Rising Titan – Mistral's Roadmap in Generative AI | Slush 2023

Summary

In a recent discussion at Slush, Paul Murphy, a partner at Light Speed, interviews Arthur, the co-founder of Mistral, a company focused on developing foundational AI models. Arthur explains that Mistral's vision is to create state-of-the-art models that are accessible to developers, allowing them to specialize and customize models for their specific tasks. They have already released a 7B model and have seen significant community engagement, with developers creating derivative works and integrating the model into various open-source projects. The conversation also touches on the importance of regulation in AI, with a focus on product safety and the need for empirical evidence in discussions about national security and existential risks. Arthur emphasizes the potential of AI to revolutionize sectors like healthcare and education and believes that fostering an open AI community is crucial for addressing global challenges like climate change. He stresses the significance of having strong European actors in AI to drive technological advancements and policy proposals that align with European values.

Takeaways

• 🤖 **Investment in AI**: Light Speed, a Silicon Valley based fund, has been investing in Europe since 2007 with a focus on AI, investing over a billion dollars into the category.
• 🚀 **Mistral's Vision**: Mistral, a startup co-founded by Arthur, aims to develop state-of-the-art AI models quickly and provide open access to developers for specialization.
• 🧠 **Open Source Models**: Mistral's approach is to create easy-to-use open source models that enable developers to tailor large language models for their specific applications.
• 💾 **Data and Training**: The success of Mistral's 7B model was attributed to a good team, a machine learning Ops system, and a focus on creating high-quality datasets.
• 🔍 **Community Engagement**: The release of the 7B model led to thousands of derivative works, showcasing new capabilities and integrations within the open source community.
• 🌐 **Upcoming Developments**: Mistral has plans for new models, techniques, and a platform offering hosting capacities with fast inference capabilities to be announced by year-end.
• 🧑‍💼 **Building a Company**: Arthur highlights hiring the best talent and community engagement as key challenges while building Mistral.
• 📈 **Regulation and Safety**: There's a call for hard regulation on the product side for safety and compliance, with a focus on empirical evidence over speculation.
• 🌱 **Open Science**: Open source principles have accelerated AI advancements, and Mistral aims to continue this tradition by fostering knowledge circulation.
• 🛡️ **Product Safety**: Arthur suggests that regulation should focus on the application layer, ensuring that deployed applications meet safety standards.
• ⚖️ **Independence in Regulation**: For effective regulation, there's a need for independent oversight, possibly state-funded, to prevent industry bias and pressure.
• 🌟 **Positive Impact**: AI has the potential to revolutionize sectors like healthcare and education, enabling more efficient and personalized services.
• 🌍 **European Leadership**: It's important for Europe to have strong AI actors to drive technological advancements and influence policy to reflect European values.

Q & A

What is the name of the company that Paul Murphy is a partner at?

Paul Murphy is a partner at Light Speed, a Silicon Valley based fund with investments in Europe since 2007.

How long has Light Speed been investing in Europe?

Light Speed has been investing in Europe since 2007.

What is the focus of the company Mistral, as described by Arthur?

Mistral focuses on developing state-of-the-art models quickly and aims to provide open access to these models for developers to specialize and make them their own, thus creating more human-like intelligent applications.

What was the first major milestone for Mistral after securing their seed round of funding?

The first major milestone for Mistral was to build their 7B model, which they achieved in less than three months.

How did Mistral manage to develop their 7B model so quickly?

Mistral managed to develop their 7B model quickly by having a good team, creating a machine learning Ops system, focusing on good training and inference code bases, and dedicating a large part of the team to curate and optimize datasets.

What has been the community's response to Mistral's 7B model?

The community has been very engaged with the 7B model, with thousands of derivative works where developers fine-tuned the model for their specific tasks or datasets, resulting in new capabilities and applications.

What are the next steps for Mistral?

Mistral is working on new models, techniques, and the beginning of a platform. They plan to offer hosting capacities for their models with fast inference capabilities and are expected to announce these developments before the end of the year.

What are the two main challenges that Arthur identifies for building the company?

The two main challenges Arthur identifies are hiring the best engineers and scientists in a competitive landscape and creating an engaged community around their open-source models.

