00:00all right Steve Nellis here from the
00:03information in the offices of founders
00:06bond here in San Francisco to talk about
00:09automation and how it's going to change
00:12the way that we live in the work I've
00:15got with me - awesome entrepreneurs
00:17where I am from in both therapeutics and
00:20Nick from plethora and I just before we
00:23get started on the discussion maybe just
00:27like a minute or two on what each
00:30because - you know how the company got
00:32started and if the problem that you're
00:34trying to trying to solve do you want to
00:36know any perps yeah sure oh yeah
00:38we haven't work I think it's the cutest
00:40route I guess so we actually started as
00:44a therapeutic research company very long
00:47time ago now and there was as much law
00:50than the company about what we were
00:53doing is how we were doing it so the
00:54therapy except is still self it's a very
00:56long term thing but we built this thing
00:58along the way we call the cloud lab now
01:01Wow which was a basically the ultimate
01:04tool for doing research where the vision
01:07was how little manual lab work do we
01:10have to do can we get to the point where
01:12were really just giving instructions via
01:14the computer that describe how the
01:16experiments should be run without any
01:17ambiguity and then that's something we
01:19could essentially automate or
01:21industrialized so you can do it at scale
01:23and that became a standalone product
01:25itself and it flips of the company and I
01:30feel like it's it's interesting how much
01:31similarity there are between our
01:33companies I mean so essentially the
01:34biotech version are serve the whole
01:36civil manufacturing version where you
01:38know our customers are people who are
01:39making prototypes for you know like
01:41physical production of things so
01:42anything from yell biotech robots
01:44sometimes flying cars all sorts of stuff
01:46they make they need to prototype it very
01:47quickly and usually that involves a
01:49phone call it's like an old guy in a
01:51shop somewhere we've basically automated
01:53the whole process of the pricing these
01:55sort of feedback on the part and then
01:57also the factory will automatically
01:59configure itself and then produce the
02:00parts you want so our whole business is
02:02thinking about like when you engineer
02:04how can you give sort of superpowers to
02:06the engineer the designer totally like
02:08express themselves in the physical world
02:09we started with the production process
02:12things around like the interfaces and
02:14infrastructures that you need to do to
02:15make that possible so first thing that
02:19appears about is what were the you know
02:22what would the technical challenges
02:23they're things that are new or different
02:26technologies that you that weren't
02:32around before that allows you to do this
02:35and I mean why hasn't anybody done this
02:37before I mean you know the biotech
02:39industry in particular right I mean it's
02:41one of the most expensive expensive
02:43industries there is it requires all this
02:46upfront capital why don't you set to
02:47trial but but but there's a big
02:50component that revolves around testing
02:51like there's a lot of monetary incentive
02:53for people to to get this right so what
02:56were the things that fell into place
02:58that enables you to to certificate
03:00systems together so in Emerald space
03:06there are really two big kind of
03:08innovations one we inherited and the
03:10other one we had to build on our own
03:11it gets important this was a possible
03:13product to like build with the first
03:16place so one was we in many ways we
03:19inherited some of the robotic hardware
03:21from there's a subset of biotech called
03:25- food screening which is where you take
03:27giant libraries of compounds half a
03:29million compounds and you screen them
03:31through the same assay like a million
03:33times to try to get early leads for what
03:36might make a good therapeutic and then
03:38it has to get optimized for the process
03:40but that star point was just very
03:42repetitive conducting those acids is
03:44just basically like it's liquid handling
03:46and then incubation times and then some
03:49kind of imaging usually at the end or
03:51some downstream assay that's that works
03:53at scale okay right and so we inherited
03:56a lot of that hardware and a lot of it
03:58was just it was doing the right physical
04:01motions that we needed the problem was
04:03it was designed to do the same thing a
04:05million times and we needed to do a
04:08million things once so the big layer
04:11that we kind of had to add was this idea
04:12of having a a called a programming
04:16language with some sort of control flow
04:17a way to express what you wanted to do
04:20in the lab and give it an instruction
04:23that can get converted through a layer
04:24of software to what eventually
04:26would be command on robots but it gets
04:28all the way down at the hardware level
04:29at the last second but that's not the
04:31thing you want to expose the find user
04:33because you want to express your
04:35experience in a very high-level language
04:36the way you talk to another fellow
04:38scientists okay but I run something and
04:40then it figures out I fill in all the
04:41blanks of okay this motors and I write
04:43this force for this long that's been
04:45draw this something like that okay
04:46so all of that it's a job that you had
04:49to rip out oh yeah it was a lot of
04:52software there's some hardware when we
04:54had to make parts occasionally and it's
04:56useful to have an ecosystem out there to
04:58go through for that but it's a it's like
05:01it probably 8020 software hardware okay
05:04okay interesting and were there any
05:06software breakthroughs that allowed you
05:08to offer using some of the standard
05:14things that everyone's using yeah
05:17definitely like everything lives on AWS
05:19so we take advantage of all of on it's
05:21up there I mean we're using a lot of
05:24modern open-source things so the
05:26platform itself is little on end of UJS
05:28so it's this thing that lets us be
05:30cross-platform very easily
05:31yeah much like what like Spotify doesn't
05:33stuff like that and you know we take
05:37advantage of if we do some high
05:39performance computing parity and with
05:40bottom use docker for that there's some
05:43tools that take advantage of that and
05:45then the big one is we it's it's a we've
05:48really taken imagine this thing called
05:50the Wolfram language oh yeah used to be
05:54not surprisingly math lives behind most
