysis to true pattern recognition done by machines. Numerai uses cutting edge encryption technology (a variation on homomorphic encryption), machine learning, and pays out using the bitcoin blockchain. It’s a trifecta of things that are interesting to us, as recognized in this tweet by MIT’s Technology Review.
In both cases, USV invested in the “GP” of these funds. In Polychain, we also made a small investment in the fund itself. We did not do that in Numerai, although it is possible that we could choose to do that in the future. Our investments are largely in the companies that manage these funds and they are a bet that these new technologies will offer new ways to build fund management businesses that over time will be highly valuable. These are both startups and we think of these investments as classic early stage investments with all of the risk and return that comes along with them.
Comments (Archived):
Very interesting. How about investing in a startup that uses AI to invest in startups? Then you and partners can retire.
There are some people working on that. Our portfolio company CircleUp is one of them
I spoke with Numerai’s CTO in March in SF after he presented at an AI meetup and shared that their own model is ranked low on their league tables.Numerai is a variant on http://www.hedgehogs.net, whose founders I know from 2009, with more data science tools. There are merits to both models.I said to Numerai’s CTO that financial data is only a fraction of what’s needed in a comprehensive prediction model.
Where’s the fun in that?
Retirement is for people who don’t like what they do. 🙂
AI doesn’t have the tools to assess team chemistry or product-market fit in the same way as human investors (yet).They will, though.
Polygolf, never ending Polygolf.
Fore!
So, to confirm, with Polychain (perfect name)…—-> USV is betting on a startup———> Which is betting on securing investors—————> Who are betting on buying tokens——————-> That are betting on creating protocols————————–> Which are betting on powering the future of technology.F**king love it.
We do too. For the record Polychain is not the only find doing this. We know of at least two others and like the founders of those very much too
Where is JLM these days #JLMMissing?
I believe he took a post in November as an invitation to not participate in AVC…..and he took it seriously, which would be his personal MO.
Cool, but my head hurts.
Fred – your Numerai investment was also picked up by the Financial Times in an article yesterday – https://www.ft.com/content/…
That’s great to hear
Skate to where the puck is going.
O Canada, nice hockey reference. The family is heading to Quebec City later this winter btw. I know it’s not your home base, but do you have a top tip to share, M. Mougayar?
Spent New Year’s in Quebec City a few years back. It was 25 below zero, but still a lot of fun. The city has a lot of charm and a strong Euro feel. The cold didn’t stop the locals from playing hours of pond hockey, though. Darwin at his best.
Salt Shaker:any destination that hasn’t reached the minimum weather of 75 degrees Fahrenheit or 23.8 Celsius is not a destination we couldever visit. We love Canada and the Canadian people but dislikethat awful weather.
Quebec City is so charming. The hotel Frontenac and views from it are amazing. Aside from the known places to visit, the food is excellent, but try to avoid the tourist spots where it will be a notch down. Have fun, and hopefully the temperatures won’t be so bad.
Here is another team in an equally large but relatively untapped investment market: http://www.myfanfund.com
Tom:www.myfanfund.com (underserved market of Florida and invests in high growth companies in the technology and life science sectors)The Florida reference was all we needed before even researching the firm.The Contributors on this blog are just a little more sophisticated in their investing approach. That would be a negative. COPY?All we can ask is REALLY?
I have no idea what you are talking about. They are crowdsourced AI sports investing.
Tom:It appears we were not more direct. That is not a financial vehicle that fits our investment principals nor style.No further response required.Thanks
Not sure I’d advise anyone on taking an investment from you based on this thread. I’ll be more direct. We thought the company was cool and out of solidarity wish them the best in this market.
The new Long Term Capital?
Soon the ecosystem of cryptocurrency will get deeper. One of the initial hurdles was “Where do I hedge my risk?”. Starting to find places to do that. Next place that needs development is reward for holding otherwise known as interest bearing notes.
I wonder how the use of ML predictive models will affect the fundamentals of investing and what that means for the companies (or assets) being traded. Instead of thinking about where a company will be in 6 months, I should be thinking “what will the ML signals be in 6 months”.
Are you implying that even ML algorithms can collectively fall into their own self fulfilling rabbit (w)holes ?
A lot, because you can create Complex strategies with machine learning.Hence why de Shaw has been hiring phds for years
I will try to write more, but I am so proud and impressed of this decision. As someone working to launch a similar GP type fund business, it has been very frustrating to hear investors explain why it is not appropriate as an early stage investment. I’ve just emailed several people on my team to say — YET AGAIN, USV leads the way. BRAVO!
I saw reporting that you had only invested in the Polychain fund, which didn’t make sense to me. Investing in the firm makes a bit more sense.It will be fascinating to see what kinds of returns a VC-type firm generates when it achieves liquidity in its investments almost immediately.
So USV is a Poly LP?
according to this post, both GP and LP
A leak is not a good start, and I can foresee a polynumerai of rumours being spread by wannabe blockchain startups that their tokens have been snapped up OTC.Is Poly going to be announcing its investments?