How does Arthur view the concept of open source in the context of AI models?

Arthur differentiates between open source software and open weight models in AI. While providing the weights allows for modification, it doesn't necessarily enable full understanding due to the opacity of the models. He also mentions a balanced approach between openness and maintaining a competitive edge.

What is Arthur's perspective on the role of regulation in AI?

Arthur believes that regulation should focus on product safety, ensuring that AI applications meet certain safety standards. He also emphasizes the importance of empirical evidence in discussions around national security and existential risks related to AI.

How does Arthur envision AI improving society in the future?

Arthur sees AI as a tool that can revolutionize healthcare, education, and enable more creative thinking in society. He also believes AI can contribute to addressing existential risks like climate change by unlocking new scientific discoveries.

Why is it important for Europe to have a strong presence in the field of AI, according to Arthur?

Arthur believes it's crucial for Europe to have strong technological actors in AI to drive the field forward, shape the technology according to European values, and ensure Europe is not just a spectator as the technology transforms society.

Outlines

00:00

🎉 Introduction and Company Vision

Paul Murphy, a partner at Light Speed, introduces himself and the company's investment history in Europe since 2007, highlighting their broad sector and stage investment approach. He emphasizes Light Speed's significant investment in AI, mentioning a decade of experience and over a billion dollars invested in the category. The conversation shifts to Arthur, who shares the story of Mistal's founding six months prior, with a vision to innovate foundational AI models. Arthur outlines their strategy to create open-source models that are user-friendly for developers, allowing them to specialize and optimize the models for their tasks. The rapid development of their 7B model is attributed to a dedicated team and efficient machine learning operations system.

05:02

🚀 Community Engagement and Future Models

The discussion moves to the community's engagement with Mistal's 7B model, which has seen thousands of derivative works and integration into numerous open-source projects. Arthur shares the new capabilities enabled by the model, such as longer context understanding and better instruction following. He also teases upcoming announcements of new models and techniques, hinting at a platform that will offer hosting capacities with fast inference capabilities. The focus then shifts to the challenges of building a company, with hiring and community engagement being top priorities for Arthur.

10:03

🤔 Open Source Philosophy and Differentiation

Arthur clarifies the concept of 'open weight' in the context of AI models, distinguishing it from traditional open-source software. He explains that while the weights of the models are made accessible for modification, full transparency is not always possible due to the models' complexity. The open weight approach is shown to be beneficial for bias control, interpretability, and security through red teaming. Arthur also discusses the business and ideological advantages of open-source, emphasizing the importance of knowledge circulation in accelerating AI advancements. He differentiates Mistal's approach from competitors by targeting developers and focusing on specialized, efficient models that can be customized using proprietary data.

15:05

🌐 Regulation and Safety in AI

The conversation delves into the topic of AI regulation, with a focus on product safety, national security, and existential risk. Arthur advocates for a balanced approach, emphasizing the need for empirical evidence to guide discussions and regulations. He suggests that the application layer should bear the responsibility for safety, with model providers offering controllable models and evaluation tools. Arthur also calls for independent regulatory bodies, possibly state-funded, to monitor AI safety without being influenced by industry pressures.

20:07

🌟 Positive Impacts and Utopian Vision of AI

Arthur envisions a utopian future where AI significantly improves various sectors, such as healthcare and education, by providing personalized assistance and enabling more creative thinking. He also sees AI as a potential tool to address climate change by accelerating scientific research and innovation in fields like chemistry and material science. Arthur stresses the importance of fostering an open AI community to drive advancements and overcome global challenges.

25:09

🇪🇺 The Importance of a European AI Champion

In the final segment, Arthur highlights the importance of establishing a European presence in the AI field. He argues that having a European champion is crucial for shaping technology according to European values and ensuring that the continent is not just a spectator as the AI wave progresses. Arthur sees the development of strong European technological actors as essential for driving policy and technological proposals, influencing the direction of AI globally.

Keywords

💡AI Investment

AI Investment refers to the financial backing provided to companies or projects that are focused on developing or utilizing artificial intelligence technologies. In the video, Light Speed, a Silicon Valley-based fund, has been investing in AI for nearly a decade with over a billion dollars invested in the category, showcasing the significance of AI in modern technology investments.