05:56of the science stuff that we do so
05:57inheriting all of those that
05:59mathematical toolkit and the data
06:00analysis and there's like image
06:03processing plotting all kinds of bottle
06:04all the functions that we call on the
06:06yeah stuff it's been very helpful so I
06:17guess the sort of shoulders of giants we
06:19stand on or kind of like on the one hand
06:21you have the design software so 3d
06:23design software yes 10 years it's got a
06:25pretty low adoption until nearly
06:27universal adoption by like professional
06:29so that's one thing it's okay there's an
06:30install base it's almost like the
06:32browser that leads the browser you start
06:34doing something yeah and the second
06:36the actual hardware itself which has
06:37been computer program for maybe since
06:40the 60s around that yeah and then
06:42there's a language from 60s it looks
06:43like it that you can program that again
06:45and you know that's sort of where you
06:47start so I can buy equipment that like
06:48this really shitty language interface
06:50exists on and then we had to kind of
06:52like borrow and steal our way into the
06:54codes to get the API on and then the big
06:57thing that we do is normally the person
06:59who would program that machine is a
07:00human yeah you know so a nest and so we
07:03replicate their brain and say like how
07:05are you gonna do it so a big part of
07:07that is checking the design from the
07:09design software making sense of it and
07:11saying how would we make this and so our
07:13whole system that's the novel thing we
07:14build on the shoulder of giants in the
07:163d world but we have a lot of novel
07:18algorithms that do that in tell us this
07:19it makes a whole like kind of plan and
07:21then I'm gonna talk of individual
07:23instructions so I gotta take this
07:24cutting tool this way and you know all
07:26the different things that a machinist
07:28without their design and program like a
07:31compiler and for solving overheard and
07:34so in both cases you guys are kind of
07:36like ripping out these machines didn't
07:39that are essentially Dom ripping out the
07:42middle of them and you haven't seen
07:44thinner diamond you got some good
07:45interfaces over here and you're ripping
07:47out the middle income like let's do
07:50something that what's this thing that
07:51everybody uses but there's design
07:53software Mathematica which is universal
07:56tools and like how can we actually use
07:59them to manipulate these damn machines
08:00so if that's what you're doing is this
08:03even is it automation yet are we to
08:06automation yet are we just to
08:08augmentation at this point and getting
08:11stuff out of the way it depends from
08:13whose point of view right let's automate
08:15right I mean you ask the question is
08:17super automated not to the driver but
08:19definitely here that user is pushing the
08:21button yeah so from the standpoint of
08:23our users it's like the ultimate
08:24conclusion of automation and on the back
08:26end where they just type in this is the
08:28experiment on them and they get the
08:29result they want yeah on our end we have
08:31to build mostly software but some
08:34hardware to try to make that as
08:35automated as possible soon our whole
08:37facility's 15,000 square-foot facility
08:39it takes two people to keep it running
08:4224/7 right now so it's not like
08:43lights-out but they're advertised
08:45there's still two people running around
08:46like trafficking stuff to get
08:49yeah so two people and how many people
08:52how many people would there be in a
08:54traditional environment for the number
08:56of experiments we clear at least ten to
08:59a hundred times that okay in terms of
09:02the but it's it's a different set of
09:04people too because one of these we
09:06always have to get away from is that
09:08operations team which is what handle
09:10that's the group that handles logistics
09:11of keeping the sample spooling is not
09:13the same as a nutritional environment
09:16you might have a hierarchy of scientists
09:17where there's like P I is at the top and
09:19then there's a bunch of staff scientists
09:21who work for them and then maybe there's
09:22some lab techs beneath them but they're
09:24all adding their scientific contribution
09:26up and down the stack they just have the
09:28program's not a different level of the
09:29problem yeah usually yeah in our cases
09:32we just cut directly to its full control
09:34over to whoever's like stuff in an
09:37Operations is all about logistics that's
09:39all that we get anything quickly but
09:41it's never adding any scientific so is
09:45that a is that a job that you know more
09:50people so is it a lower-level job and
09:54say being a lab tech would've been isn't
09:55comfortable job is it it's well and
09:58salary it's it's a little better
10:00actually but it's like it's a different
10:03job we call it like our analogy when we
10:05go to hire people it's to describe what
10:07it's like it's like being first
10:10responder of science it's like being you
10:12know in EMT where your job is just about
10:15quickness efficiency never making errors
10:17and moving things forward but you don't
10:19want to be making you know start surgery
10:21on the floor like right there you ship
10:24it up to you that's another layer to
10:25analyze so you know our operations team
10:28is never doing things like pipetting or
10:30anything like that it's always a robot
10:31that's doing that but they're handling
10:33usually the coordination traffic between
10:35all the robots okay so get things
10:38think about a plethora I mean are there
10:41you know you're you're moving that
10:43machine operator from the loop that
10:46usually translates what you get from the
10:48CAD design software you know and if so
10:53what kind of productivity sure I mean
10:56it's it's funny like we have very
10:58similar dynamics where the there are
11:01operators on the floor that kind of like
11:02things and you know like more like
11:04physical stuff that eventually will
11:06become robots but now people are more
11:08flexible season that way but the big
11:09thing that we automate is the program so
11:11what would take a human maybe a day or
11:13two to do we do in a few minutes of
11:15server time so that's like the
11:17difference in sort of productivity
11:18between that so that takes like nearly
11:20all what we cost to set up away which is
11:23why couldn't we just like spin up and
11:24make one thing and it doesn't cost them
11:26like it would cost to make a thousand
11:27things traditionally yeah but it's true
