Next Economy 2016-10-11

A lot of people thinking about changing economics come from the Rust Belt and know people hurt by it

Young people don't have a lot of role models to look to for ethical business

you're looking for ways to do things well, rather than just nihilism

People who did this kind of work are the healers, and I put @timoreilly in that category

your family furniture company was in trouble, so you left for your own company?

it was a company founded by my wife's father that I joined

my family's firm had been making wood furniture since 1919, and in 2001 the WTO meant china could compete

Wall Street wanted us to close or move production offshore

we had to work with all our staff, and come togetehr, and we came up wiht 5 great rules

our people thought we could not compete, but we had to convince them to try

you turn it round with leadership: don't ask anyone to do anything you wouldn't do yourself

we cut salaries before we cut wages, and made sure we were there before the workforce came in

This is something that happens more in Europe - especially in Germany: they have family firms that do this

what do you hear from your peers?

At my age I don't have many peers. I was told I was born with a silver spoon in my mouth, but I wasn't special

I was brought up to know that I owed service to the community, with the parable of the talents

I will tell them the good news, and the bad news, and never knowling lie to my team

I have found that midwest workers have a deep understanding of globalization that bankers don't see

business schools say to marshal your capital and treat labour as a cost

You have to drive change. Don't panic when people tell you what you can't do.

They tell you that wages are cheaper and regulations are non-existent. So what can we do?

they are on the other side of the world - we made sure that we could deliver to our dealers in just 7 days

My 5th rule is teamwork and communication - people hear rumours. Talk to them and show that you are on the team

Our health insurance was good, but people were dropping out. We made our own clinic in the factory

I'm the founder and CEO of guideline - a retirement company built for small businesses

small businesses often have 401k plans, but they have multiple 3rd parties involved

the 401k ssytem was set up 30 years ago, and is intentinally deceptive so people don't know they are paying fees

each of these fees add up over time, and eat away the retirement funds

we brought all those services in house, and replaced them with code

Josephine.com is an on demand food service founded in Oakland, currently closed for regulatory reasons

Josephine doesn't work with restaurants but connects you directly to home cooks

I was supposed to be joined by Crystal, one of our cooks, but she wasn't able to come

even though industrial cooking has made food cheap and available, the best food has been made in homes

we want people to tell their story behind the food, and let the cooks be the face of their own brand

the food industry is synonymous with minimum wage

in food there is a huge informal economy - there are 2 million professional cooks and 10s of millions more

We think there is a more holistic food experience that the commodified commercial businesses

our cooks are largely folks excluded from the professional world - 75% women, 30% minority

Crystal used to run block parties and local music events, and sells gumbo through josephine

the government sees the economic value, and the grey economy that was impossible to regulate before

we want to be more proactive - we are currently a private marketplace. we sponsored state level legislation

we have a new permit process for home kitchens at state level,

20% of th company is going to be allocated in cooks starting in 2017 - we are working out dividend models too

we have a cooks council that helps decide our business direction and legislative advocacy

Bryce came up with an idea a few years ago called indie VC - without the pressure fro a big exit

we have a flaming unicorn on the indie.vc site

we've invested in companies at OATV, but indie.vc takes a different path

so much of the classic VC model is based around the next fundable milestone, not a long term durable business

the trade off between what companies wanted to build versus what they had to do to scale was obvious

I was attracted to this because O'Reilly was built that way - - starting with $500 and not investment

we're trying to give companies access to a little bit of capital and a lot of mentoring -

Otto, media.net and applovin were recently acquired for large amounts with no VC funding at all

there were companies that took funding in the run up to IPO, without putting it on the balance sheet

you figured out a way to invest in companies that could stay small and not go for the casino

we aren't opposed to unicorns, but some businesses reach scale that isn't huge but sustains

it's like me or basecamp - you don't want to sell it

so what we do is once you start taking cash out of the business for yourself, we share in cash distributions

so you're sharing the proceeds of the business, not the equity of it

we have the VC funded archetypes for startup founders - we need to surface more like atlassian and esri

our next guest @Jessicalessin of The Information is exactly the kind of company that @bryce would fund

I'm in fanboy mode - I have been a massive fan of The Information since it launched

I thought bloggers were supposed to kill media companies, from the outside - what did you do different?

it's the business model. Reporting isn;t broken, the news business is broken

the vast majority of new outlets are chasing traffic and pageviews

you get 19 versions of the same story on each site and nothing in depth

charging $400 a year was something people said wouldn't work. but we doubled cashflow and have 19 employees now

we're not going to write the news of the day, we're going to write content worth paying for for decision makers

as our staff has grown, we've stuck to 2 stories a day, and made those as good as we can make them

the impulse is to publish, so it takes control to hold back like this

All the big publications have their hit stories from time to time

if you look at what is truly unique on the broader sites, our 2 stories a day are competitive