💡Mistral

Mistral is the name of the company being discussed in the video, which is focused on creating foundational models for AI differently from other companies. The company aims to develop state-of-the-art models quickly and provide more open access to developers, which is a key theme in the discussion about the future of AI development and accessibility.

💡Open Source Models

Open Source Models in the context of AI refer to the models whose underlying code or structure is made publicly available, allowing others to view, modify, and distribute the models. Arthur from Mistral emphasizes the importance of open source in AI, stating that it allows for better customization, differentiation, and community engagement, which is crucial for the evolution and adoption of AI technologies.

💡7B Model

The 7B Model mentioned in the video is a large-scale AI model developed by Mistral. The '7B' likely refers to the number of parameters the model has, indicating its complexity and capacity for learning. The development and release of this model faster than expected is a testament to the company's commitment to rapid innovation in the AI field.

💡Developer Access

Developer Access denotes the level of interaction and modification rights given to developers over AI models. Mistral's approach aims to provide developers with deep access to models, allowing them to specialize and tailor models to their specific needs. This is a core part of Mistral's philosophy to democratize AI and let developers create more efficient and differentiated applications.

💡AI Safety

AI Safety involves the practices and regulations aimed at ensuring that AI systems function correctly, ethically, and without causing harm. In the video, the discussion touches on the importance of AI safety and the need for regulation, reflecting the broader conversation about the ethical considerations and potential risks associated with advanced AI systems.

💡Product Safety

Product Safety in the context of AI refers to the reliability and safety of AI applications when they are deployed in real-world scenarios. The video emphasizes the need for AI applications to meet certain safety standards to build trust and ensure they perform as expected without causing harm or errors.

💡National Security

National Security, as discussed in the video, is a concern related to AI that involves the potential for AI systems to disseminate knowledge that could be misused by malicious actors. It's a part of the broader regulatory debate on AI, where the speakers advocate for empirical evidence and careful monitoring without stifling innovation.

💡Existential Risk

Existential Risk in the context of AI pertains to the theoretical possibility that advanced AI could become uncontrollable and pose a threat to human existence. The video speakers view this as a currently remote and speculative concern, advocating for a focus on more immediate and tangible issues like product safety and ethical AI development.

💡European AI Champion

A European AI Champion refers to a leading AI company or entity based in Europe that can influence and drive the direction of AI technology development according to European values and interests. The video highlights the importance of having such a champion to ensure Europe's active role in shaping the future of AI globally.

💡Regulation Pressure

Regulation Pressure is the influence that regulatory policies and standards have on the development and deployment of AI technologies. The discussion in the video suggests that applying regulation pressure at the application layer rather than the foundational model layer can foster innovation, competition, and safety in the AI industry.

Highlights

Paul Murphy, a partner at Light Speed, discusses the firm's extensive investment in AI, totaling over a billion dollars.

Mistal, a company founded by Arthur and his team, aims to create open-source models that are accessible to developers for specialization.

Mistal's vision is to democratize access to large language models, allowing developers to create more personalized and efficient applications.

The company has rapidly developed a 7B model, showcasing their ability to innovate quickly in the AI space.

Mistal's 7B model has been well-received by the community, leading to thousands of derivative works and specialized applications.

Arthur emphasizes the importance of a dedicated team and good data sets in achieving their rapid development of AI models.

The company is focused on building a platform that offers hosting capacities with fast inference capabilities for their models.

Hiring the best talent is a significant challenge for Mistal, as they aim to stay competitive in the European tech landscape.

Mistal is proactively engaging with policy matters, advocating for hard regulation on the product side to ensure compliance for application makers.

Arthur differentiates Mistal's approach from competitors like OpenAI by targeting developers and enabling them to build specialized, efficient applications.

Open weight models are crucial for Mistal's strategy, allowing for modification and customization to align with developers' values and use cases.

Mistal sees open science as a key driver in the rapid advancements of AI and aims to contribute to this tradition.

The company is committed to addressing biases and ensuring model interpretability through their open weight approach.