11:30the operators are differently trained
11:32like we are entry level people need to
11:34know nothing and then other people need
11:35to know a little bit more to like
11:36maintain the machines and stuff and we
11:38kind of actively bring on into a little
11:39people send them to like school and we
11:41just kind of on board them as like a
11:43conveyor belt of like okay raw material
11:45that has no training and then just keep
11:46moving them through yeah I miss did you
11:49have a sense yeah like you used to take
11:51this many people to run a prototyping
11:53facility and handle this many orders a
11:55day or a week that's a good question
11:59thank you oh you get 10 X 200 X yeah
12:06it's at least 10 X that we don't have
12:08like it doesn't believe Don was mature
12:10you guys are ahead of us like what are
12:11people going to do with death
12:13what are new things that we might see
12:15come out of these industries life
12:19sciences or you know physical things
12:22like and I guess the analogy I'm
12:27thinking about here is software totally
12:30Amazon Web Services which guys are both
12:32probably using and has totally changed
12:35the process for how you create software
12:38in terms of not having to wait around
12:40for weeks for vision new hardware you
12:43get it right when you need it and and
12:45it's made different things possible it's
12:48made it possible to fail faster and it's
12:53totally you know changed the way that
12:56software business works and I'm just
12:58curious if you guys have any insight
13:00Ghana goes very very leyland you've just
13:02you know it early saves development but
13:04are you seeing anything that customers
13:06are doing with this that gives you hints
13:08about what it might allow them to do
13:10analogous to how cloud computing change
13:13the way we did the other software and
13:15education services I mean I think the
13:17large thing there's like a meant to
13:19parts of these for us like one is by
13:21adding the sort of infrastructure so
13:23flexible you can do lots more narrations
13:25in the same level of time so you can
13:26have either a better product when you
13:27get to market faster that's one thing um
13:29on the other end that sort of like
13:31building these interfaces that make it
13:32so people don't have to know the exact
13:34process now so in the manufacturing
13:36world there's a process called DFM
13:38design for manufacturability and that's
13:40like every little thing in your product
13:42is evaluated to see if it'll work you
13:44know that scale or whatever thing you
13:45want to do and so by analyzing it
13:47automatically with our interface we're
13:49able to let me believe that in real time
13:50or normally I might be like a couple
13:51days of like okay like oh you send your
13:53file to some guy I mean we'll use it
13:55then you have a call tomorrow
13:56that is instant so if you don't know how
13:58to do a given process our thing will say
14:00oh start just trying to designer and
14:02then we'll say like hey that's a problem
14:03here's what what's wrong and here's how
14:04to fix it so it's almost I've got a
14:06learning tool and so you can actually be
14:09like right now these industries are very
14:11siloed there's like a metal engineer for
14:13just you know and just machining and
14:14just plastic like what if you can do all
14:16of those things yeah you can have like
14:18superpowers that like one or two people
14:19can like you know say build an app they
14:21can't build a car right now um what if
14:23you could do that and only a few people
14:24could start entering markets and now
14:26require so much money in front you just
14:28lower those barriers of entry of the
14:30design and prototype to get to something
14:31that works and then of course scaling is
14:33the infrastructure is so flexible then
14:35just happens as you need so it's a lot
14:37less sort of cost of capital required to
14:39do I think the the kind of like
14:46emotional feeling among the users is
14:48some combination of like relief and
14:50freedom science is one of these things
14:53that's ironically is supposed to be very
14:54intellectual activity I think every
14:56first-year grad student comes in grad
14:58school bright-eyed bushy-tailed and my
15:01family did these science and and you do
15:03that when you first get there like my
15:05experience was I designed for the first
15:08I would say month to two months I design
15:11laid out everything that was going to be
15:13in my thesis so five years of physical
15:14labor got laid out of us two-month
15:17period ended up being this 500 page
15:20monstrosity several papers out of that
15:22but to get to that realization then you
15:26know very long nights in the lab just
15:28mechanical labor to get there and then
15:30dealing with inefficiencies of the
15:31mechanical laborer probably the system
15:33tons of troubleshooting on the equipment
15:35stuff that was kind of like below the
15:38level of warehouse remnants it is like
15:39pay to play so getting your thermocycler
15:42to work or getting your HPLC to not leak
15:45during a run those are valuable things
15:47you need to do to get data out of your
15:49system but it's not you don't lean on
15:51advancing the science by doing those
15:52things getting to the starting line so
15:54that you can then do it yessuh tell I
15:57think like you see I was having a
15:59conversation my friend and he was saying
16:00he's like 10 returning it up what kind
16:02of crazy company and he's like you know
16:04give me two great software engineers but
16:06give me 10 mediocre mechanical engineers
16:08and the idea is that like there's not as
16:10much automation in mechanical so it
16:12sometimes it's just like find the bolts
16:13that work and that's what you need if
16:16you don't need this champion you know
16:18sort of mechanical person where software
16:20tools that way automated so you want
16:21looking at people who couldn't really
16:22make use the tools and yeah scientists
16:24have to do all this manual like
16:25work is like you know really unfortunate
16:28they're like their brainpower has to do
16:29this you know right and think about how
16:31to fix that they're most likely or
16:32literally annoying so I think any
16:34industry where you see that it's like
16:35okay yeah that's obviously going to
16:37stifle stuff to work yeah yeah that
16:45makes a lot of sense you know another
16:48thing that that occurs to me too
16:50it's confidantes processes is that the
16:53best we can do is not screw the buck so
16:56like if this was this section under Jeff
16:58lost in the sea of Twitter or hello was
