davidshapiro_youtube_transcripts
/
Proposing GAIA Global AI Agency to research AI and solve alignment mitigate existential risks_transcript.csv
text,start,duration | |
morning everybody David Shapiro here,0.12,4.5 | |
with your weekly video,2.1,6.72 | |
so today's video is about layer six of,4.62,6.059 | |
the gato framework which calls for,8.82,4.5 | |
international treaties uh or in other,10.679,5.821 | |
words Global AI based agencies so,13.32,5.76 | |
today's uh topic is going to be talking,16.5,5.539 | |
about what would those look like because,19.08,6.539 | |
uh in many cases people are just not,22.039,6.281 | |
aware of existing work that has been,25.619,5.341 | |
done on an international or global scale,28.32,4.98 | |
so I figured we would talk we would uh,30.96,5.279 | |
talk about those and um and then also,33.3,4.74 | |
you know first talk about what exists,36.239,4.561 | |
and then also talk about how uh a,38.04,4.5 | |
version of those for AI would look,40.8,3.96 | |
apologies,42.54,4.8 | |
um so real quick just before we jump in,44.76,4.74 | |
I'll plug my patreon I have a private,47.34,3.66 | |
Discord server for all patreon,49.5,5.699 | |
supporters and I also have weekly office,51.0,6.12 | |
hours for the premium tier patreon,55.199,4.321 | |
supporters I also do have a higher tier,57.12,4.919 | |
which allows for one-on-one calls but at,59.52,4.679 | |
least as of the recording of this I have,62.039,5.4 | |
no lots available that being said I do,64.199,6.121 | |
add slots occasionally you know there's,67.439,5.341 | |
Churn it happens okay moving right along,70.32,5.7 | |
so for some background a few weeks ago,72.78,5.4 | |
Sam Altman Gary Marcus and Christina,76.02,3.959 | |
Montgomery testified before Congress,78.18,4.5 | |
about artificial intelligence and one of,79.979,3.901 | |
the things that they asked for was,82.68,2.78 | |
regulation,83.88,4.02 | |
all all three of them asked for it in,85.46,4.96 | |
certain respects Sam Altman obviously,87.9,6.0 | |
the CEO of open AI has been calling for,90.42,5.42 | |
various kinds of regulation for a while,93.9,5.64 | |
Christina who comes from IBM she was a,95.84,5.739 | |
little bit more let's say corporate,99.54,3.96 | |
sanitized in the way that she asked but,101.579,4.981 | |
she basically said that IBM's policy has,103.5,7.02 | |
always been to be more in favor of of,106.56,6.18 | |
Regulation including going back to,110.52,4.559 | |
social media and I checked on IBM's,112.74,3.659 | |
website and they actually do have some,115.079,4.5 | |
policies and recommendations obviously,116.399,5.101 | |
there they have to be very diplomatic,119.579,3.18 | |
about it because they are not a,121.5,2.82 | |
political entity and of course we,122.759,4.201 | |
already we have a lot of in my opinion,124.32,6.359 | |
too much uh interference between the,126.96,6.18 | |
corporate establishment and American,130.679,7.261 | |
politics that being said uh you know the,133.14,8.099 | |
IBM has expressly asked for more,137.94,4.62 | |
regulation,141.239,3.841 | |
um on social media Ai and other uses of,142.56,4.08 | |
technology for a while,145.08,4.32 | |
so Gary Marcus is a professor and,146.64,5.4 | |
researcher so this this panel if you,149.4,6.24 | |
missed it was a good mix of Private,152.04,7.559 | |
Industry as well as Academia and they,155.64,5.4 | |
were all basically saying the same thing,159.599,4.681 | |
so Gary Marcus explicitly said we need a,161.04,5.22 | |
new cabinet level agency so cabinet,164.28,4.86 | |
level agency is like FDA you know or or,166.26,5.46 | |
some someone who's the head of which,169.14,4.92 | |
reports directly to like the White House,171.72,4.799 | |
or or the uh like you know Joint Chiefs,174.06,4.679 | |
of Staff or whatever I know the Joint,176.519,4.561 | |
Chiefs of Staff is uh for military,178.739,3.961 | |
um but you never know like we created a,181.08,2.879 | |
space force,182.7,2.88 | |
um you know NASA was the research branch,183.959,3.42 | |
and now we have a space force who knows,185.58,4.5 | |
maybe we will ultimately have an AI,187.379,4.981 | |
division for military,190.08,4.56 | |
um but anyways so one of the when,192.36,4.739 | |
pressed Gary Marcus said that you know,194.64,4.56 | |
one model that we could pursue would be,197.099,6.541 | |
the FDA but for AI and the the uh,199.2,6.24 | |
example that he gave is that when you,203.64,3.9 | |
release a drug to you know you know 300,205.44,5.1 | |
million people you you make sure that,207.54,4.74 | |
it's safe you make sure that it has that,210.54,3.9 | |
has really passed muster,212.28,4.379 | |
one of the Senators interviewing him,214.44,4.079 | |
said he disagreed but you know he he,216.659,3.961 | |
appreciated the sentiment,218.519,3.421 | |
um and then there was also talk about,220.62,3.899 | |
well what what other existing entities,221.94,5.04 | |
could be empowered right like what about,224.519,6.841 | |
the FTC or uh SEC or whatever and uh,226.98,5.52 | |
there's actually a lot of people that,231.36,3.0 | |
Advocate that say like yes all,232.5,4.739 | |
government agencies must become AI aware,234.36,6.659 | |
AI literate but another point is that AI,237.239,6.36 | |
is such a distinctive and powerful force,241.019,6.0 | |
that it warrants its own uh agency just,243.599,5.22 | |
like how uh America has the department,247.019,4.321 | |
of energy which which focuses,248.819,4.441 | |
specifically on generation and,251.34,5.28 | |
management of energy open AI has also,253.26,4.979 | |
explicitly called for an international,256.62,5.82 | |
agency they call for IE iaea but for AI,258.239,7.321 | |
