Maxim Fores on Leading PAI: Coordinating Global AI Safety Efforts

A lot of people think of the situation as a race between different labs or a race between [music] the US and China.

That's not reality. The reality is there is a race between people who are trying to build a godlike [music] entity that

will most likely kill everyone and us, the people who are trying to [music] coordinate to stop them. So what we need

to do is get this coordination to be on an exponential path as well. So exciting news this month at Pori. We

are now announcing um a new director who's going to be coming in. Many of you will already know him as Maxim Fores who

has been the director of Porzi France for a while now. Um he's going to be coming in. Um, I'm personally quite

excited to see how things will change and how we can make a stronger case for the need for AI regulation going

forward. Thanks for joining us today, Maxim. Um, how's it going? >> Thank you. Uh, it's going well. I am I'm

not sure I would say that I'm excited to take this role just because it's a I think it's a tricky word when we are

talking about the stakes that we're facing. Um, but I would say that I feel very aligned.

I feel I feel the weight of the responsibility and it's heavy, but there is a a specific kind of energy that

comes when you know that you're doing exactly what you're supposed to be doing. Um, so yeah, I don't do this

because it's fun, but I do it because it's necessary, but when I look at the team that we're

building and the momentum that we have, I feel like determined, I would say, and and ready. So, in the in the last couple

weeks, we've seen uh Google released their new model um Gemini 3. Um this was after PAI's protest um back in the

summer, given that they they hadn't um lived up to some of the safety commitments that they had previously um

promised to do. Um and we've also seen some quite drastic improvements in the capabilities of um Gemini 3. So, what

what was your take on the release of Gemini 3? Well, I mean, first I would say it's

positive that uh Google has decided to release the um the safety card after all demands. On the other hand, this looks

like the most unsafe model that has ever been released. So, what we have gained in transparency, we are losing in the

amount of risk that is being taken. Uh I say that it's also I'm not sure it's going to be a nail on

the coffin of the people who believe that AI is not improving and is going to plateau.

Uh what I would say is that I hope it's going to stop improving. I hope it will plateau soon. But uh really we shouldn't

we should we should prepare for the case where it doesn't it doesn't change our actions much much because if it actually

plateaus then that's amazing. It gives us it gives us more time but we're still late right if we get AGI in 5 years

that's tricky. If we get it in 20 years that's better but we still need to prepare right now even if that is the

case. I believe you've been um head of Borai France for almost two years now. I may be wrong about that. Um but

obviously before then you also existed and you were doing things in the world. So give us a little bit of the the story

of Maxine. Before Pai existed, what were you up to and what we do? >> So before Pai, I was an engineer and

researcher in artificial intelligence. Although so I was I've been doing that for about 10 years. I wouldn't call it

artificial intelligence for most of my career. I think uh I started call calling it this way around 2020 when uh

it became a bit clearer to me that we were getting closer to the real deal. [snorts] Um

and I so I was heading a research and development team in AI in a systematic hedge fund called the true sigma. Um I

think I I started understanding what was happening better around 2022 where I started really feeling the acceleration.

Um and this is when I started uh coming to my senses and thinking about what we were doing. I it's not the same thing if

you're working on a technology that you think will actually arrive in 200 years versus when you think it's going to

arrive in 10 years. I took a I took a sabbatical to think about it and then GPT4 came out and uh that's the time

when a lot of people started waking up to uh the reality and that's when I I realized the the incredible amount of

risks that we were taking. uh and I shortly after that decided to join POSAI as a volunteer.

