Live Chat with Professor Alex Pentland

Earlier this week, our Telegram group (https://t.me/EndorcoinGroup) welcomed Endor Co-Founder, Prof. Alex “Sandy” Pentland for a lively and insightful AMA (Ask Me Anything) session.

This was a great opportunity for members of our rapidly-growing Endor.coin community to ask prof. Pentland questions about our science and technology directly. Every question during this live session was submitted by members of our community.

Read on for a transcript of the session:

Yaniv Altshuler @ Endor.coin

Hello Endor endeavors!

Thank you for joining us to this exciting happy hour with Endor.coin co-founder, MIT Prof. Alex “Sandy” Pentland!

Background:

Prof. Pentland is a member of the US National Academy of Engineering, was recognized as “one of the world’s 7 most powerful data scientists” (Forbes). Director of the MIT Media Lab Entrepreneurship Program. Founding member of the Advisory Boards at Google, AT&T, Nissan, and the UN Secretary General. Serial Entrepreneur and a Leader at the World Economic Forum.

I had the privilege of spending 3 exciting years of research at MIT working with Prof. Pentland, which led to the development of the new field of Social Physics, and to the founding of Endor.

I would like to thank Prof. Pentland for attending this session!

Question: What is Social Physics and how are you using it at Endor / Endor.coin?

Answer — Professor Alex Pentland, Endor: Social Physics is the use of statistics to understand human behavior, but specifically the discovery and use of statistical “laws” of about how trends and fads begin and end. Endor uses these to discover when the “next thing” is coming, or when the current fad is running out of steam

Q: What problem are you tackling and how will Endor.coin solve these problems?

A: The problems to date have been about detecting scams, figuring out why customers of big companies are changing behavior and such. The coin offering is to bring this technology to everyone, so that startups and even individuals can detect fads and trends to help their business and investment

Q: And how does Social Physics differ from other technologies used in predictive analytics? How is it translated into value?

A: The big differences are that other technologies can’t run on encrypted data, which Endor can, and that other technologies requires really huge long-term amounts of data and as a consequence aren’t good for detecting trends or fads

Q: Why did you choose blockchain technology as the next phase for Endor?

A: Everything is becoming blockchain and coin. At Davos this year (world economic forum) there were blockchain talks and demos everywhere you looked. In a few years a lot of the world will be encoded on blockchains, and if you want to understand what is happening you have to be able to analyze the encrypted transactions logged on blockchains

Q: Interesting, can you please provide a few scenarios that illustrate blockchain’s potential for predictive analytics?

A: Well the obvious one is crypto trading, but the one that might have the most impact is management of logistics chains. Everything we eat, buy, and sell is part of a long chain of cooperating businesses and individuals, and today we are mostly blind to where things come from and where they end up. WIth blockchains the “great material continuum” will become more trusted, efficient, and (by using Endor’s tech) much more predictable.

Q: What is the purpose of the crowdsale? Why a token sale and not VC route?

A: Endor already got started using VC money, and serves big guys like Walmart and Coke, but the real potential is machine learning (and particularly trend prediction) for everyone, not just the big ones. The crowdsale is a move to democratize and open up super advanced AI/machine learning tech for everyone.

Q: How is Blockchain, in your opinion, best harnessed to serve the financial community?

A: My article this month in Scientific American (online at Endor.com I think) shows how we can now bring groups of little guys (small businesses, small countries, farmers, etc) together to pool their assets and actually outcompete the big guys. Imagine if small farmers could go head-to-head with Monsanto, or a group of small nations could compete head-to-head with the US dollar!

Q: How do you define predictive behavior?

A: Predictive behavior is detecting trends or fads before they really get started, or predicting when they will run out of steam

Q: What non-standard data has Endor been able to utilize in its predictions, compared to traditional predictive models?

A: The big difference is that Endor can use encrypted data, and does not need to carefully prepare and clean the data beforehand. You just dump your encrypted data in, show some examples, and it tells you what other items are part of the same group. So you can use all sorts of data, anything that people leave behind as they go through life, to predict things. Funniest recent example is using amount of trash companies produce to detect when they are about to launch a new product or have a new business relation.

Q: That’s an interesting example. Endor is using any data pre-processing (such as parsing on text) before feeding it into the Endor system?

A: Generally, Endor does not need to do pre-processing, but it can clean things up and improve performance.

Q: How Endor ensures accuracy of the data being fed into the ecosystem?

A: Endor can’t assure that the data you feed in makes sense, but it can tell you that it doesn’t work very well. This lets you figure out if your data is junk or not. Usually “quality” of data also means how many mistakes or missing data fields there are; Endor is very insensitive to data quality in this sense.

Q: Which industries and sectors do you see will adopt Endor first and alpha/beta testing it with much enthusiasm? Which industries do you see Endor will have the most disruptive potential once it becomes mainstream? What are you most excited about the project personally?

A: My guess is that the first will be cryptocurrency trading (of course!) and then as large firms begin using permissioned blockchains you will see analysis of bank performance, logistics firms, etc., etc.

What am I excited about personally? The ability to combine encrypted data sources and look for patterns in just about everything using Endor technology.

Q: Can Endor be remodelled to utilise data sets regarding non-human behaviour, for example, crunching data regarding the weather, and say, applying Social Physics to social platform data, to help a grocery chain predict inventory stock and help marketers design ad hoc online campaigns?

A: Social platform data, grocery, etc are human data and Endor is great there. Weather and jet engines not so much….but we haven’t tried much and maybe we would be surprised at how well it works there too!