so the iaea is the international atomic,262.44,5.479 | |
energy agency which has to do with,265.56,5.46 | |
inspecting nuclear sites,267.919,5.801 | |
so the conversation is there the,271.02,4.26 | |
conversation is happening people are,273.72,3.919 | |
calling for National and international,275.28,6.26 | |
levels of AI uh not just,277.639,6.34 | |
regulation yes but I'm also calling for,281.54,4.06 | |
AI research,283.979,3.181 | |
um and I've got plenty of examples of,285.6,3.42 | |
existing International research bodies,287.16,4.44 | |
uh to share so first we're going to talk,289.02,6.72 | |
a little bit about uh the iaea it is was,291.6,7.5 | |
established in 1957 under uh United,295.74,6.36 | |
Nations it is an autonomous entity,299.1,5.099 | |
um its primary purpose is of course to,302.1,3.659 | |
promote the peaceful use of nuclear,304.199,3.361 | |
energy inhibit its use for military,305.759,3.781 | |
purposes and ensure nuclear Safety and,307.56,4.079 | |
Security you swap that out with,309.54,3.719 | |
artificial intelligence and that's,311.639,3.481 | |
exactly what we need,313.259,4.681 | |
uh it is governed by the General,315.12,5.46 | |
Conference of the UN and has a Board of,317.94,4.319 | |
Governors,320.58,4.38 | |
um that is appointed uh from 35 members,322.259,4.321 | |
of the UN,324.96,5.16 | |
um as of 2021 it had 171 member States,326.58,5.82 | |
and its annual budget is only 500,330.12,4.919 | |
million Euros which is probably lower,332.4,4.32 | |
than you might think,335.039,3.121 | |
um especially when we're talking about,336.72,3.84 | |
the astronomical amount of money that,338.16,4.08 | |
artificial intelligence could possibly,340.56,3.78 | |
generate and so then if you say okay,342.24,5.28 | |
well we are asking for the foundation of,344.34,4.859 | |
something similar but for artificial,347.52,4.38 | |
intelligence uh maybe we start at the,349.199,5.0 | |
same part at about half a billion euros,351.9,5.04 | |
if the entire world chips in that's,354.199,4.84 | |
actually a pretty trivial you know,356.94,5.22 | |
amount of money so you divide 171 member,359.039,5.94 | |
states by 500 million euros and like 3,362.16,5.039 | |
million Euros or three million dollars,364.979,5.701 | |
per member State uh that's a drop in the,367.199,6.861 | |
bucket so we could easily easily easily,370.68,6.54 | |
afford to fund such an agency so you,374.06,4.419 | |
know well who's going to pay for it I,377.22,2.46 | |
don't even want to hear that argument,378.479,4.381 | |
like that is a trivial like Nations,379.68,5.04 | |
could sneeze that and not even remember,382.86,4.2 | |
that they spent that money,384.72,5.819 | |
um one of the things that they do is as,387.06,5.4 | |
I mentioned they do inspections to,390.539,3.72 | |
verify compliance with nuclear,392.46,5.76 | |
non-proliferation they also assist in,394.259,5.88 | |
the development of peaceful nuclear,398.22,4.5 | |
programs so they have kind of a one-two,400.139,3.961 | |
thing where they're they're regulating,402.72,3.479 | |
but they're also educating and,404.1,4.68 | |
supporting and it is based in Vienna,406.199,4.021 | |
Austria,408.78,4.259 | |
so you might say okay well yeah but it's,410.22,4.56 | |
the United Nations it has no track,413.039,3.421 | |
record or the United Nations isn't,414.78,4.199 | |
effective or whatever uh you know but,416.46,4.98 | |
there is quite a few things that uh the,418.979,5.541 | |
iaea has helped with uh so the nuclear,421.44,5.22 | |
non-proliferation treaty,424.52,4.72 | |
um it actually has a dedicated,426.66,4.979 | |
um organization within itself,429.24,4.88 | |
um based upon nuclear non-proliferation,431.639,5.821 | |
it helped after the Chernobyl explosion,434.12,6.46 | |
which of course you know the IAA iaea,437.46,4.62 | |
could not prevent the Chernobyl,440.58,3.72 | |
explosion but it helped respond and so,442.08,4.08 | |
that's another key thing is that is that,444.3,5.04 | |
a a global entity shouldn't just be,446.16,5.099 | |
about regulation and enforcement but,449.34,4.38 | |
also like disaster mitigation if or when,451.259,3.681 | |
it does happen,453.72,4.56 | |
nuclear Test Ban Treaty was instrumental,454.94,4.42 | |
in that,458.28,4.5 | |
uh North Korea of course has uh nuclear,459.36,6.839 | |
Ambitions and the iaea provides constant,462.78,5.639 | |
monitoring for that it was also,466.199,3.661 | |
instrumental in helping with the Iran,468.419,3.12 | |
nuclear deal which of course the,469.86,3.239 | |
relationship between Iran and the rest,471.539,4.081 | |
of the world particularly the West uh,473.099,4.281 | |
has been very strained for many years,475.62,5.22 | |
but it hasn't been in the news lately uh,477.38,6.219 | |
due largely to that nuclear deal that,480.84,5.34 | |
happened under Obama I'm not saying that,483.599,4.261 | |
Obama is responsible for it it was an,486.18,3.6 | |
international effort but just pointing,487.86,3.66 | |
out he was president at the time,489.78,4.919 | |
uh the Fukushima uh disaster so that was,491.52,6.78 | |
after the tsunami that hit Japan uh the,494.699,5.581 | |
iaea was part of the coordination effort,498.3,3.78 | |
and then finally,500.28,5.16 | |
um the director general of uh of the,502.08,5.88 | |
iaea Mohammed al-barati,505.44,5.34 | |
um got the Nobel Peace Prize in 2005.,507.96,5.28 | |
so just want to point out that like yes,510.78,4.02 | |
this organization does have a track,513.24,4.44 | |
record and has made a name for itself,514.8,5.94 | |
now one of the organizations that I like,517.68,6.84 | |