17:02saying any interview that I was reading
17:04the other day about when he should you
17:07know build forces by software and and he
17:10said the best thing you can do is not
17:12screw it up then I wish you'd just buy
17:14it and the best thing you can do with
17:17some sort of actual improvement over it
17:19then you should build it yourself it
17:22seems like there's an average of hear
17:24what things automate these are all
17:25processes where the best thing that the
17:28user can put forward it's just getting
17:30it right you know you're not like as a
17:34scientist you're not adding any value to
17:38but getting a gel the receivers yeah
17:41making sure the machine doesn't leak is
17:43that what kind of cure anybody to
17:45previous illness but it is but it is
17:48so are they that hard after people seems
17:51like it makes a lot of sense
17:53I think it's that's definitely the first
17:55level is just taking away barriers to
17:58entry for sure I mean there's there's
18:00also just a physical cost constraint of
18:02like the average first round financing
18:04for a biotech company is like 7 million
18:06and yeah the soccer world it's like 1
18:08million so right it would be nice for
18:09the same amount of capital get 7 times
18:11as many biotech startups you could
18:13imagine that that would propagate down
18:15the line sure and have a pretty good
18:17effect on the ecosystem yeah but then
18:19there's also from the highest level I
18:22think it does something kind of it
18:24corrects a big kind of rowing problem
18:26you had in science which they sometimes
18:30sometimes gets expressed as
18:31reproducibility sometimes gets expressed
18:33is like this frustration with grants or
18:35I'm trying to get very very large labs I
18:37think some of the most interesting
18:39insight I've had there is I was lucky
18:40enough to at some point get introduced
18:42to Sydney Brenner he's one of the
18:45founding fathers of like their biology
18:47Nobel Prize winner and so you know we
18:50met a couple times and having lunch with
18:52him and I got to ask him questions
18:54abstruse tired of hearing the sort of
18:56things but I the thing I really wanted
18:57to ask him is what what does he lament
19:00is most different today than from back
19:02then because it was like the heroic
19:03agent on to the biology 56 yeah we made
19:06this like very rapid progress very
19:08quickly and he had a lot to say
19:11some of us focus on the way grant
19:13funding works but one thing that I
19:16really keyed in on which I think had a
19:18nice parallel here was that said back
19:21when a lab was very simple so when they
19:24were doing research the lab was like
19:25benches and glassware so anyone who had
19:27one it was pretty much the same as
19:28someone else who had one right there was
19:30much to it he said the biggest struggle
19:32he had was trying to get a dishwasher
19:34during the war so like the metal surplus
19:36to get the dishwasher was like a big
19:38deal and that was like the most escapees
19:40to admit they're trying to get and when
19:43that was the case his argument was that
19:44science is very top-down that you the
19:47business of a scientist was to come up
19:49with the most elegant set of experiments
19:52the truth so they would start with a
19:53high-level question something like has
19:56the information make it from the DNA
19:57molecule to the protein molecules very
19:59important and then your job as the
20:02scientists have come up with the no loss
20:05minimal way to get to that treatment
20:06definitively say that this is the
20:10fastest this technique called strong
20:12inference they used to do that but his
20:15lament is like nowadays we're still
20:18trying to the same objective right
20:19versus wrong move science forward faster
20:20but we've become instead of being top
20:22down and become very bottoms up so
20:24because the thing has become so resource
20:25intensive when we have all these tools
20:27that they wouldn't dream of back then we
20:29have math specs NMR is very expensive
20:32you know half a million dollar-plus
20:34machines we have teams that are enormous
20:37so they're 10 times at least the size of
20:40its things that they used to be and with
20:42that I mean he said this kind of a
20:44pejorative he's like the scientists now
20:46big ones they look more like CEOs than
20:47they did back and I like here the bad
20:49thing for them yeah kids you have to
20:51manage all this resource yeah and so the
20:53thing of science nowadays his mind is
20:57about taking from the bottom opposite I
20:59have this freezers have this many these
21:01piece of machinery people with this
21:03expertise have this much grant money
21:04okay what can I do from that that would
21:06advance the science and it's trying to
21:08give it the same goal but the argument
21:09is it could come from the top down
21:10there's no waste it's only the minimal
21:13set of experiences you to advance the
21:14thought from the bottom is claw it
21:16towards this towards that are we sure
21:18we're making progress
21:19you know papers that don't get
21:20referenced yeah stuff it's a partial
21:21solution that doesn't really tell us
21:23that much so by going to a world where
21:29there is one industrial resource right
21:31to produce all the experimentation
21:33whether you're Sydney Brenner with this
21:353040 person lab or you're a first-year
21:38graduate student you're connected to the
21:39same resource so the second you walk in
21:41there we've pushed aside all the
21:43grinding of handling the I have this
21:45part of equipment in terms of cost in
21:47terms of get anything running in terms
21:49of all the troubleshooting you have to
21:50do in advance the staffing is no longer
21:52an issue you don't have to hire 30
21:54people to execute on your ideas you can
21:55do it by yourself and that allows you
21:58just to go back to a focus where it's a
22:01much more meta kradic thing right it's
22:03based on is this experiment actually
22:05dancing us in science right so you don't
22:08have to worry about you know well I
22:11can't ask this question right right like
22:15a parallel to the parallel experiments
22:18and you hear this frustration across the
22:21whole ecosystem like you talk to
22:22entrepreneurs they're big stressors or
22:24how do I get the data I want and it's
22:26gonna require getting its machinery in
22:28getting access to these samples and
22:30getting this many people and if you talk
22:32to Ti is it's very similar like well how
22:34am I going to get access to enough grant
22:35money to do this thing and enough people