Arthur discusses the need for regulation in AI, particularly focusing on product safety and the importance of empirical evidence in policy discussions.

Mistal believes that regulation should primarily target the application layer, fostering competition and promoting safer AI technologies.

The potential for AI to improve sectors like healthcare, education, and address global challenges like climate change is highlighted as a utopian future.

Having a strong European presence in AI is seen as critical for shaping technology according to European values and ensuring a leading role in AI innovation.

Transcripts

[Music]

[Music]

[Music]

okay um welcome everyone really nice to

see you uh very very happy to be back at

slush especially this time with Arthur

uh my name is Paul Murphy I'm a partner

at light speed based in London uh just a

real quick uh bit about light speed for

those that don't know uh we have

actually we're a Silicon Valley based

fund but we've been investing Europe uh

since 2007 we have over 30 companies now

um in

Europe uh and um yeah we're investing in

pretty much every sector uh and every

stage um we're talking about AI today

and I think I think it's important to

put some context around that from our

perspective we actually have been

investing in AI for nearly a decade we

have about 50 companies um and have

invested over a billion dollars into the

category and that context is relevant

because uh when we met Arthur and his

co-founders we thank you we immediately

fell in love with the vision of mistol

um and so I thought the the best place

to to start would be to ask you Arthur

to tell us a little bit about what

you're building at

sure thank you very much slush for the

invitation thank you Paul as well um so

yeah we started mistal six months ago uh

with guom and timot and our vision was

that we wanted to make the foundational

models a bit differently from the other

companies uh we've been in the field for

almost a decade now and we've seen it go

from a cat and dog detector to something

is very close to being humanlike

intelligent or atast at least looks like

it and we knew that with a very

dedicated team we could develop uh

state-of-the-art models very very

quickly and we could actually take the

field into something that is that would

be more open where would give more

access to developers so that they could

specialize the models make them their

own make them as small as possible to

solve their task and for us the good way

of doing it and the good way of starting

that was to ship the best open source

models create models that would be very

easy to to use by individual developers

and from then on build onto an

Enterprise play to sell a platform that

allows developers to take large language

models and to make them their own to to

create some differentiation on the

application they're making and that's an

differentiation which is currently hard

to do when you only access apis of a

couple of providers but if you have a

deep access to the models you can create

things that are much more interesting

and this is what what we want to enable

so when we we LED your seed round it

wasn't that long ago you told us that

you're first thing you're were going to

do is to build your 7B model and then I

think it was it was like 3 months from

when we signed the docs on that round uh

we got our message saying hey we're

ready uh it's ready and it was faster

than we had expected it was already

incredibly ambitious I'm just I think

everyone's probably wondering how you

did how you did that so quickly well I

think the secret is to have a a good

team uh so we were joined by our first

employees uh a dozens of them at the

beginning of June and nobody took

holidays uh we Rec created the wall what

we call the machine learning Ops system

so that's actually Fair simple you you

need to create a very good Training code

base you need to create a very good

inference uh code base uh to to deploy

the models you need to be able to

evaluate the models and the one thing

you do need the most and where we

actually dedicated 80% of the team on uh

for three months is to have some very

good data sets so we we went to the open

web took public domain knowledge created

it so that we could just get the best of

it filtered it did everything to get

something very good did some work around

how to better optimize the models and

combine all of this uh and then train

the model to get the 7B and we continue

doing it uh with the new models we'll be

soon announcing like when you say it's

fairly easy I think maybe some people

would disagree with you on that but you

definitely made it look easy I think

that's true um so I'm curious uh the

community you know was has been very

engaged with 7B model Since You released

it I think it was you know trending on

hugging face for multiple days top you

know top top models um what kinds of

things have you seen that have been

interesting so far from the community so

we've seen I think thousands of um D

derivative work so uh Developers that

took mistal 7B and fine tuned it on

their task or on their data sets to make

it special so we've seen new

capabilities like longer context uh

better instruction following capacities

uh we've seen uh like new topics so

we've seen like occult specialized

models able to talk about uh post test

experience and the like much better than

what MB was able to do before so many

kind of different applications uh some

of them useful some of them

just funny um we've seen integration in