to think about is CERN so CERN is a,520.74,6.0 | |
European so it's it is international but,524.52,5.7 | |
is not Global but it is there to study,526.74,5.039 | |
particle physics,530.22,5.16 | |
and so this was established in 1954 by,531.779,6.901 | |
12 member countries it also now has 23,535.38,5.7 | |
total members including observers,538.68,3.96 | |
there's a few categories so there's the,541.08,3.24 | |
founders which are Europe and then,542.64,4.5 | |
there's contributors and observers which,544.32,4.92 | |
includes the United States Japan and a,547.14,4.62 | |
few other nations uh and so this is an,549.24,5.219 | |
international effort to understand uh,551.76,4.32 | |
the fundamental structure of the,554.459,4.021 | |
universe right its entire mission is is,556.08,4.68 | |
literally understand the fundamental,558.48,3.6 | |
structure of the universe specifically,560.76,3.66 | |
the particles that compose matter and,562.08,4.879 | |
the forces that hold them together,564.42,5.88 | |
so as of 2020 for for whatever reason,566.959,4.961 | |
finding the budget of certain is is,570.3,3.3 | |
rather difficult,571.92,3.78 | |
um All Nations report in terms of like,573.6,4.38 | |
percentages of how much they support it,575.7,3.78 | |
but like getting a total number like,577.98,3.419 | |
what was your budget hard to find for,579.48,3.24 | |
whatever reason,581.399,3.481 | |
um anyways I think that the annual,582.72,4.739 | |
budget of CERN is about 1.1 billion,584.88,5.1 | |
dollars and of course they run,587.459,5.041 | |
um uh Large Hadron Collider and a few,589.98,5.34 | |
other major major experiments,592.5,6.42 | |
uh so CERN if you are not aware of is,595.32,5.28 | |
the organization that invented the World,598.92,4.979 | |
Wide Web this is a copy of the original,600.6,6.9 | |
proposal that led to the World Wide Web,603.899,6.541 | |
and at the top someone wrote vague but,607.5,4.2 | |
exciting,610.44,5.82 | |
um so that is uh that is uh uh many many,611.7,6.24 | |
people as particularly in computer,616.26,4.16 | |
science will understand that reference,617.94,5.22 | |
so this was the initial proposal for the,620.42,5.46 | |
World Wide Web uh it's discovered in a,623.16,5.58 | |
whole raft of fundamental Elementary,625.88,7.24 | |
particles bosons Higgs boson creation of,628.74,7.32 | |
anti-matter using neutrinos all sorts of,633.12,5.159 | |
stuff and of course we've got the Large,636.06,3.899 | |
Hadron Collider,638.279,4.201 | |
which generates many many hundreds of,639.959,5.041 | |
terabytes of data and it's interesting,642.48,4.799 | |
to bring this one up because actually I,645.0,4.38 | |
you know as a science nerd I was,647.279,4.081 | |
watching a documentary about CERN and,649.38,5.579 | |
LHC a few years ago and the LHC actually,651.36,6.06 | |
generates so much data that they need,654.959,4.44 | |
artificial intelligence to sift through,657.42,4.62 | |
the data to help find the phenomenon to,659.399,4.981 | |
find the anomalies and so there's,662.04,3.96 | |
actually a very tight feedback loop,664.38,3.899 | |
between for instance nuclear research,666.0,5.279 | |
and artificial intelligence,668.279,6.06 | |
so those are two organizations that have,671.279,4.74 | |
kind of the research mandate and then,674.339,3.841 | |
the regulation and safety mandate and I,676.019,3.481 | |
think we need both for artificial,678.18,2.399 | |
intelligence,679.5,3.36 | |
there's a few other International bodies,680.579,4.44 | |
that we can pay attention to that are,682.86,5.039 | |
probably either going to be good models,685.019,5.461 | |
that we can base this on or they could,687.899,4.38 | |
actually even be partners,690.48,4.38 | |
so one another Global organization that,692.279,4.981 | |
is more for emergency response is the,694.86,4.979 | |
World Health Organization which of,697.26,4.44 | |
course helped coordinate Global the,699.839,4.701 | |
global response to the covid epidemic,701.7,6.18 | |
but they've also responded to SARS MERS,704.54,6.94 | |
Ebola and other stuff so my uncle is a,707.88,6.0 | |
microbiologist and so a few years ago,711.48,4.14 | |
when there was a couple of outbreaks of,713.88,4.5 | |
Ebola he actually traveled the world to,715.62,3.959 | |
help be part of the World Health,718.38,4.68 | |
Organization response to contain Ebola,719.579,5.7 | |
and then of course MERS is something,723.06,4.019 | |
that you might not have heard of but it,725.279,4.381 | |
was infinitely more dangerous than kovid,727.079,4.621 | |
the mortality rate of MERS can be over,729.66,4.02 | |
10 percent,731.7,4.5 | |
um and so like it's a there's some,733.68,4.44 | |
really horrible diseases that you have,736.2,3.36 | |
never heard of because of the World,738.12,3.54 | |
Health Organization,739.56,4.32 | |
um there's two Financial uh Global,741.66,4.38 | |
entities both were created at Bretton,743.88,3.959 | |
Woods after World War II,746.04,4.02 | |
and so World War II caused a lot of,747.839,3.961 | |
people to come together and say we need,750.06,3.899 | |
to prevent this from happening again and,751.8,3.719 | |
so one of the things that they decided,753.959,4.201 | |
after World War II was that economic uh,755.519,4.981 | |
Improvement and economic stability and,758.16,5.34 | |
economic cooperation were critical to,760.5,5.22 | |
maintaining Global Peace which has more,763.5,3.839 | |
or less worked up until that up until,765.72,2.94 | |
now,767.339,3.481 | |
um but it's you know time will tell,768.66,3.96 | |
because we're facing a backlash against,770.82,4.199 | |
globalization anyways getting lost in,772.62,3.48 | |
the weeds,775.019,3.961 | |