>> Just out of curiosity, what were your kind of timelines before the release of um GPT2, GPT3? Um because I think the

timelines shortened for almost everyone um but what how far in the future were you looking when we're talking about

things like AGI or super intelligence? I think uh before GPT2 well yeah I think the key moments for me were when uh when

Leedle the the the um one of the best world players in Go was beaten by AlphaGo at this point I

started thinking huh uh maybe we will eventually create an AGI I think before that to me it wasn't entirely clear

whether there would be a a hurdle it would just be impossible forever or something like that which could sound a

bit naive now but uh so at this point I thought maybe okay 100 years from now we we'll be there then when uh when GPT2

came out and when people started realizing the that it had emergent uh skills like it was able to build

websites run a prompt stuff like this then I started thinking oh maybe I'm going to actually see that when I'm

alive uh so I thought maybe 50 years and It's only but I wasn't thinking about this too much, right? It was really low

in my the priorities in my life just down at the back of my head. Uh I think it's only when GPT4 came out that I had

a very sudden realization that we may be 5 to 10 10 years uh from AGI. And after the the kind of um huge boom

in in AI capabilities we saw in 2022 2023 more experts began to kind of come out of the closet as it were and began

to warn very seriously about what the risks may be of advanced AI. Um and then early on in uh 2023 um you founded uh

Paul's AI. Um how long after the the founding of Porzi did you join and what were your reasons for joining? Why did

you think Paulai was an organization that was needed at the time? I >> remember correctly. I joined in November

of that year and I joined because um well I I tried to figure out how I could have the most impact. Like I thought the

situation was pretty bad. Uh but I've always been very optimistic. I think that uh people who think that we are

doomed and there's nothing to do it's mostly an excuse because uh it's very handy right if you think that then you

don't have any responsibility on your shoulders you can just chill and and look at the events unfold and like be

very cynical and and and be like oh I told you so uh I'm more of an optimistic myself I think that we can change the

course uh I noticed that a lot of people were already trying to change the course, change change the trajectory,

but they were going through the uh inside game. So, um building think tanks and trying to speak directly to

politicians, trying to speak to people in the labs, which is great. I think a lot of people should do this, but I I

thought that what was really missing is just public awareness. Um, so I I was looking for a movement or

anything that that would uh try to achieve this and POSI was here and yeah, I think it's founded on the correct

idea. I think that you cannot um put this technology on pose if you don't

have a fairly global uh awareness of the risks in the general population. So that's why I joined this movement.

>> Um, and obviously perhaps the main thing that PSAI is known for are our our protests, right? Um, and for me

personally at least, I had never attended a a protest um before joining PSAI because I had never felt um

strongly enough about something in order to feel it was um apt to to go to a protest. Um, do you had you been to a

protest previously? And why is it that protests specifically are something that you think may be needed in order to

raise public awareness and and perhaps show politicians that this is something that the public care about?

>> So yeah, first I had not I had never joined a protest in my life. uh and I ended up the first protest I joined was

actually a protest that I organized in France uh when I when I started the French branch of POI. So that was a

quite a steep learning curve for me. Um but why I think they are necessary? I think that protests are a great way to

raise attention on an issue. I tend to be more inclined towards very peaceful protests. I think they they carry more

strength. um than the more disruptive ones although um I'm open still to to uh to more ideas on that domain. Uh but

basically they have a huge uh value over cost ratio. Uh the first protest I organized in France, for example, we

were only 10 people uh in the street for one afternoon and we managed to get articles published in multiple

um multiple high-profile French newspapers thanks to this action. So that's that's really not a lot of work

to uh be heard. And you you mentioned that that protest was one of your first actions uh in PAI.

Um you've since um been the head of PAI in France um one one of our largest um national chapters. So what has your

experience been like as the head of PAI um France? Um and what was like some of the unique challenges in France? You

you've told me before that um some of the prominent figures in AI in France have been maybe a bit more skeptical

than they have been in in the UK and in America figures like Yan Lun. Um so how have you dealt with those challenges and

what's your experience been like? >> Yeah. So France was um let's call it hard mode. So yeah, you have to

understand that uh as you said France is the home of Yandun. Uh it's the home of Mistral uh which is probably the most

misaligned uh AI company in the world. Although I don't mean OpenAI is is fighting hard to to be on the first spot

at the moment. But um uh there is also a big problem of like understanding of AI because there is no there's no really

there's not much AI talent in France. uh everyone has left to go to San Francisco or London. So uh in house you have

mostly people who are advising the government who have basically no idea what they are talking about. Mostly uh

older people who are used to the old school AI which has nothing to do with uh deep neural networks and they still