22:37to get this thing to move forward and
22:39even in pharma companies it's a lot of
22:40well how are we gonna get 40 people to
22:43pivot around this idea I have to go
22:45convince 40 people it's a good idea to
22:47do this kind of radical thing that's a
22:49very difficult thing to do anything what
22:55are some of the challenges there I mean
22:57one thing it seems like you're really
22:58focusing on is being able to connect put
23:02the designer and take out the process
23:04for a person next to total design right
23:10next to the actual fashion I tend to see
23:14it as like there's a continuum of
23:15engineering sort of like expressivity or
23:17you can start it to something like on
23:18the one end you have like computer
23:20science where it's like every minute you
23:22can iterate you know all the time and
23:23you're so expected to and the other one
23:25looks like civil engineering you get one
23:26bridge you mess it up then you're fine
23:29you know and it's very conservative in
23:31there we use methods from 50 years ago
23:32to do that and then sort of how it goes
23:34is like a civil mechanical electrical
23:36and then computer science and so sort of
23:39what we think about it's like how can we
23:40look at it mature science which gets to
23:42iterate faster and it's kind of
23:43development world lives in the future
23:45and how do we portal those analogies to
23:47the other spaces so you know it's like
23:48it's biotech I don't know where it would
23:49be on that maybe it's like mechanical oh
23:52take a long time and they're kind of
23:53expensive I don't know but it's
23:55somewhere on that that conservatism and
23:56I see it as our job to make
23:58hardware C's get software so if I can
24:00write now the limitations of software
24:02just you know hit a button and it works
24:03that is not possible at hardware they
24:06never but if you can get faster than you
24:08can write the code so in our world looks
24:10like can you CAD things faster than the
24:12machines can make them and we're
24:14actually almost there a plethora
24:15you can like we can beat them like they
24:17can can't design parts that's enough
24:18that we can't make them and send them to
24:20my career that's now and at that point
24:22you're pretty much okay it's just your
24:23brain you now should be able to learn
24:25the tools and then beyond that that's
24:26like the filming it's like yeah like
24:28tooling community designers like how can
24:30you make a designer really know that the
24:31whole space you know what we're used to
24:33in in a very basic way like Amazon will
24:35say maybe you need this you know you
24:37bought the remote do you need the
24:38batteries you know or something like
24:39this such doesn't exist in industry and
24:41it's actually a lot of problems where a
24:43lot of your job is just looking through
24:44catalogs and like understand automatic
24:47fittings of the fittest pump that you
24:48know and you don't know and some
24:50questions are more more like
24:51semantically generated it's like how
24:53would I move the shovel on this thing
24:55you know and you don't know you could
24:56use all these different methods and how
24:57do you decide if they're there even down
25:00to the methodology of engineering
25:01anything has to change so like sort of
25:03upper part of the topic of work like all
25:04these methods of work like design
25:06thinking and scrum and agile and all
25:08these things actually to be very
25:09relevant of how do we I mean a prepper I
25:11think how do we build a hardware world
25:13that has like the ultimate way of doing
25:15it because we have this as tool makers
25:16the ability to really have this giant
25:18lever on humanity you know if if
25:20technology is one of these huge things
25:23that humanity forward our standard of
25:24living and our sort of expressivity of
25:26the individual until like take the sort
25:27of freedom and capabilities you have and
25:29express it in to the environment and to
25:31make everyone's lives better hopefully
25:32um by making that faster and more
25:34expressive and a better solution
25:35shouldn't be as toolmakers be able to
25:37just prove you know everything and
25:39that's you know sort of a giant's need
25:40to do it and I think um you sometimes
25:42it's like how do we even know what
25:43problems we need you know that's one
25:45question and how do you know okay this
25:46problem how do you know the problem
25:47space how do you think okay this is how
25:49I'm gonna solve it from knowing that
25:50space how do i implement
25:51said device poem like whatever you think
25:54is the way of solving it and then
25:56actually how do you make it sustainable
25:57business or whatever you know I think
25:59that's the thing the economy just says
26:01that and by producing these solutions
26:02that we get a better GDP and you know
26:04individual activity everything that
26:06should drive our standard of living
26:07leaves that's kind of high he said
26:09something very interesting in there that
26:10I think ties into sort parallelism
26:14because often it was interesting that
26:15when you put this you stack it up from
26:17like mechanical engineering down there
26:19so yes electrical engineering was less
26:20was closer on the iteration totally yeah
26:23yeah which is a shame because if you
26:24think of it from just like a purely
26:26perspective seems like you know C&C he's
26:30got to be easier than making a microchip
26:31but because the ecosystem and what
26:33Taiwan Semiconductor and virtualization
26:35is done do you want to talk a little
26:37about like it's interesting I mean you
26:39look like a constant why is why is it
26:45actually probably harder to ensure
26:48didn't chip right and I would say that
26:51look and most electrical engineering is
26:53actually not making the check its making
26:54the board yeah so those like the two
26:55problems there's a lot of stuff like
26:57antennas and whatever but like really
26:59you're probably can boards and the chips
27:00probably even exist you know so like
27:02yeah one you left PGA's which like we
27:03can trigger about computer chips that
27:05you have for prototyping now so you need
27:06to make that ship until you know now you
27:08can get a die share meaning like a
27:10really low kind of cost set up for
27:12making a new pad ship for I have a 50
27:14grand for basic tech not just looks like
27:15a man saying you can make a check for 50
27:17grand but you can um you can make board
27:19to like 200 bucks yeah so and you can
27:21you know if you ever try to make I need
27:23this stuff so Cody just do this and