so the point of the IMF is to stabilize,776.1,4.56 | |
the global economy through cooperation,778.98,4.32 | |
trade exchange and so on so members of,780.66,7.5 | |
the IMF will get loans the IMF has been,783.3,6.9 | |
responsible for some bailouts they also,788.16,4.38 | |
have been responsible for,790.2,5.04 | |
um not necessarily economic sanctions,792.54,5.16 | |
but economic requirements so for,795.24,4.62 | |
instance during the global recession the,797.7,4.259 | |
IMF was the entity responsible for,799.86,5.159 | |
telling Nations such as Greece you need,801.959,5.041 | |
austerity measures before we're going to,805.019,4.38 | |
bail you out and so the idea of the IMF,807.0,5.7 | |
is that it uses that that ability to,809.399,7.141 | |
withhold funds in order to to shape,812.7,6.72 | |
National policy more towards stability,816.54,5.4 | |
rather than things like hyperinflation,819.42,5.039 | |
or economic collapse,821.94,5.16 | |
and uh you know the situation in Greece,824.459,4.201 | |
is still not good,827.1,4.919 | |
um after many many years that being said,828.66,5.88 | |
that you know we don't hear about uh the,832.019,4.141 | |
Greek economy collapsing in the news all,834.54,3.12 | |
the time granted I don't pay attention,836.16,3.299 | |
to Greek news but it doesn't make it,837.66,4.5 | |
into uh you know American uh financial,839.459,6.841 | |
news uh the World Bank uh is uh has a,842.16,5.82 | |
slightly different mandate than the IMF,846.3,4.2 | |
so while the IMF more focuses on,847.98,5.52 | |
developed Nations the world bank has the,850.5,5.399 | |
goal of ending extreme poverty,853.5,4.56 | |
um through financial means,855.899,4.201 | |
um and also to boost Global Prosperity,858.06,4.62 | |
through lending and Advising so the,860.1,3.66 | |
World Bank,862.68,4.14 | |
um is more likely to issue loans to,863.76,4.68 | |
developing nations,866.82,3.12 | |
and the reason that I bring both of,868.44,3.72 | |
these up is because artificial,869.94,4.32 | |
intelligence has the ability to,872.16,3.84 | |
profoundly impact artificial,874.26,4.86 | |
intelligence all over the world and and,876.0,5.339 | |
greatly Advance the missions of both of,879.12,4.2 | |
these entities and one of the best ways,881.339,5.341 | |
to incentivize aligned development is to,883.32,4.98 | |
pay for it and we'll talk about some of,886.68,3.0 | |
those tactics,888.3,4.02 | |
um near the end of the of the slide and,889.68,3.959 | |
I also have other videos that I'm,892.32,2.94 | |
working on talking more about some of,893.639,3.781 | |
these specific tactics and then finally,895.26,7.019 | |
uh the another research body is eater or,897.42,7.26 | |
iter which is the inner is the,902.279,5.041 | |
international Fusion research entity,904.68,5.399 | |
it's got 35 Nations working together to,907.32,5.759 | |
solve fusion and so the point point,910.079,4.5 | |
being is that there are actually many,913.079,4.44 | |
many examples of International and,914.579,6.301 | |
Global cooperation in order to provide,917.519,5.94 | |
safety stability and research as well as,920.88,6.0 | |
regulations so when we propose to do the,923.459,4.981 | |
same thing for artificial intelligence,926.88,4.38 | |
this is not like this is It's not like,928.44,4.199 | |
this has never been done before we've,931.26,3.78 | |
actually done this many times before,932.639,3.961 | |
so let's talk about how these,935.04,5.159 | |
hypothetical Global entities might work,936.6,5.46 | |
um so the first one that I propose is,940.199,4.681 | |
Gaia the global AI agency,942.06,4.5 | |
um and of course Gaia also means Earth,944.88,4.98 | |
so the Mandate for Gaia would be very,946.56,4.98 | |
very simple study and prevent,949.86,4.08 | |
existential risks from AI That's it that,951.54,5.039 | |
is its primary purpose uh it would,953.94,4.5 | |
publish open source scientific research,956.579,5.161 | |
papers data sets and align models and by,958.44,5.519 | |
making those models publicly available,961.74,5.7 | |
to everyone it will that will one make,963.959,5.341 | |
access to artificial intelligence much,967.44,4.019 | |
more democratic but it will also,969.3,4.56 | |
saturate help saturate the world with,971.459,4.44 | |
alignment research and Alignment data,973.86,3.539 | |
and Alignment models,975.899,4.38 | |
it would also participate in global,977.399,5.341 | |
policy recommendations advising Nations,980.279,5.221 | |
advising militaries and it would also,982.74,6.599 | |
establish best practices and guidelines,985.5,6.839 | |
that could be used by everyone Nations,989.339,6.24 | |
corporations and so on and so by by,992.339,6.42 | |
creating a global Authority a global,995.579,5.7 | |
scientific Authority that says yes like,998.759,6.921 | |
we understand and the idea is is just to,1001.279,7.321 | |
comprehend this problem not to not to,1005.68,5.019 | |
punish anyone not to put on the brakes,1008.6,5.7 | |
but actually to focus exclusively on,1010.699,5.88 | |
Research initial funding should be,1014.3,4.74 | |
around 500 million dollars similar to,1016.579,5.401 | |
the iaea and membership should be,1019.04,5.879 | |
probably pretty similar to CERN or NATO,1021.98,4.859 | |
which are international but not,1024.919,3.481 | |
necessarily global,1026.839,4.261 | |
and the reason that,1028.4,5.82 | |
um I suggest that is because there is,1031.1,5.099 | |
presently a competitive dynamic in the,1034.22,4.02 | |
world and so basically in order to,1036.199,4.62 | |
benefit from this Global AI research,1038.24,5.339 | |
effort then um Nations should have to,1040.819,4.5 | |