How does it feel to be disrupted by blocktech?
I’ve had an insider view of several of the industries most prominent ICO launches. In each case I’ve witnessed insiders guaranteeing their allotment of tokens before the ICO, participating in the hype cycle, and selling tokens during the initial pump. I’ve seen tokens used to pay contractors, incentivize journalists, and solicit partnership announcements. The lure of easy money perverts incentives. I’m hopeful that USV’s involvement in this sector signals an end to this type of behavior.Several industry leaders are working on new legal and regulatory frameworks for tokens. There’s an opportunity to demonstrate a new level of transparency. Technology allows you to publish a cryptographically verifiable table of all insiders that participate in the ICO. You can also lock tokens in a manner resembling a founder’s vesting schedule.Good luck with the new ventures! I have friends who have invested in both companies and will be following them closely.
Wayne Vaughan:”during the initial pump”We guess if dumb isn’t added it is viewed positively at how a sector or venture is pumped.
With regard to Numerai, when did a quant become a machine learning expert? Most trading algorithms have little to do with machine learning?
After my maths degree, I worked in a hedge fund for a Professor of Neural Networks (Deep Learning) where we build 5+ machine learning models to do portfolio asset allocation and risk management.That was a good 10 years before the current spate of DL in hedge funds and in consumer tech.
How were the results? Is it still going on?
Results were good. The hedge fund’s data science tools are now part of IHS Markit. The Professor of NN had gone on to build ML models for Nat Rothschild.https://uploads.disquscdn.c…
Twain Twain:Is the Nat Rothschild you cited him?Nathaniel Philip Victor James “Nat” Rothschild is a British-born financier who has settled in Switzerland, and a member of the prominent Rothschild family.If so why?We always wondered how Thelonious Monk was able to keep the company of some of the Rothschilds.
A long time ago
Will Poly be voting for delegates in proof of stake blockchains?
What’s the criteria for success/failure? Are there benchmarks relative to more traditional investing?
Both very cool ideas. I would love to see Numerai go full DAO someday.
It’s worth an allocation to Polychain just to see what they’re up to. I don’t think chasing the latest trendy altcoin is such a great model for protocol investing, but allocating to such a fund *is* a great way to find out what the next trendy altcoin is gonna be.There might be a better and more holistic way to invest in protocols, which would have to involve an attack at the meta protocol level. The problem with investing in a specific altcoin is that if it presents a novel feature, either the feature will be copied, or the entire protocol will be forked (and probably dozens of times). How do you capture any meaningful portion of innovation in bitcoin-like protocols when they can be forked infitely?Another viable approach is to be an “activist” investor. Buy up tokens in abandoned or unpopular (but still functioning) protocols. Then fund basic development efforts to update wallets, negotiate exchange listings, and advertise. There are dozens of protocols with market cap under $100k that could easily be $10mm protocols. This is probably too smalltime to be of interest to USV, but I’m sure someone is doing this.
Sounds good to our ears since we are building a tokenised and community-led “fork” of the vc industry.
The hedge fund is itself a startup that will invest in blockchain startups, so its not like USV does some FoF type of deal.
USV posted before that their latest thesis is based on AI, blockchain, wisdom of the crowds and networks effects in domain-specific verticals, so Numerai hits all four bases.See Kristian Hammond of Narrative Science’s tweet about data science automation which is what Numerai does for the hedge fund industry.https://uploads.disquscdn.c…@fredwilson:disqus
It just might be an indication that USV’s success is forcing them to find scale that a pure thesis driven investment model cannot sustain 🙂 Capital needs to find good outlets and when idle $$’s start burning a hole in the pocket .. it is good to slaughter sacred cows I suppose (then again they might not be viewing picking specific companies as their sacred cow at all)
Haha, no, Charlie. You’re so so HUMAN.Your artisanal offering is about understanding the emotions and experiences of your customers. It’s not likely to get automated within the near future, :*).AI systems like Numerai’s ignore those signals and focus purely on the quant financial data and optimize for that in its prediction models.
It’s OK. At least there were no actual sharks involved
It’s funny @fredwilson:disqus references the “blind leading the blind” quote.Numerai’s system is an example of the Enlightenment method which, at the most, can only provide us with 1/2 the whole picture — the 1/2 which is the logic box (aka the Matrix, the “black box” of binaries and probabilistic derivatives that people refer to in HFT, Black-Scholes and Deep Learning models alike).The Enlightenment methods aren’t so “enlightened” because they’re blind to not-yet quantified / quantifiable data and the language sphere beyond the logic box. The sphere contains all the “black swans” of language understanding and context of art+heart in investment decisions.https://uploads.disquscdn.c…Of course, Da Vinci was more intelligent and could see more coherently than Descartes!
given up on local as an economy then?the rub is not that local doesn’t work. it is that once you move products out of a local model the margins for transportation and management make all human scale non economical.
I can imagine a group of robots gathering for Thanksgiving in the future, holding their polymer hands together, a robo-turkey at the center of the table… thanks for our jobs and for the stupid humans that created us. 🙂