27:25dinner with with electrical work you can
27:27kinda just like solder some trip
27:28together in an hour and it works you
27:29know so a lot of labs without like you
27:31know say Arduino is for example
27:33prototyping board and you just plug some
27:35stuff into it and you know how these
27:36sensors work and they're just they just
27:37work so mechanical isn't like that if
27:39you want a member making our early
27:40robotic stuff and leather and I made it
27:41on manual machines I couldn't like
27:43afford to get sand seats before we had
27:45any funding and so they're just like
27:46pounding that out and it's like God
27:48my grandfather was a machinist and I'm
27:49like how to do this for his whole life
27:51I need on these manual machines and you
27:54know a harbor even CNC is the same way
27:56you know it's it's like yeah it's so
27:58much like I prefer to this do
28:00it manually honestly bandsaw yeah just
28:09like it's like my girlfriend some art
28:10thing and I was like okay I'm gonna make
28:12this art thing god is this him forever
28:13you know and yeah it was like slightly
28:15out of the spec of a plot that I could
28:17do at the time now we can make it so I'm
28:18like okay this is so easy now it just
28:19appears but before it took all day you
28:22know yeah but yeah so your questions
28:23like why electrical is closer to see us
28:25well it's saying even in the
28:26virtualization space like there is this
28:28push a button thing with something like
28:29talent something that very cool
28:31yes sure you tape out and you send them
28:33the chip and it's an elastic product you
28:35can join a shuttle and get like one chip
28:37if you want you're scaling up to ten
28:39units you don't need to any of the
28:41binary troopers anything and the file is
28:43a big thing too so the layout of a chip
28:46and like I guess the the SLI or HDI or
28:49any B's were the other rules like how
28:51the circuit of the chip works it's
28:53portable you can take the same pin and
28:54flash it on an FPGA it's like we can
28:56figure the chip and then you can take
28:58the same thing but okay pull that worked
28:59go to time and send me and say make a
29:01million of them you know and it's an
29:02amazing thing that we're able to do and
29:04mechanical actual doesn't work that way
29:05so in mechanical the CAD file you might
29:083d print or machine as a prototype is
29:10not the thing you'll injection mold cast
29:12as the end product and everything has to
29:14be redone and optimized and that's a lot
29:16of human twittering on various things to
29:18do and it takes forever
29:20and unfortunately mechanical does not
29:22have to things like chips reports it has
29:24hundreds of different components
29:25everything you look at like this is just
29:28blow molding an injection molding for
29:30some little water who knows the other
29:31equipment needs to like make a label and
29:33everything else boards are easier it's
29:35just ok fiberglass copper put some
29:38on there it's like there's a million
29:40companies that do that Oh totally yeah I
29:49mean so there's a few things like if you
29:51start and protonating for us because
29:52that's what the painters like I don't
29:53really need fast iteration Burress the
29:55production maybe a couple weeks is fine
29:57you know yeah but when they need a tool
29:59lot though meaning like the naming all
30:00the custom hardware to make your custom
30:02hardware it's kind of meta we actually
30:04would help make and then it's like well
30:06how do you go from production over the
30:07short run some companies might make 100
30:09objects or each a million dollars that's
30:11actually more like our customers now has
30:13they're like a bunch of custom robots or
30:14flying cars or whatever they're very
30:16expensive they're not you know a dollar
30:18store junk you know that's another kind
30:20of process would make that but you can
30:21help people tool up I'm making hard for
30:23easy and then like like what you said
30:24you buy parts for your stuff so like I
30:26could I'm really happy that we're not
30:28making cheap in the dollar store
30:29I'm happy we're making like a biotech
30:30parts and robots and you know that's
30:32what I want to help in the future
30:33they're not going some future landfill
30:35the floating my factory I preferred not
30:36do that you know what of the day I well
30:42I wanted to I want to push you guys a
30:46so do the journalists thing a bring up
30:48the social implications yeah we talked a
30:50little bit about productivity Brian you
30:53mentioned uber and you know the big kind
30:57of meta theme these days is the robots
31:00gonna come take away our job there's a
31:02lot of alarmism out there but but if you
31:05think there's a serious discussion to be
31:06had about about the social impact of
31:09this when you guys think about building
31:11these businesses what's your take on
31:13that I mean it's it is if you from the
31:15place of the entrepreneur trying to
31:16build for these businesses to worry
31:18about it do you need to have a voice in
31:20you know social policy like so let's say
31:23we do make it much easier for dudes band
31:25access to assign the physical objects or
31:27running experiments but who has you know
31:31who's got the opportunity to access
31:33those things who's gonna make the money
31:35off of them yeah it's kind of a capital
31:37versus labor stuff so as you guys are
31:39going about your thing you know what are
31:41you what are you thinking about this
31:43issue even if it's not directly written
31:46no it's definitely directed I mean I
31:48spend a tremendous amount of time
31:49thinking about this isn't a job that you
31:51pick because it's like a sound rational
31:54choice from the greedy perspective of
31:56like you're gonna be you have to work
31:58your ass off take extreme risk and get
32:02out deal a lot of pushback but the thing
32:04that I think you hit on this bit is when
32:07it comes to the efforts of it right
32:09because you think very carefully about
32:10it what are the ethical implications
32:12here what are the things we need to be
32:13thinking about the long term we're
32:14aligning success the company with the
32:17world we all want to live in citizens
32:19right and i think ii think the key thing
32:22you said there was it's about access
32:23right because if you draw this to try to
32:25draw some larger sort perspective on
32:27what we're doing right if automation in
32:31general go all the way back to stuff
32:32like the industrial revolution and there
32:34we saw factories the first time making
32:36parts that that was a the the biggest I