pass certain muster,1043.579,4.921 | |
um for instance non-aggressive use of AI,1045.319,6.301 | |
um and uh and you know not actively be,1048.5,5.22 | |
you know at war or whatever something,1051.62,4.14 | |
like that I could be wrong,1053.72,4.199 | |
um I'm not uh I'm not an international,1055.76,5.46 | |
policy expert yet so it might be that,1057.919,4.981 | |
there is there are problems with that,1061.22,4.86 | |
but between CERN eater and um and other,1062.9,6.06 | |
entities it looks like that Global,1066.08,4.56 | |
Research is not necessarily something,1068.96,3.06 | |
that has been done yet but certainly,1070.64,3.779 | |
International research,1072.02,5.58 | |
um yep so that's it for for Gaia,1074.419,6.901 | |
the next one is uh Garza or Garcia the,1077.6,5.76 | |
global AI Regulatory and compliance,1081.32,5.7 | |
agency so this one would basically be,1083.36,8.1 | |
like kind of a combination of gdpr but,1087.02,6.899 | |
for the globe and for AI right so it,1091.46,4.74 | |
would inspect certify and regulate AI,1093.919,4.561 | |
hardware and software so this looks at,1096.2,5.28 | |
the at the hardware and software aspects,1098.48,6.12 | |
of AI rather than like the data or,1101.48,6.66 | |
policy but what it would use is it would,1104.6,7.199 | |
use similar litmus tests and also like,1108.14,6.24 | |
sanctions or regulations as to what the,1111.799,3.901 | |
IMF does,1114.38,3.539 | |
which basically says,1115.7,5.82 | |
um you know if here's an example in in,1117.919,5.221 | |
the investment world in the finance,1121.52,4.74 | |
world if a company is not GDP are,1123.14,5.159 | |
compliant it will not get any investment,1126.26,3.72 | |
why because investors say you're not,1128.299,3.181 | |
compliant you're at risk of getting shut,1129.98,5.22 | |
down the same is also true of ESG which,1131.48,6.42 | |
ESG is a as a inside the industry,1135.2,5.219 | |
standard so ESG means environmental,1137.9,4.44 | |
social and governance so the idea is,1140.419,3.841 | |
that if you don't get that ESG you know,1142.34,3.78 | |
stamp of approval you're going to have a,1144.26,4.02 | |
much harder time getting investment and,1146.12,4.98 | |
so rather than rather than enforcing it,1148.28,5.759 | |
through you know more strong arm,1151.1,6.36 | |
approaches uh my recommendation for the,1154.039,5.161 | |
for the global AI Regulatory and,1157.46,3.9 | |
compliance agency is to perform those,1159.2,4.32 | |
inspections to perform those tests to,1161.36,3.78 | |
basically be an underwriter a global,1163.52,4.32 | |
underwriter of AI companies products and,1165.14,3.779 | |
services,1167.84,4.94 | |
and so then if if the Garcia Agency,1168.919,7.14 | |
decertifies a company then it you know,1172.78,4.66 | |
one their share price is going to go,1176.059,2.941 | |
down because everyone's like oh you know,1177.44,3.479 | |
they're not AI compliant they might get,1179.0,3.66 | |
shut down they might get raided you know,1180.919,4.981 | |
or or investigated or sued or whatever,1182.66,6.78 | |
um but then if uh conversely if they are,1185.9,6.18 | |
if they do get certified as AI alignment,1189.44,5.099 | |
compliant then you know they get more,1192.08,5.76 | |
grants more handouts more whatever,1194.539,6.481 | |
um and so this is this is why I included,1197.84,5.1 | |
the IMF and the World Bank is because,1201.02,5.279 | |
you can you can absolutely steer policy,1202.94,5.28 | |
research development and other behaviors,1206.299,5.161 | |
with financial incentives and that also,1208.22,6.5 | |
creates a relatively lightweight agency,1211.46,5.94 | |
uh that that they don't have to do a,1214.72,4.18 | |
whole lot of heavy lifting themselves,1217.4,3.659 | |
but rather they provide that,1218.9,3.72 | |
underwriting that certification and,1221.059,3.901 | |
decertification which can then steer the,1222.62,5.88 | |
behaviors of the rest of the world,1224.96,6.48 | |
so this overall strategy is about having,1228.5,4.86 | |
gas and brakes because one thing that,1231.44,3.78 | |
people are legitimately worried about is,1233.36,3.54 | |
that if we slow down too much someone,1235.22,3.56 | |
else is going to pick up the speed right,1236.9,4.92 | |
in a global competitive landscape the,1238.78,5.08 | |
nation that slows down is the one that,1241.82,5.28 | |
loses so we need gas but we also need,1243.86,5.88 | |
Brakes in order to win any race you,1247.1,4.079 | |
actually need both,1249.74,4.26 | |
so I play Forza I've actually been I've,1251.179,5.581 | |
I've played Forza since the very uh,1254.0,4.2 | |
original Forza and so if you're not,1256.76,4.74 | |
familiar Forza is a simulation grade uh,1258.2,6.359 | |
racing game and uh you know if you're,1261.5,5.58 | |
driving a Ferrari or a Lamborghini or,1264.559,4.801 | |
you know whatever else a you know high,1267.08,4.92 | |
performance you know Mercedes AMG around,1269.36,5.16 | |
a track you don't just use the gas if,1272.0,5.34 | |
you just use the gas you crash so in,1274.52,6.42 | |
order to win a race you actually do need,1277.34,6.959 | |
both gas and brakes and so this is the,1280.94,5.94 | |
one-two punch of creating two different,1284.299,6.181 | |
entities one for acceleration Gaia and,1286.88,6.36 | |
the other for breaking Garcia and the,1290.48,4.5 | |
idea is that they will have independent,1293.24,3.9 | |
oversight and funding,1294.98,5.16 | |
um which means that also because they,1297.14,4.86 | |
have entirely different mandates those,1300.14,3.019 | |
entities are going to have different,1302.0,3.96 | |
methods of acting as well as different,1303.159,5.14 | |