feel like it's the same like it's just a series of instructions that a machine is following. Uh so that's that's a massive

problem. uh because of these people the government has no no clue what's happening or had at least no clue what's

happening when I was starting um but like despite this hostility and also part of the hostility is is the

the fact that the French government wants to push super hard on AI and and thinks that France should join the race

and and become a serious uh actor in developing uh in developing uh advanced AI In spite of this, we managed to in

just one year starting from zero. So at initially just myself trying to recruit a bunch of people uh we organized three

protests. We got coverage probably uh I don't have the number in mind but something like 30 uh articles in in

newspapers about us over the full year. We wrote a 100page counter expertise to the government's AI commission report.

Um, we managed to accumulate more than 1 million views on our media appearances. I became the uh the most well-known

public figure in France advocating for uh catastrophic risks of AI. Uh we organized a series of high-profile

conferences that culminating in a organizing a conference at the French Senate uh start like last month.

So I think that what really worked was professionalism like my strategy in France was we need

to grow across two metrics. The first one is visibility with just media appearances, protests etc. But the

second one which I think is is quite important is uh more like reputation. um being seen as credible actors who know

what they are talking about and getting access to um to basically politicians uh this way. So yeah, in a year we built a

community with about 2,000 people uh 100 paying members of the of the of POSAI France and dozens of volunteers working

in a in a very structured manner. So yeah, I really think that if we can shift the narrative in France uh which

was actively actually act actively trying to silence us, we can do it anywhere. Uh I think this is a good

proof of concept. >> Yeah, absolutely. I mean I' I've been very impressed by how well um you and

the rest of uh the people in PAI France have done. Um so what is both uh in terms of PAI in France and Pori

globally? Um we've been going for roughly two and a half years now. Um lots of things have changed um outside

in in the AI space and in uh regulation. Um so what what is your evaluation of PI's impact? What do you think we have

done well and what do you think there are where do you think there are areas for uh improvement? I'm going to give a

a fairly short answer to that one for now. But uh I think that what we've done well is gathering a large group of

people uh who understand what is happening and generally raising awareness. Uh where we have not done so

much is I think in terms of uh our way of organizing. I think that at the moment we are still

basically a group of people who are aware uh of what's happening and who want to to help.

Uh I think what we are going to do better from now on is organizing coordinating all of these people to

maximize their impact starting a ch a chain reaction. Um that's that's going to be my focus.

Mhm. >> And you said to me in a conversation, I believe this was um last week, that uh

solving the existential threat from AI is actually an easier challenge than it is to solve climate change. Um why why

is it that you think that? >> Yeah, I think it can sound counterintuitive u but

>> you have to think about the the mechanics of the problem. If you want to solve climate change and I'm not saying

that we're not going to be able to solve climate change but uh that means you that means that basically 8 billion

people need to change the way of living. You need to to change the way we use cars, change the way we eat like these

are as far as I know the two biggest uh uh factors. Um it it requires massive distributed sacrifice and you need to to

change the way the industry works uh as a whole in order to chop to stop AGI. You you don't need 8 billion people to

change. Actually you need very few people to change. You need to to stop a few maybe a few thousand people in two

countries basically. So it's a very concentrated supply chain. uh the talent is very concentrated the the capital

that flows is fairly concentrated still so we know we know where the data centers are we know who the CEOs are

that there are very few so it's all it's all bottlenecked and so because it is bottlenecked a small determined group of

people I think can push the dominoes that will end up applying enough pressure to squeeze that bottleneck

We don't need to convince the the the entire world. We we just need to convince the people who have the power

to shut it down. >> And so we already it seems like the polling shows that we pretty much

already have the support of the public. Um there are a larger and larger number of politicians it seems every month who

are beginning to take the extreme risks of advanced AI seriously. Um what do we need to do as PAI and as the the anti-

super intelligence movement um to accelerate the the process of politicians not only understanding it

but um being willing to to publicly state their their opposition to a reckless unregulated race to super

intelligence? What what are the steps that we need to take? >> Yeah. So I want to be transparent here.

