can you review carefully?
Forwarded message
From: Greg Brockman <gdb@gregbrockman.com>
Date: 13 September 2017 at 19:01
Subject: elon notes
To: Sam Altman <shga@ycombinator.com>
the following are the notes i took in real-time. call yday:
- elon: talked to sam a fair bit. did you have a chance to connect?
- gdb: yep.
- elan: he was pretty hung up on ceo thing less night, seemed last so
now heading into din. need at beginning who is running the company,
sam was saying everyone would know it's elon, e said don't think
should be diff from reality and title. i don't like being ceo of
anything. intent to not be ceo of tesla once done with model 3 in
particular, he was hungup, (he said showstopper).
- gdb: thought you talked about swapping the title.
- elon: not a certainty, not like he is guaranteed the job. if the
consensus view of the senior team is that they want him as the ceo,
inclusive as elon, right guy to be ceo, would support. don't want to
be ceo for the sake of it, fucking hate being ceo, first startup
tried extremely hard, did not want to be ceo, brother was ceo,
wanted to be basically programming, vcs invested and hired a ceo,
discovered that if you are not the ceo, people can put some bozo in
charge who make you do crazy things that suck. so then after that
was like goddamn it, i don't want to be ceo, if i'm stuck on the
ship and can't get off the ship, then gotta be ceo. given the
importance of this particular issue, can't see why it'd make sense
-- it'd be highest priority thing. arguably is right now but have
short-term obligations i can't release. this is not a case of sam
gets to be ceo automatically. down the road.
- gdb: what would that look like?
- elon: this is wearing on me. if sam wants to join that's great, join
as president/coo, great. if you wanna do something with me, i've had
enough of dancing around the bush. not doing succession
planning. I'm gonna go do my own my thing. am i clear?
- gdb: very clear.
- elan: have enough shit to deal with without this on my
plate. dealing with someone's insecurities not something i want to
do. can be this path or another path. not going to engage in
complicated.
- gdb: understood. let's close out our conversation.
- ilya: so greg willing to do loan for 5.5M, securitized by YC
stock. 2.5 from me. 12M.
- elon: what's the 2.5M that's not YC?
- ilya: i have 7M, half in tesla stock.
- elon: haha really? thank you. vote of confidence.
- ilya: can't sell it. i talk to you a lot. it's done very well for
me
- elon: i support it, when all the cards are on the table, can explain
it to the rest of the team.
- ilya: gave up google cash, got a lot of YC stock.
- ilya: if i understood correct: 12M grant, 2.5M cash, 5.5M willing to
do securitized YC stock (10x larger than the ones from others).
- elon: sounds good. would ask jared what occurred with neuralink. for
any pledging exactly what assets are pledged. that works for me.
- ilya: to make sure on same page: current picture 10M, 10M, 12M,
$2.5M from me, and $5.5M from greg securitized against YC stock.
- elon: think it's something that can be explained.
- ilya: question: i'm taking out the loans. would prefer to return
sooner-ish if it's possible. what are your views on early liquidity?
- elan: yes, absolutely. the way it works at spacex. have been able to
CONFIDENTIAL OPENAI_MUSK00017450
get the best of all possible worlds. the liquidity of a public
company without all the pain. e.g. stock moving randomly up or
down. can be a mood thermometer. at spacex external valuation done
every 6 months, liquidity event 2x a year. spacex participates in
stock buybacks every financing round. want to keep # of shareholders
under control. will do the same thing here. aspire at least once a
year (maybe 2x) can sell stock at a valuation determined by an
outside company.
- gdb: at spacex, was a while into the company right?
- elon: first 5 years, trying not to die. the first liquidity event
where people able to sell 2008. 2006 brink of destruction. to some
measure of prosperity. i suspect 1-2 years. could accelerate that if
there.
- gdb: assume $10M is comfort limit.
- elon: subsequent rounds, i'll invest there.
- gdb: you'll be in spotlight,
- elon: i would like to be in the spotlight the least amount
possible. would like to highlight you. anything you can think of.
- gdb: means quite a bit.