Q: Where could someone without a background in data science start to learn about predictive analytics? What sort of educational trajectory or resources do you recommend studying for a career in social physics?

A: Well, you could start with my book “social physics” (Penguin press) and that would give you the general idea. As for traditional classes, statistics and network science classes are good, but also reading intro to social psych might be good (although a lot of that is not high quality).

Q: Are there plans in the long term pipeline to deal with non-human data? Is Endor ‘locked in’ to Social Physics, or other models and frameworks could potentially be applied at a later stage?

A: In the long term we will probably add capabilities for data like weather or jet engines, but since most data in the world (and most value in the world) comes from people that is first. Weather, for instance, is only interesting because of what it does to people…so detecting how people react to weather is more interesting than the weather itself.

Q: What would be the limiting factors to adoption of Endor in the current predictive analysis environment?

A: Limiting factors: you have to have data that has information about the phenomena you are interested in, and you have to have examples of the sorts of things you are looking for. This is how Endor avoids worrying about semantic interpretation: you provide the examples and that captures the semantics.

Q: You said one of your goals is to bring Endor to small businesses. Does this mean, a normal local businessman dealing with hundreds/thousands of transactions on a regular day can use ENDOR and find out what exactly is being bought by the customers and work around with the small dataset that he has? Also let’s say a farmer wants to go ahead and grow some Corp and he has the data of different varieties of crops grown in the area for last 10/20 years. Can ENDOR predict what crop yields better profits for next year or current year ?

A: Endor to small biz: yes, the local store can use Endor to predict what to stock up on, or the individual farmer can use Endor to predict what crop will give the best price. That is the quite disruptive capability the Endor brings….custom AI for the little guys, not just the big guys.

Q: On a first sight, looks like that the quality and qualitative interpretation of the data, makes the real difference in what Endor wants to reach, but they’re definitely not under your control. Don’t you think that it could be a very threatening issue for the success of the project?

A: Quality and qualitative interpretation: not particularly relevant. Endor is extremely tolerant of noise in data (“quality”) and by picking examples that capture the “qualitative” character you want Endor avoids needing to specify semantics

Q: What kind of data yields better results ? Encrypted/Blockchain/Normal ?

A: Data type and result: Probably normal, but not much better. Endor is looking for patterns and similarities, not specific values.

Q: Prof. Alex, I suppose that there must be some proprietary algo involved. Is there any comparison in the accuracy when applied to various data from varying industries. How will the algo and its accuracy differ from neural network which is very popular in both academia and practitioners?

A: Proprietary algorithm: Of course the details are proprietary, but the general ideas and math are in the books and papers we have published.

Endor vs neural net: Endor requires much LESS data, so it can be much better detecting trends.

Q: How will data be fed in the system? Is this via independent actors or will big companies also plug in?

A: Data fed into system: Just upload (encrypted) data to cloud, give question via examples, and hit “go”.

Q: Why a project that has that much of a track record and that many partners why are they ‘still for sale’ in an ICO round?

A: The point of the crowdsale, as I said earlier, is to take Endor capabilities out of the realm where only the “big guys” can use/afford it, and make it something that everyone can use

Q: In the MVP video, I did see that it took quite some time for the results. Not sure what was the size of dataset? Do we come across similar experiences while using? To put it straight — are there any metrics regarding the time taken by Endor to give our results for different size of datasets?

A: Time for results: It you think that is long you haven’t done much in machine learning! That is several orders of magnitude faster than standard methods. In addition, the computational cost is almost all a one-time upfront cost, and additional questions are blazing fast

Q: Is there a limitation load regarding simultaneous questions being asked across the entire system, and say per user? Will it affect the speed of delivery of results?

A: Limitations on questions: There is a one time “ingestion” cost (not too bad esp. in comparison to standard machine learning) and then it is blazing fast…like google search. So, no limit is expected

Q: Help me understand, does this mean, every time you look for an answer from the dataset you provided — the user has to use the token ?

A: Current plan (as I understand it, I’m a science guy) is pay per use.

Q: Can you elaborate more on the roadmap. what has been done and what is upcoming. and when can the EDR tokens be used for in the endor protocol ?

A: Roadmap: Again, I’m science guy not CEO, but idea is to take current cloud implementation and open up. The simple version of that isn’t too complicated and should go fast.

Q: Have you encountered or envision any regulatory hurdles for usages of Endor for malicious groups? Example — stolen social media data sets used to predict best times to rob a house. Extreme example — nations using social media data sets to suppress freedom of speech and high risk targets. Suppose Company A use Endor will data record will be stored? In system?

A: Stolen data sets and malicious use: Yes, we need to keep track of things, but the fact that the data will uploaded to cloud means that it will be available (even if encrypted) for detecting bad actors.

Q: Now looking at the whole scene, am sure there will be huge data available with Endor going down the line (hundreds of thousands of datasets from various businesses or people asking for questions and getting results) how are you going to use this data? Also is there a facility for other users who can use somebody else’s dataset? Something like have all the datasets at one place and use whatever one needs. What are your plans on this?

A: Data storage: The most exciting idea is that we build a repository of all sorts of data (in encrypted form) so that anyone can do what only the richest hedge funds can currently do: look for predictive signals everywhere. If someone uses your (encrypted) data you get paid. Imagine that a bunch of individuals around the world create a joint “prediction vector” that everyone uses…they could make their living off of being good predictors!

Yes, the idea is that people get paid if their (encrypted) data is useful to others. Result is a data marketplace for everyone, based on the usefulness of the data.

For more in depth conversations about Endor and the Endor.coin Protocol, join us on Telegram — https://t.me/EndorcoinGroup