32:40think kind of thing that makes this a
32:42very different revolution than that is
32:45the customizability is the elasticity
32:47and scale the fact that you get down to
32:49individual parts and one person can
32:51control an individual chart is a
32:53radically different then you have to
32:55have this massive machinery to get you
32:57know Henry Ford in both cars and you can
33:00mom's black right make the one thing and
33:02many many of them we're talking about
33:04many many things wants that changes a
33:07lot and so it does become a question who
33:09has access to it so personally from my
33:11record I don't know if you thought about
33:12this at all but actually you said
33:14something interesting
33:17I think a lot about that because it's an
33:19often a very practical question that
33:21comes to us is to what to do with IP
33:23right there's there's competing voices
33:25in the room there's one group that says
33:27hey no you should be agnostic to that
33:29there's another gift that says hey no
33:30you should take IP and the winners and
33:32that's how you're gonna profit and goal
33:33on yourself with your customers is what
33:36do you mean by that so so here here's a
33:40deal that we could offer that I've
33:42actually I think is unethical and long
33:44so I've decided not to write will take
33:46five percent of your company and give
33:47you free experiments for this much to
33:49return or something like that I see this
33:51is a dangerous thought from my mind
33:53because because the thing that makes it
33:57so controversial Matt is that who has
33:58access if it's his thing that anyone can
33:59roll up and like with not that much
34:02money they can make their own thing
34:03it's hard to think of that as a negative
34:05from a societal perspective we've seen
34:07that in we've definitely seen that in
34:09like I said electronics Hardware where
34:11anybody can pretty much make a chip and
34:13that's led to a huge booming creativity
34:16a lot of what we think about it boys law
34:17is the hack itself it's just ubiquity of
34:20being able to make a competing
34:21architecture and put it into stuff
34:23that's been like a wonderful thing very
34:25different it was much more fraught in
34:26the Industrial Revolution where it was
34:28like we're replacing manifold mechanical
34:30stuff but there are these few
34:31controllers people call them robber
34:33barons right we've run these big
34:34factories it's a different aspect and so
34:38the taking amputee piece for me is about
34:40I actually don't want to decide what
34:42people are good and what their piece
34:43actually has its own therapy initiative
34:45that's doing its own right but my
34:47co-founder and I often describe this is
34:49we're trying to like so we get through
34:51the first thing through and that's how
34:52we saw a lot of the problems that made
34:54us build the system we did well or
34:55honestly trying to build a highway
34:56behind us right where anybody can come
34:59in where any creative can come in and
35:01not just like new entrants but even
35:03people at some of the great formal
35:05companies don't have to arrange
35:07themselves around only one objective you
35:09have 10 times as many interesting
35:10projects going on at once
35:11small groups so do you have a vision for
35:14this where high school kid is pretty
35:17talented can no hop on the Sasse the
35:24liquid version of your software and it's
35:28definitely possible I wouldn't I would
35:31never stop anyone I mean I just have but
35:33they do it like there's there's already
35:34like a very basic kind of genetic
35:36engineering you can do with I gem which
35:37is yet the system where you know what
35:38kind of building blocks for do yeah high
35:40school kids do do this so I mean and
35:42they make some you know not
35:43sophisticated things but certainly in
35:45the future leader stuff they will be I
35:47mean that's kind of freaky to me a
35:48source IP goes right I mean there's a
35:50lot more you know maybe the worst you
35:52can do it clutters make like killer
35:53robots or something with this any famine
35:56or pestilence or you know whatever
35:58there's some amazing things you can do
36:00with this kind of responsibility
36:05yeah and I think that like IP is
36:07interesting as well like yeah like it
36:08just DNA is like if you've got the code
36:11you can do it and they're like maybe
36:12trying to control some of it now we have
36:14some other things for like everyone of
36:15course needs to keep their design secret
36:18the same kind of thing there's like
36:20discussion the board was like oh
36:23you long should you try to get equity in
36:26we're good people assume I mean I think
36:29we've never had the like that particular
36:31discussion it's always around like
36:33biggie enterprises like GE when I was
36:35like you know sign to that black our
36:37files never leave our computer unless we
36:39my was a huge together with varmint a
36:41that very early Rex is that no one can
36:43see this totally totally and so like you
36:45know we built art things so that it runs
36:46locally for the analysis of only when
36:48you hit order does it come in but I'm
36:50interested to in like in hardware and as
36:52an I guess biotech everything's open
36:54source like you could like you know
36:55sequence some bug and figure it out
36:56maybe in the physical world you can just
36:59open up the electronics look at the
37:01board look at the clamshell and say okay
37:03and there's machines like industrial cat
37:05scanners they can actually completely
37:06rip the internals without even opening
37:08you know in a China will just do it on a
37:10map before you bring a 10 and grandeur
37:12versus engineer something so I think
37:13that like we're almost getting to like
37:14mp3 ripping territory with some of these
37:17things hopefully going to get more yeah
37:19products and biological products it'll
37:22be really interesting to see like how it
37:23and I got you can do chemicals to rate
37:24you could like get have some you know
37:26chemical thing as well so you can kind
37:29we take allows that that and like
37:30whether or not look one of the freakin
37:32topics of discussion we're talking
37:35should there be a data marketplace the
37:38cell data I mean and how do we encourage
37:40people to publish actually that's one of
37:41the big things I get AG anemia already
37:43has a way to share a scientific
37:45measurement it's an optional literature
37:46so we actually took it's really borrowed
37:49from like the github come home yeah
37:51where we said alright there's like a