- elon: not looking for awards/praise/anything else. my reasons for
doing this are because my #1 thing keeping me awake at night is
AGI. what the fuck are we gonna do. sure hope this ends up
mitigating that. #1 buzzkill in life. don't want any awards, any
anything. nothing.
if there's anything you can think of that dimishes me and
highlights
- gdb: what would your role look like, as you scale up?
- elon: want to amp up security.
- gdb: over weekend was home and reviewing my NSA stuff.
- elon: playtime is over on this matter.
- gdb: know how to do infosec. less so the physical security.
- elon: amping that up for myself.
- gdb: (reminds quesiton)
- elon: needs to gain a much better understanding on AI on a
fundamental level. want to get to the same level of understanding as
of rockets + electric cars. in order to make right decisions quite
fundamental. a lot of info to upload. way it worked at spacex/tesla
from beginning involved in countless design meetings of every aspect
of the rocket + car. not something that can be created
overnight. present, listen and ask questions, read whatever i can,
look at the software, understand the math, understand the
fundamental algo.
- ilya: are you considering getting hands on exp with writing the
software?
- elon: think i should. wrote all the code of zip2, with lousy
computing tech. lots of video games. stopped programming seriously
in mid- to late-20s. gotta do some of that.
- ilya: think it'd be really helpful.
- elon: gotta close loop around ground truth. gonna make my brain
hurt.
- gdb: definitely made my brain hurt, but not that hard if you really
dig in.
- elon: gotta do it. kind of thing i like. like technology. load up
the old brain. huge part is gonna be a talent war. how do we
convince the best people to join us.
- gdb: if you and sam get more involved, will be unstoppable.
- ilya: new structure will help a lot too.
- elon: alright sounds good. game is afoot. gonna be battle.
- gdb: pretty soon will remember the days when enemy was just DM.
- elon: hah. don't want to pave road to hell with good intentions.
- ilya: don't create the AGI before making it act in our best
interests. should be fundamental tenant. especially once we have the
10GW datacenter, can do a lot of smart things, things they do should
be good for us.
- elon: i feel fear. don't know about you guys.
- gdb: used to feel born 10 years too late. missed the 90s. just
grateful to have an app to help this go well.
- elon: hard to believe this is real. reality is increasingly surreal.
- ilya: let's hope it doesn't take a negative turn.
- elon: don't approach with too heavy a heart. it's a serious
matter. if things do go awry, enjoy these years on earth.
- gdb: mix of hedonistic parties and heads down writing code.
CONFIDENTIAL OPENAI_MUSK00017451
- elon: have a few parties, celebrate the wins. reasonably good time
along the way.
- gdb: over upcoming weeks, how much of your time should we plan for?
- elon: coming weeks, top priority. gotta figure out how do we
transition from non-profit to something which is essentially
philanthropic endeavor and is B-corp or C-corp or something. must
tell the story and not lose moral high ground. absolutely vital.
- gdb: yep.
- elon: need to understand B-corp situation.
- ilya: i have *some* thoughts. one idea keep non-profit and have
C-corp. but i find this approach less appealing because in some
sense C-corp not beholden. one formulation: mission is to minimize
ex risk by building friendly AGI. could be a good way to go. like it
more as new entity beholden to the mission.
- elon: agree, does sound better. never even heard of a B-corp until
sam brought it up. it does sound like right move. would not shut
down the non-profit, should still exist in some form.
- ilya: no opinion, as long as the main entity has something
fundamentally philanthropic.
- elon: any prefs on name?
- ilya: openai. continuity of the mission. it's all the same.
- elon: i agree, think that makes sense. well, i'm pretty excited
about doing this with you.
- ilya: all that's left is the minor thing of actually doing it.
- elon: alright cool. I'm gonna figure out the details of the whole
B-corp thing tomorrow and get that process under way. and then let's
just stay in frequent touch. make this happen as quickly as
possible.
- gdb: super excited.
- ilya: most drastic thing that can possibly be done.
- elon: call quality is degrading. maybe too many people spying on the
call.
call today:
- elon: this is just to go over the board stuff, you may have other
things you want to work. intended to reflect the principle, year to
18 months, that would erode to not having control. over time
dropping to 25%. which is what he has at tesla.