incentive structures,1305.96,4.32 | |
uh so they have those different mandates,1308.299,4.86 | |
they have different kpi now that being,1310.28,5.34 | |
said they also both would have an,1313.159,4.861 | |
international or Global scope and the,1315.62,4.02 | |
reason is because AI is going to affect,1318.02,3.48 | |
everyone on the planet whether or not,1319.64,3.6 | |
they participate,1321.5,4.74 | |
um and then finally uh part of this,1323.24,6.54 | |
strategy is is both risk mitigation and,1326.24,6.419 | |
encouraging Innovation and so if we do,1329.78,5.94 | |
both then we should be on the right,1332.659,4.801 | |
track,1335.72,4.199 | |
um so here's a few more uh tactics so I,1337.46,3.719 | |
promise that we would talk about some,1339.919,3.661 | |
various tactics that they can use so the,1341.179,4.5 | |
first tactic is just grants,1343.58,4.38 | |
if you want to fund research you hand,1345.679,4.021 | |
out money to do it it's really that,1347.96,2.82 | |
simple,1349.7,3.54 | |
um you uh you incentivize the behavior,1350.78,4.68 | |
you want to see with you know you know,1353.24,3.78 | |
you dangle money and people will go do,1355.46,4.92 | |
it certification and compliance like I,1357.02,5.34 | |
said gdpr is a is a pretty good example,1360.38,5.039 | |
because gdpr is a pretty powerful lever,1362.36,5.939 | |
in order to get companies to comply with,1365.419,6.0 | |
data policy training and education so,1368.299,4.801 | |
many of the organizations that we talked,1371.419,4.201 | |
about do provide training and education,1373.1,5.28 | |
and so by by providing training and,1375.62,4.559 | |
education you can raise the global,1378.38,4.2 | |
literacy on artificial intelligence,1380.179,4.62 | |
whether it's just AI ethics and safety,1382.58,5.52 | |
or the existential risks so by by having,1384.799,4.801 | |
global authorities that provide this,1388.1,4.14 | |
training and education and certification,1389.6,5.04 | |
programs you can say like you know I'm a,1392.24,5.52 | |
Gaia certified AI existential expert or,1394.64,5.58 | |
whatever because right now we don't have,1397.76,3.36 | |
any,1400.22,4.14 | |
global standard of yes this is what the,1401.12,5.82 | |
global experts agree on are the issues,1404.36,4.92 | |
and the concerns and stuff certainly we,1406.94,4.619 | |
have those conversations right if you're,1409.28,3.899 | |
watching my channel you probably watch,1411.559,3.841 | |
like Robert Miles and AI explained and,1413.179,3.36 | |
everyone else,1415.4,3.48 | |
um Eliezer like so you probably are,1416.539,3.661 | |
aware that these conversations are,1418.88,2.82 | |
happening but there is no coherent,1420.2,3.479 | |
Authority there is no establishment,1421.7,5.219 | |
saying yes we are working on this which,1423.679,4.441 | |
is part of the reason that a lot of us,1426.919,2.701 | |
are like hey we need to do something,1428.12,4.62 | |
about this another uh possibility is,1429.62,5.4 | |
that you can actually do competitions,1432.74,5.64 | |
so in terms of AI one of the best,1435.02,6.6 | |
competitions is kaggle which is a it's a,1438.38,5.64 | |
competition platform where independent,1441.62,4.62 | |
sponsors can come in and say hey we want,1444.02,4.38 | |
you to compete to try and solve this,1446.24,4.74 | |
problem over here open AI is doing their,1448.4,5.04 | |
competition right now which is based on,1450.98,4.26 | |
grants for the independent uh or,1453.44,4.38 | |
Democratic inputs to AI DARPA is another,1455.24,4.26 | |
one so DARPA is the defense Advanced,1457.82,4.979 | |
research research projects agency here,1459.5,6.24 | |
in America which funds uh amongst many,1462.799,5.581 | |
many other things Advanced research on,1465.74,5.28 | |
things like PTSD regeneration but also,1468.38,6.9 | |
self-driving cars so the DARPA like I,1471.02,5.279 | |
can't remember the name of the challenge,1475.28,2.399 | |
was like the desert challenge or,1476.299,2.641 | |
something,1477.679,3.12 | |
um started many many years ago was about,1478.94,4.8 | |
creating self-driving cars and so by,1480.799,5.401 | |
creating competitions you get all kinds,1483.74,4.319 | |
of teams coming to you whether it's,1486.2,4.74 | |
universities and independent uh research,1488.059,5.821 | |
groups corporations all participating in,1490.94,5.52 | |
the research for the lure of a prize,1493.88,5.039 | |
which of course uh is a good way to,1496.46,5.52 | |
surface innovative ideas uh putting on,1498.919,6.601 | |
conferences right now there are um there,1501.98,5.28 | |
are international artificial,1505.52,4.62 | |
intelligence conferences like nureps but,1507.26,4.98 | |
none of them are focused exclusively on,1510.14,5.039 | |
global safety whether it's existential,1512.24,6.299 | |
risks from you know autonomous AGI or,1515.179,5.281 | |
even just the escalating risk of,1518.539,3.781 | |
autonomous and semi-autonomous systems,1520.46,4.74 | |
such as weapon systems and so on,1522.32,6.06 | |
so by creating International or Global,1525.2,5.4 | |
conferences where you deliberately get,1528.38,3.539 | |
all the experts in a room talking,1530.6,4.079 | |
together uh that is a good way to,1531.919,5.041 | |
advance the conversation policy,1534.679,3.721 | |
recommendations I already mentioned this,1536.96,4.199 | |
earlier in the video where if you have a,1538.4,5.04 | |
global Authority on something if a,1541.159,4.861 | |
nation comes to you and says hey you,1543.44,4.38 | |
know we want to we want to you know,1546.02,4.92 | |
spurn AI development what kind of,1547.82,5.459 | |
policies do we do in order to attract,1550.94,4.44 | |