37:52data storage fee that you have to pay
37:54and you generate a ton of data when you
37:56read it's like terabytes of stuff yeah
37:58and you have to pay so we have to pay us
38:02to store the thing but the deal is we
38:03made this this pack that said if you
38:06agree to publish your data which bras
38:07just means you release access to anyone
38:09any user can see that data then we'll
38:12pay the data storage fees yeah so we're
38:15trying to do to encourage you go visit
38:16you know it's not just a charitable
38:18thing it makes the platform better if
38:20you can access all the shared data if
38:22people say silos it's a lot harder to
38:24get stuff and then you will get groups
38:26that if they have all the collection of
38:27data that's quite a collection of power
38:28to at the end of it yeah so it's nice to
38:31get people to sort of share that more so
38:3320 years my jobs these industries that
38:37you serve interesting person I mean I
38:45think in 20 years there's definitely
38:46more I mean I don't believe the thing of
38:48likely is it in the long run which all
38:50I'll say twenty years it's like pretty
38:51maybe midterm but at least pretty long
38:53run that I am scared of the short term
38:56in this like I think that there's that
38:58kind of a few things going on you like
38:59the rate of which automation takes jobs
39:00a rate of which jobs are created by new
39:02industries or at the economy growing and
39:04then of course the rate of training and
39:06so I'm not sure we're good as a good at
39:08training right now as we are at
39:10automating things in some ways but
39:11they'll be unemployment rate in the
39:12States does not bear that out it's still
39:14relatively low there may be inequalities
39:15going up because worst jobs are
39:17happening so I do think that lake
39:19musical are not able to like realize
39:20that productivity the otherwise could
39:21have so I'm kind of interested in like
39:23what those equations look like so one
39:25hand like the curve of training I always
39:27talk about like how the interface
39:28to make the machines that are like my
39:30machines that sit on my floor that make
39:32this stuff are probably hard to
39:33understand and I have a copy version I
39:35bring all new employees in like 10
39:37minutes you can use this thing why it is
39:39really not that different it took the
39:40same thing so I think that in some ways
39:42we we as a like you know economy have to
39:44figure that out but in the long run I
39:46think we will there's a lot of pressure
40:04I sure money into garbage cans we're
40:08seeing them I mean I think that right
40:21now I mean when I talk to the government
40:23I was just at this thing of the White
40:24House talking about the future of work
40:25in manufacturing and stuff as well
40:27and you know the whole point of their
40:29stuff is like we don't know what to do
40:31you know and like all of these trade
40:32school leaders around the country trying
40:34to figure out like oh you know what
40:36welding is now becoming robotized and
40:38you know what are we gonna do and you
40:39know for me I think that like the trades
40:40as an idea is outmoded and we're not
40:42gonna have these like 50 year long
40:44professions where you like you know
40:45you're trying to die being the same
40:47thing we're gonna have to get trained
40:49all the time so these four plethora
40:50we've we've never on training programs
40:52cuz our jobs are novel and you can take
40:54anyone with a good attitude in and yet
40:55the train about so I do think that the
40:56industry because some of this the
40:58government seemingly is reactive to
40:59industry and is saying please help us
41:00because they know it's a problem you
41:02know but I wouldn't say they actually
41:04know a lot of solutions not not that I
41:05found in this thing science is like a
41:08very unique I think that's like a much
41:11more machine like a larger discussion
41:13around automation whole science is this
41:14weird unique space where first of all
41:16the the trend you're saying if we don't
41:18know how people is the opposite yeah so
41:20like if you look in the landscaping
41:21cracks exactly two men Eastern
41:23aggression and not able to execute on
41:25their ideas to get from point where they
41:27can produce something valuable for
41:28society because guys jack bears so we
41:30have this problem where I think in the
41:31latest statistic was like in the last
41:34like three years there's been a seven
41:36percent of drop in salaries for PhD
41:38scientists so my market value is going
41:43it's very different and a lot of it has
41:46to do with the fact that we can only
41:47pursue so many projects at once and
41:49there's this other trend coming on give
41:51your hurt even's law okay so you've
41:54heard of Moore's law right yes yes so
41:55you as laws it's kind of a corporate
41:56joke it's more backwards in in the
42:00biotech industry if you look at the the
42:02amount of money has to spend for
42:03approved drug yeah that exponent
42:06it's every nine years it doubles and
42:08it's been doing it since 1950 yeah right
42:10so the didn't number if you think of
42:13this as like a big pipeline of you know
42:15the value that's being pulled in today
42:16is what we're providing health care for
42:19the most part if ran biotech right if
42:22that is so slowing down the whole
42:25ecosystem has a problem right so in many
42:27ways we to get almost no complaints at
42:29all because anyone who speeds up and
42:31gets faster to getting to the point
42:34where this is like a real medicine that
42:35we can actually hand to people I think
42:37people will be quite happy to pay for
42:39cures to HIV and these sort of things
42:40and that whole thing helps the ecosystem
42:43as a whole and I'll get something faster
42:45so it's a I think it's a very different
42:47industry because it's not like like
42:49we're having a problems producing it
42:51Simon ran out of ideas or like problems
42:52to solve or it's like wow we're like do
42:55we really need another here to be like
42:57yes absolutely no I'm saying I don't
42:59know that right so like there's
43:00definitely a demand on the other end
43:02this is essentially a vertical command
43:03to pull this but our inability to supply
43:06it is creating the problem so whenever
43:08you lower those various entry when you
43:09speed up the the entrance to get there
43:12that you go system to hole test that
43:13grow convict well I think that it
43:17concludes the amount of time we have
43:19today but guys thank you so much for
43:22joining us and father spend thanks for