I often say we'r competing with instagram - we want our email in your inbox to be as compelling as instagram

a lot fo companies say "i'd start with an instagram account, not a website"

that's crazy - you need to have a website and be so good that the readers come to you

you need to have control of what you are publishing, and not have it stripped out by the generic

Vice and buzzfeed and Vox have raised 100s of millions of dollars -how are they constrained where you aren't?

i don't think news is a venture backed business - news is not capital intensive

you can seed your company in myriad ways - not raising money from VC investors

if you look at the big ones you mentioned, they have flip-flopped to social, to video to conferences

companies tend to measure themselves by the amount that they have raised - what do you want to do that you can't

I ask myself all the time if there is anything missing, but so far we aren't missing out

our growth strategy isn't to hire 50 reporters tomorrow, because we don't know we could find them

right now I care a lot about subscriber growth - it is revenue, but it also shows engagement directly

I imagine you are getting lots of calls to raise funding. Why nto take some?

because we want control of our destiny, we want to keep it growing the way we wnat

when the human agent overrides the AI, that's when the AI learning kicks in

we make money at Lola the same way my last company Kayak did -we get commission from the hotel sale

at the moment we employ travel agents and have a rating scale to pay them if they make customers happy

Lola is designed to be an open platform to let people become travel agents -the 1st 20 are employees

a travel agent is a special customer service -will you do other things?

initially, we thought of Lola as a secretarial service, then one of our investors suggested travel focus

our ai engine is based on recommending from past travel, and also on natural langauge understanding

so if there is an easy question the bot will answer it, but the human agent can overrride

part of it is the agent understanding my values - I don't want a remote resort but to be part of the country

the humans can do things the ai can't - when you have analysis paralysis they can help you decide

the humans can also negotiate on your behalf with the hotel, which an AI can't do

I have always done philanthropic work with things like Meals on Wheels - now I can do more with more money

Partners in Health is a deep metaphor for the next economy - rethinking the business of healthcare

there is a project in Boston now that is copying what we did in Haiti

if you have a human who visits you in the home you can debug thinsg earlier than landing in the ER

A key next economy company trait is building platforms - these people built the platform for the Sanders campaign

Becky Bond and Zack Exley built a more complex platform to organise the campaign

normally what people do after a campaign is start a consulting company, we want to open source it

the idea that bernie supporters won't vote for clinton was mostly a myth - moe than clinton did for obama in '08

we were able to connect people to each other with slack and facebook and gogole docs and they still have those tools

the grassroots network is still out there and they are organising things themselves

Trump is not going to win, I don't want to jinx it

we must always believe that everything we do matters, and collective action only works if everyone steps up

I'm worried about 2020 and 2024 as Trump has let the genie out of the bottle

we have this campaign called "brand new congress" started - I'm sure Becky will come too

we're running 400 candidates in one slate with one plan - to rebuild the economy

youre not talking about picking and choosing candidate, but running a new slate entirely?

we want to take 400 ordinary people who aren't politicians or narcissitic billionaires to run on one slate

we have found that people are willing to things that are a big effort, rather than things at the margin

we embrace failure on the campaign - we were trying to do the moonshot, and ti was worth

My favorite poem is by Rilke - what we want is to be defeated decisvely by successively bigger beings

Yesterday the IBM CIO was talking about turning his IT team into squads of 10 with autonomy - did you do that?

the uber analogy - one of the chapters in rules for revolutionaries is the revolution will not be staffed

normally we work with organisers to climb the ladder of engagement, but we don't have any staff

so we had to have a system which is more like the Uber or lyft dispatcher that was decoupled from staff

the support infrastructue we had enabled us to hold public meetings with 500 people to commit to each other

we've always worked on the technology side of politics, but what matters is getting people out to vote

it wasn't just a CRM, but extended to person to person meetings to follow up to make stronger connections

the stronger connections between real people in real life, then captured what they wrote on paper online

we evolved the process to be like an old-time revival format - trying to do something big

we sent out emails to do phone banking, and the response rate was minimal, in person it was huge

if we were able to explain in person 'this is part of the plan to win ohio'

we'd ahve 400 people in the room and say 'we need 10% of you onsateg to organise" then the rest to meet them

so getting them to meet in person and commit to each other meant that they would fulfil it

to get people to get their meeting right we had to talk about it like a technology

a sheet would be filled out and then everybody on the sheet would get an eamil, by photographing them

this wan't possible in 2008, but slack and google drive and phone cameras meant we could do the data entry

we had these 22 rules in the book - the enterprise will not be staffed -only hire people who believe that

I'm going to talk about my robot lawyer and the future of the legal profession

I was a terrible driver and got a lot of parking tickets, and my parent said I had to pay

I worked out how to get out of parking tickets by reading obscure legal precedents