that Talent right that's a that's an,1553.279,5.041 | |
example of a policy recommendation or in,1555.38,5.34 | |
other cases where a nation might come to,1558.32,4.5 | |
these these International agencies and,1560.72,5.76 | |
say Hey you know we have a lot of AI,1562.82,6.18 | |
abuse going on people are using it to be,1566.48,4.38 | |
exploitative how do we Tamp down on that,1569.0,4.08 | |
successfully while avoiding unintended,1570.86,3.84 | |
consequences,1573.08,3.599 | |
industry Partnerships are another way,1574.7,5.579 | |
that uh that these agencies could uh,1576.679,4.86 | |
kind of help bring about the change,1580.279,5.76 | |
which uh so an example is the IEEE which,1581.539,6.961 | |
is a which is a uh was it the,1586.039,4.5 | |
international engineering something or,1588.5,5.46 | |
other anyways sets a lot of Standards uh,1590.539,6.541 | |
for and workshopping for standards um,1593.96,5.219 | |
basically created Wi-Fi Bluetooth a,1597.08,3.42 | |
whole bunch of other stuff that you are,1599.179,4.141 | |
familiar with which also helps establish,1600.5,3.96 | |
standards,1603.32,4.26 | |
uh and and uh interoperability,1604.46,6.599 | |
uh consultation and review so this is,1607.58,5.28 | |
something that like I do at a very low,1611.059,4.321 | |
level through my patreon is I have a,1612.86,3.9 | |
bunch of small and medium businesses,1615.38,2.82 | |
come to me where they're they just want,1616.76,4.799 | |
consultation how do I align AI uh how do,1618.2,6.12 | |
I how do I use AI responsibly but also,1621.559,5.281 | |
make money at the same time,1624.32,3.959 | |
um and I would certainly encourage,1626.84,5.04 | |
anyone who has expertise in AI to build,1628.279,5.52 | |
a startup or something around this idea,1631.88,3.659 | |
but this is also something that can be,1633.799,3.38 | |
done at the national or International,1635.539,5.221 | |
level where not just providing you know,1637.179,5.74 | |
policy advice but actually providing,1640.76,4.14 | |
technical assistance,1642.919,4.801 | |
um where where required especially in,1644.9,5.22 | |
the deployment of AI safety and,1647.72,4.14 | |
Alignment research,1650.12,3.419 | |
published standards and guidelines,1651.86,3.299 | |
already mentioned that and then finally,1653.539,3.781 | |
public awareness and messaging that's,1655.159,4.38 | |
what I and a lot of other people uh do,1657.32,4.68 | |
uh with our with our YouTube platforms,1659.539,5.101 | |
Tick Tock and everything else but again,1662.0,5.34 | |
there is no Global Authority uh taking,1664.64,5.399 | |
any responsibility for this messaging uh,1667.34,6.06 | |
which we view as problematic because you,1670.039,5.64 | |
know basically it comes down to it feels,1673.4,4.139 | |
like there's no adults in the room and,1675.679,4.801 | |
that's really scary can you imagine if,1677.539,5.161 | |
you know we lived in a world still where,1680.48,3.72 | |
the World Health Organization didn't,1682.7,4.68 | |
exist even for some of their faults and,1684.2,5.4 | |
failures you still prefer that the World,1687.38,4.5 | |
Health Organization exists because like,1689.6,3.78 | |
I said there's a lot of diseases that,1691.88,2.52 | |
you've probably never heard about,1693.38,3.08 | |
because of the work that they do,1694.4,4.32 | |
likewise you want to live in a world,1696.46,4.42 | |
where if there are nuclear weapons and,1698.72,4.079 | |
nuclear reactors you want to live in a,1700.88,4.38 | |
world where the iaea exists,1702.799,5.101 | |
and so because of those because of the,1705.26,4.38 | |
existence of those organizations we have,1707.9,3.18 | |
this feeling that there are adults in,1709.64,3.3 | |
the room and that and that there are,1711.08,3.54 | |
people paying attention to it and they,1712.94,3.3 | |
have the resources that they need in,1714.62,4.5 | |
order to affect uh good change and,1716.24,5.1 | |
guidance and so this is why one of the,1719.12,4.38 | |
primary things that that I advocate for,1721.34,4.68 | |
and as part of the gato framework the,1723.5,5.159 | |
global alignment taxonomy Omnibus is the,1726.02,3.779 | |
establishment of these International,1728.659,3.9 | |
treaties whether they're modeled on gdpr,1729.799,6.36 | |
or these other organizations this is uh,1732.559,6.181 | |
what we see as a critical path towards,1736.159,4.981 | |
AI alignment and this is not just you,1738.74,5.1 | |
know mitigating uh just for mitigating,1741.14,4.8 | |
existential risk this is also to avoid,1743.84,4.8 | |
dystopian outcomes of hyper you know,1745.94,5.339 | |
corporations becoming quadrillionaires,1748.64,4.44 | |
while the rest of us you know are poor,1751.279,4.801 | |
and live under bridges or whatever,1753.08,5.219 | |
so one thing that I will say is that,1756.08,5.699 | |
these are necessary but not sufficient,1758.299,6.721 | |
and so what I mean by that is that as my,1761.779,4.981 | |
work has progressed as my messaging has,1765.02,3.96 | |
progressed the conversation has been,1766.76,3.48 | |
shifting,1768.98,4.14 | |
so for instance after my axiomatic,1770.24,4.98 | |
alignment video came out a lot of people,1773.12,4.62 | |
reached out saying like Okay yes you,1775.22,3.959 | |
know but there's there's a lot of,1777.74,3.6 | |
challenges like you know who gets to who,1779.179,4.081 | |
who does this research,1781.34,6.3 | |
um but also like what about what about,1783.26,5.76 | |
right what about all these other stuff,1787.64,3.779 | |
and not every what about argument is um,1789.02,5.46 | |
is disingenuous in these cases the,1791.419,4.86 | |
conversation has shifted to these other,1794.48,4.5 | |