Um, I'm at the moment when we are recording this, I'm not even the CEO yet. This is going to start next week.

So, I'm still in a in a listening phase. So, I'm interviewing everyone. Uh, I'm looking at the data. I'm trying to

really rock the situation at POSAI and the global situation. So, I'm not at the stage where I have a a 10-point policy

plan that I can give you today, but I would say at least I have a fairly clear vision for what we need to build.

Um, the key, I think, is to really build turn POSAI into a well a welloiled machine, right? We need to build a

structure that can handle scale. H we need to move from a group of activists to a proper organization with uh

stronger governance professional tools. Uh [snorts] we need to hire the right people and once we have these

foundations in place this is where we can really start a chain reaction. I think this is what needs to happen. uh

like a lot of people think of the situation as a race between different labs or a race between the US and China.

That's not reality. The reality is there is a race between people who are trying to build a godlike entity that will

basically most likely kill everyone and a race between us the people who are trying to coordinate to stop them. So

what we need to do is get this coordination to be on an exponential path as well. um start an exponential

where you have more and more people like the the rate of the people who are aware of the problem needs to grow

exponentially and the rate of people who are not only aware but then decide to take action join POAI or something

adjacent needs to also grow exponentially. I think that's the that's the vision. Obviously there there are

lots of other organizations other than Porzi who are involved uh in this space um all of whom are doing um different

things. So what is the niche of Porzi? What what are the unique actions that we can offer um that make our existence uh

important? >> Yeah well that's the thing most of the organizations that uh exist in this

space are very much playing the inside game. You have a lot of think tanks, you have a center for AI safety, you have

organizations like this. There are at the moment very few movements like POSAI who are trying to be very um very

popular uh to to uh enable everyone to take action. So that's our niche. That's our niche is

like going to any human and making them understand that they have a problem and it's the we have a shared responsibility

to address this problem and they can help. So the key is that we have to set up the systems so that anyone in the

world can join and start taking actions that go in the right direction and be part of the this uh this chain reaction.

>> What what are the um things that are coming up next for Porsai? So we have Porscon Brussels uh coming up um early

next year. we have the no longer the safety summit but the AI impact or AI action summit in in India. Um what what

are the things that are on the horizon that uh people in PI should know about? >> Yes. So postcon Brussels I think

February will be a very interesting month because we have a postcon in Brussels that is coming up. We now have

Stra Russell that is that is confirmed as a speaker. So that's going to be very interesting. we are uh reaching out to

other high-profile speakers and uh we will be in contact with the European Parliament. Uh I think that Europe has a

a big role to play in in what's going to unfold and around the same time there is the AI summit in India. uh and I think

that India is an interesting place because it's kind of like the bridge between

um the superpowers and the third world countries. So I think that there there are actions to be taken and we will

definitely organize some international action at this point. Uh before that uh I think starting now probably depending

on when we are releasing this video uh we are going to be launching our Christmas fundraising. So I I I

mentioned that we need to build the machine. Well machines need fuel and so um we we are currently hiring some key

roles to make the this movement professional and for this we need money. So, uh we will put forward profiles of

uh of some volunteers that have uh that have helped us uh grow so far. And uh I think with this fundraising, we're going

to try to um enable more volunteers to to help us using steepens and microrants.

Um yeah, I think if you're watching this and you've been waiting for a sign to get to get involved, I think this is it.

I think this is a good time uh now that we're going to have a new leadership. Uh the wind is at our backs and you can

join us now. Uh start being active or donate. That's a great time. Of course, links for everything Maxi mentioned

there will be available in the description and in the comments both to to join PSAI become an active

participant in in how this all plays out um and not be on the sidelines um and also if you want to uh help um fund

other volunteers um there'll be a link to the Christmas campaign as well. Um so thank you very much for joining us today

Maxim. Um looking forward to seeing what what we can do uh in Pauli going forward. Um, and yeah, thank you for

watching everyone. Make sure you join PZAI. Make sure you donate. >> Thank you. And thanks for thanks to

everyone who is watching and is aware of this issue. That's the first step.