- ilya: we were a bit surprised because thought it was going to be 3
not 4.
- elon: wouldn't be appointing people right off the board, would be 4
of us assuming altman's in. you will meet everyone who is added
chance to say like them or not, if you don't like them won't put
them on. not gonna railroad in people without good judgement. while
i'm ordinarily not an advocate of a large board, because quite
unwieldy, even 16 person board relatively small. real challenge, how
to find enough people w/ good morals, tech, reasonable understanding
of AI. all the properties, really fucking hard to find such people.
- ilya: to make sure that we understand, previously was 3 board seats,
now 4 board seats, but get to veto people with bad morals, anyone i
suggest that you don't like, then can veto them.
- teller: do you mean collectively?
- elan: way i have it set it up here, 2 functioning neurons right
now. if all but 1 of other boards members are supportive. avoid the
one stick in the mud problem.
- ilya: make sense. want to think it over.
- elon: like i said, as we play this out into the future, if we don't
catch up and not relevant who cares, not gonna matter. if we are
actually successful, gonna be a push for >16. then the challenge is
gonna be how do we find, who should decide the fate of the world. i
don't want to decide, want to weigh in. some of these things not
obvious.
- gdb: how do elections work exactly?
- elan: normally in private company, couple of founders seats, when
vcs invest, take board control at the series A. 2 founders seats, 3
vc seats, one reserved for bozo CEO. with each subsequent round,
lead in that round can't be gotten rid of. we can do it in a way
where we can get rid of them. all but 2. don't want to be stuck with
rogue terrible board member. when goes public all stock goes to 1
class. i would be very surprised if we wanna go public. public may
demand that we go public. people vote on future of the world with
dollars.
ilya: makes sense, different from the old proposal, need to think.
elon: way it works in practice, unanimous consent. no one against
adding. only moved forward when the board was in favor. zero
dissenting voices. mostly while looking at candidates.
ilya: i just want to understand, greg/sam/ilya get 3 board members
CONFIDENTIAL OPENAI_MUSK00017452
of our own. do we nominate? how does this work?
- elon: i think that what we would do, say who do we think would be
great to add. if there was a strong disagreement for common stock
seats, could move forward and put that person on the board against
my reservations, but suspect that's unlikely.
- ilya: specific thing: appoint board member
- elon: in practice that's what occurs
- gdb: want to talk about the adversarial case. like with vantage.
- elon: not board member himself, but head of the firm.
- gdb: does the concern make sense?
- elan: certainly worth considering. don't think it's gonna come down
to tiebreak, highly unlikely but possible. think we'll see it
coming. unlikey to be something that springs out of nowhere. if
there's a board member who is consistently showing ways of thought
or
- gdb: vesting. let's say we're fired. should we assume common shares
are forfeit?
- elon: for sure if live. if fired for fraud, then no acceleration. if
fired for angry disagreement, some reasonable acceleration is
probably ok, but comes back to have to defend that with other
people. if say don't have acceleration for termination that's not
for cause, then people will think about it. you guys are not gonna
get fired. have you seen any indication that you'll get fired.
- gdb: if we're not doing well, then should be fired.
- elon: would recommend no acceleration. not gonna be fired
immediately. half of stock is vesting, half investment. extremely
unlikely that we would be fired. if we don't have acceleration,
strong moral basis for refusing that with others. that outweighs the
outcome of being fired. if slight diff
- gdb: common vest over next 4 years, no cliff.
- ilya: last question on how board members are appointed. if i'm
appointing, say i really like this person. others gotta approve.
- elon: essentially, you/greg/sam, 3 board members. I'd be me. gotta
find 3 board members, slight bias towards me, but really who would
be the most useful. anyone you think is not right, not gonna
push. if sam is not in favor but you guys are, would say let's put
them on the board. what we're looking for here.
- ilya: makes sense, need to think on it, since different from old.