what abouts not because of Doubt but,1796.279,4.441 | |
because of like okay we overcome that,1798.98,3.6 | |
problem what are the next problems and,1800.72,4.74 | |
so this list is some other challenges,1802.58,6.66 | |
above and beyond the uh the you know,1805.46,5.459 | |
let's let's say you know we get to wave,1809.24,4.02 | |
a magic wand in tomorrow the two,1810.919,4.38 | |
organizations that I recommended exist,1813.26,4.56 | |
okay still what's gonna happen first is,1815.299,4.5 | |
economic barriers so what I mean by,1817.82,4.739 | |
economic barriers is that,1819.799,4.921 | |
uh not every nation is able to,1822.559,4.86 | |
contribute or compete at the same way uh,1824.72,5.28 | |
certainly the the economic lure is there,1827.419,4.62 | |
for artificial intelligence that being,1830.0,3.84 | |
said there might still be some economic,1832.039,5.101 | |
barriers especially when uh some people,1833.84,5.64 | |
are going to be uh throwing on the,1837.14,3.779 | |
brakes right,1839.48,2.699 | |
um one thing that people are concerned,1840.919,3.541 | |
about is regulatory capture if you raise,1842.179,4.141 | |
the bar so high that nobody else can,1844.46,3.54 | |
participate except the largest players,1846.32,3.719 | |
they have a de facto Monopoly which is,1848.0,4.919 | |
not a good thing cultural differences so,1850.039,5.061 | |
the work that I've been doing on axial,1852.919,5.041 | |
axiomatic alignment and convergence has,1855.1,5.14 | |
to do with finding the underpinning and,1857.96,4.38 | |
Universal cultural values that all,1860.24,4.5 | |
humans share after all yes there are,1862.34,4.62 | |
many differences between nations and,1864.74,4.02 | |
cultures but we're all still the same,1866.96,3.78 | |
species and we're all still on the same,1868.76,5.88 | |
Planet so that based on those,1870.74,6.84 | |
assumptions uh well or facts based on,1874.64,5.279 | |
those facts I assume that we can find,1877.58,4.8 | |
some common ground somewhere,1879.919,5.041 | |
that being said there are still pretty,1882.38,4.08 | |
fundamental differences between some,1884.96,2.76 | |
cultures,1886.46,3.839 | |
um geopolitical tensions so this is a,1887.72,5.04 | |
very diplomatic way of saying that uh,1890.299,4.141 | |
some Nations kind of want to shoot at,1892.76,4.139 | |
each other and they are in either con,1894.44,5.04 | |
conflict or under,1896.899,6.721 | |
um competition which is not necessarily,1899.48,5.64 | |
I'm not going to say that it's good or,1903.62,5.4 | |
bad it's problematic in some respects,1905.12,5.9 | |
and then of course uh scientific,1909.02,4.8 | |
breakthroughs all of this presumes that,1911.02,5.32 | |
with the adequate funding and research,1913.82,4.079 | |
that we will have scientific,1916.34,5.219 | |
breakthroughs on alignment on AI safety,1917.899,6.301 | |
uh but this again is aspirational just,1921.559,5.041 | |
like how how um the eater experiment the,1924.2,5.16 | |
the nuclear fusion experiment it is,1926.6,4.319 | |
aspirational we don't know if it's,1929.36,3.66 | |
actually possible we are hoping that it,1930.919,4.86 | |
is possible uh and then on the next,1933.02,6.0 | |
category is Game Theory so one of the,1935.779,4.801 | |
most common whatabouts that I get now,1939.02,3.48 | |
after axiomatic alignment came out is,1940.58,3.599 | |
what about people that just are not,1942.5,4.08 | |
going to play ball uh what about what,1944.179,5.581 | |
about the uh down Downstream effects of,1946.58,5.76 | |
toxic competition uh which you might,1949.76,4.08 | |
have also heard you know we talk about,1952.34,3.6 | |
moloch and other things so toxic,1953.84,3.959 | |
competition is a very simple way of just,1955.94,2.88 | |
saying that,1957.799,3.541 | |
in a competitive environment people it,1958.82,4.219 | |
creates a race to the bottom basically,1961.34,3.839 | |
perverse incentives and unintended,1963.039,5.74 | |
consequences these are all kind of uh,1965.179,5.88 | |
related to The Game Theory aspect of,1968.779,3.961 | |
this which is that despite best,1971.059,3.541 | |
intentions and best efforts towards,1972.74,4.98 | |
achieving a better outcome you still end,1974.6,6.059 | |
up inevitably Falling Towards dystopia,1977.72,5.579 | |
or Extinction or collapse,1980.659,4.441 | |
uh and then of course there's reach and,1983.299,3.181 | |
limitations,1985.1,2.699 | |
um just because we create these,1986.48,2.88 | |
International organizations doesn't mean,1987.799,4.161 | |
that they're going to work uh you know,1989.36,5.52 | |
if if you get adoption if you get buy-in,1991.96,5.439 | |
if people uh Blacklist the organizations,1994.88,4.679 | |
or whatever and then finally unknown,1997.399,4.561 | |
unknowns uh you know we're still we're,1999.559,4.86 | |
still working our way to the Future,2001.96,5.4 | |
um so we need constant vigilance uh,2004.419,4.561 | |
anyways thanks for watching I hope you,2007.36,3.419 | |
got a lot out of this,2008.98,4.199 | |
um you know in the long run I'm hoping,2010.779,3.9 | |
that all of my videos are completely,2013.179,3.24 | |
irrelevant because some organizations,2014.679,4.261 | |
like this get created and then uh the,2016.419,5.64 | |
real experts get to uh get to get to,2018.94,5.219 | |
comment and um kind of steer the ship,2022.059,4.021 | |
which we don't have right now which is,2024.159,4.5 | |
really terrifying so thanks I hope uh I,2026.08,3.78 | |
hope this made you feel a little bit,2028.659,2.701 | |
better at least in terms of options that,2029.86,5.179 | |
we have before us thanks for watching,2031.36,3.679 | |