I became the local parking ticket guru, and I realised I could automate this

I ask a series of questions about the parking ticket and make a legal dcument you can send in

it has now contested 180,000 tickets and saved $5M for consumers

I have done the same thing for delayed flights - automating compensation for this

people started asking me for help with repossession and eviction

I worked with Centre Point the homeless charity to automate homeless appeals

government is going to get much more efficient with the help of bots

Lawyers are going to be consumed by technology. I am just 19 and I have made lots of enemies in the legal world

you took over Watson - what has happened?

an example is how Watson diagnosed the illness of a 60-year old woman in minutes and she is now getting better

we're automating the base level grunt work - we can see what we understand already and what we don't

and focus the human work on the unknown parts

freeing human capacity to solve hard problems is what we want to do

when the doctors overrule the radiology, what do you learn then?

one example is 30% of diagnose were made by humans, and can improve over time

how do we grow this as a symbiosis - it becomes clear that we need more data, but how do we increase capability?

we're bringing down the cost of things which provides more access - India has only 1000 oncologists for 1Bn people

by doing more diagnoses automatically and screening we can save the oncologists from visiting small villages

if people work with AIs how does that make the humans more performant?

we're not necessarily looking for the AI to provide the answer but to ask better questions to help us see it

IBM has been working on AI for six decades - it is the bulk of our research today

the other tend to start from a big consumer franchise with a big knowledge graph

there are a lot of domains - medical, legal etc that don't necessarily fit into those categories

people are usually worried about regulatory burden from government

rules and regulations are things bots can do well, especially with things like auditing

The emirate of dubai uses Watson to help you fill in forms to start a business there

regualtions create guidelines, so you can learn them better

when Watson finds that regulations that contradict each other it ranks them by degree of penalty

the mess around the world where different countries have different regulations also shows up

can you teach Watson to be a Congressperson?

well, we elect congresspeople but watson could help lawmakers and their staff make better regulations

we're really trying to open Watson up to developers so more people can make sense of it

what is the future of Watson, Tensorflow, azure cloud etc? what will that unleash?

there are 3 patterns: learning rules and how to comply with them - legal, medical and more

2nd explore and discover to find new ideas - combining molecuels for medicine or ingredients fro cooking

3rd is bots and agents for creating interfaces that people can use for specific domains

what are the most interesting things to come from increasing data and sensors?

I love sensors - that was what I learned working in weather

if we're all wearing sensors in our bodies, we can know cancer, leukemia and fitness ahead of time

we're also going have augmented reality in our glasses to make the world clearer for us

when is watson going to get a PhD or be certified as a radiologist?

we will have watson take the radiology exams and pass the bar and qualify as an accountant

Urs Hölzle has been vp of infrastructure at google since the beginning - what are you builing now

we started out looking at costs, but once adwords kicked in, we looked at doing things that were impossible

now we are looking at making machine learning possible and efficient

wsatred out wiht mapreduce it seemd impossible, now it is homework, adn we even have BigQuey to do it for you

what si the architecture of modern in the world applications liek cars?

used to have to do it in the datacentre, but now you want to do more at the cleint side too

traingin a large model is super computationally intensive - trillions of flops

we ahd to make special purpose hardware fro the neural network tasks so that the costs din't blow up

we ahve so many translations and speech recogntion to do we need to automate further

our datacentres are pretty automated, but our Iowa datacentre has 100s of employees there too

we want the workflows to be as automated as possible so the peopel can work on the exceptions

it's not a cost thing so much as an agility thing -you need to follow the curve when you have demand

the datacenters are in out of the way places, located near cheap energy -can you make the grid moe efficient?

a lot fo the grid is built on 50 year old technology

we were able to use AI to find the right knob to make cooling 40% more efficient, could do that fro the grid

IT is very inefficient, especially on-premise - by concentrating in the cloud you can mix workloads

if you do the heavy lifting in the cloud you can have a phone not a PC, and save 2x energy

its amazing how long it has taken to move data to the cloud

The key thing is to make cloud really accessible fro the average company - thousands of companies can benefit

someone was able to build a satellite imagery cloud removal tool in cloudML - improved it by a factor of 4

a novice was able to make a huge improvement using an alpha version of the tool we have

just as mapreduce made it possible to process a lot of data as a non-systems person, cloud machine learning can

there are big opportunites in understanding data and improving search - the idea behind google now and assistant

you don't want to waste time copying and pasting as a consumer - the same is true for enterprise

you're limited by your ability to apply technology, not the opportunity to make a difference

Google was about organising information and making it accessible; cloud does this for technology too

this is different from 'here's virtual machines' - here are rich tools

what do executives need to do to be prepared for technology infused future?

this is not the future, this is the present. this si doable by the average company- the tools are available

you need to try things out - the Airbus cloud removal didn't think it would work that well

machine learning has an opportunity to create jobs by doing things more practically than before

I'm an optimist by nature, by running large infrastructure you have to have a sunny approach to survive

There is a big change coming over the next 20 years, but we are better prepared than the industrial revolution