- elon: see how it evolves over time, much more collegial than one
would think. really about creating alliance of smart, strong-willed
peolpe who will advance the cause of the company, make the decisions
for the future of the world. hard to find 12-16 people in that
category. some understanding of tech. nailing board together is
gonna be difficult. as soon as we find someone with. if you think
i'm starting to lose my mind, let me know. i really trust you
guys. think you have great abilities, great morals, strength of
will.
- gdb: have you talked to sama?
- elon: haven't, sent him a note. I'm really surprised how hung up he
is over the title. down the road if team wants you to be CEO, not
gonna stand in the way. am an anarchist. personal philosophy is
anarchy. never ran for president of any club, or class president. or
any of those things. have to be boss in certain cases, VC put
asshole as CEO. terrible decisions that would destroy the company,
create products that he didn't know existed to keep the company
alive.
- gdb: my interpretation was sam wanted title to be out there talking
to gov/etc
- elon: not true. if they do, they can ping.
- teller: gwynne does it a lot.
- elon:
never any
question mark at spacex. particularly for any product currently in
operation. at spacex i
that was a case where hard to understand
what questions to ask. now confident something pretty great. at
least version 1. if launch anomoly, something explodes, then get
involved
- gdb: did you say that to sam?
- elon: didn't go into details. for some reason, gwynne started out as
VP. in 2008 when had no choice but to go CEO at tesla, had her
become pres/coo. over time build credibility with nasa, heads of
senate committees, etc. almost never unwound decisions of
gwynne's. once every 3 years.
- gdb: if you don't mind me asking, what's an example decision you
unwound?
CONFIDENTIAL OPENAI_MUSK00017453
- elon: gotta think about it. (spends a minute thinking.)
few cases like that
where super rare
to disagree. put us in two separate rooms and ask a bunch of qs and
get the same answers. 90-95% fidelity.
maybe sam should talk to gwynne. this
isn't meant to mean way that gwynne and my working relationship to
be like how gwynne and i work. sam would get a bit more into the
tech than gwynne does. she is an engineer -- 10 year on spy
sattelites. she understands at a deep level how the tech works, but
i think sam would be much more closely involved in that. this is *a*
model of how it has worked. alleviated a huge amount of time for
me. because had to fly to washington dc a lot less. didn't have to
go to conferences. with very few exceptions don't go to
conferences. don't have to meet politicians. sam seems to like being
out there and being gregarious and meeting people, and being a
connector. i don't really like doing that. that seems to give sam
energy. for me that's painful.
- gdb: i really hope sam feels good, think that the four of us will be
unstoppable.
- ilya: team of the four of us together has the best chance.
- elon: i agree, tell sam the same thing. really value his
capabilities, trust his judgement, enjoy working with him, like you
guys said significantly improve p(success). bring
- elan: think he'll come around, not 100% sure. figure out today. said
trust me, you will be a happy camper. you will not be unhappy, it'll
be cool, will work out well. good feeling about this. neural net
giving a thumbs up. high probability. not indentured servitude. if
he doesn't like it, 6 months down the road, take a while to unwind
other stuff anyway, stop doing that and go in a cliff direction.
- ilya: the band must stick together
- elon: i can whistle.
- ilya: before you hang up, one q. woj really cares about having
co-founder title. from my perspective, very comfortable. if we do to
woj, then should john. to me seems like a great gesture, doesn't
cost us much.
- elon: if you think true reflection of reality, then i support. love
woj, haven't talked to schulman.
- teller: middle ground: founding team.
- elon: in this case, go with co-founder.
- elon: co-founders for the 6 of us, founding team for the others who
join off the bat. if there are a few aces out there where you think
founding team status.
- gdb: how big of a window for stock?
- elon: six months
- gdb: longer than i thought
- elon: 409a come in under stock price. common stock. depends on risk
of IRS battle. optimize the best thing for people now. not for IRS
arg 5-10 years down the road. initial stock: 0.1 or 0.01c, then to a
penny, then 2 pennies, write it as low as we can.
- gdb: at stripe when i bought my shares, i put $6 on
patrick's desk. openai non-profit didn't have a window.
- elon: now have a window. time to pounce is now.
- gdb
CONFIDENTIAL OPENAI_MUSK00017454