a lot of llm Open Source projects so the

open source world around Genera T is

pretty is is pretty involved already so

you have retrieval augmentation systems

you have projects that allow to deploy

the models on your laptop you have all

of these things and they adopted M 7B

very quickly and I think it was the

field was really missing an actor that

would produce the best open source

models and actively engage with the

community and that's what we we uh we we

are enabling okay and so now 7B is out

there what comes next so we have um

nothing announced yet but we we do have

things in house that we'll be announcing

before the end of the year uh new models

uh new techniques uh and obviously the

beginning of a platform so we're

actively working on the product uh we'll

be soon offering uh hosting capacities

for our models uh with very fast uh

influence uh capabilities and yeah

that's for uh very soon okay okay I'll

watch the space um so you're also while

you're doing all this incredibly what I

think most people would think of as

quite challenging technical work you're

also building a company and I know

that's not easy haven't done it myself

before um what's keeping you up at night

right now what's your biggest headache

um so hiring is obviously a very big

challenge I think the only reason why we

got there so fast is because we hired

the the best engineers and the best

scientist in the world it's a very

competitive landscape uh Europe is full

of talent especially the junior ones uh

and so we we are this is some like a

very big preoccupation for us like I'm

constantly working on it so that's one

thing um the other thing is like

creating the community engaging with it

uh so we started with the with mral 7B

but we really need to uh yeah well

facilitate the life of our users uh have

them engage facilitate Upstream

contribution facilitate the emergence of

IDs that we could help enable

so that's another thing we have a lot of

um I guess policy matters uh that we did

not expect but obviously this is an

agenda that you don't select um there's

we we so there's there's different

tracks you have in the US you have in EU

um we've been uh vocal about the fact

that we wanted to have hard regulation

on the product side because it's very

important and we see ourselves as the

provider of tools and a big enabler of

compliance for the application makers so

we've been saying that uh constantly and

and and we've seen like the debate uh

progress on these topics and so this is

something that yeah we're very keen on

trying to enable from a technical

perspective because it's important that

you have technical Founders that

participate in that discussion uh and so

that that has kept me up at night uh for

for a while and I think you know the

ambition was certainly to be able to

build something that could rival other

large companies like open Ai and I'm

just curious what do you view as a

differentiating philosophy or approach

to companies like open AI I think a

differentiating philosophy is that we

really Target the developer space and we

really think that when you're making an

application that you want to put into

production you do want to have several

specialized models that are as many

chips you you should see them as chips

that you assemble in an application and

it's actually not easy to make a very

good chip for the use case you want so

you can start with a very big model with

thousands of billion well with hundreds

of billions of parameters it's going to

solve your task maybe but you could

actually have something which is 100

times smaller and when you make an a a

production application that goes at

scale and Target a lot of users you want

to make the choices that lower the

latency lower the costs uh and leverage

the actual proprietary data that you may

have and this is something that I think

that that's not the the topic of our

competitors they're really targeting

like multi-usage very large models AGI

we takeing very much much more pragmatic

approach in enabling super useful

application today uh that would be cost

efficient that would be very low latency

and that would enable strong

differentiation through uh proprietary

data okay and you've talked I think

another key difference you've talked a

lot about open source as being a core

part of your DNA um and I think question

I sort of wanted to ask uh Arthur by the

way wouldn't look at these questions

beforehand so he wasn't expecting this

one but I understand the concept of Open

Source software I think we all do we see

the code you kind of can take it modify

it um and use it but in the world of AI

and and models the concept of Open

Source just feels like it's maybe a bit

different because actually some things

you do keep for yourself or you have to

what does open source mean in the

context of llms and AI so we don't

really call them open source so the the

models we provide are open weight I

think it's important to like keep a good

distinction between the like the

terminology we were using for software

and the terminology we are using for

models if you provide the weights of a

models you're enabling modification

you're not necessarily enabling like

full understanding of what's going on

but even if you do provide full

transparency on the data sets and

training you don't know what's going on

cuz it's it's a bit opaque by Design so

it's an empirical science when you

create a model the only way to verify

that the model is doing what you expect

is to measure it with with with

evaluation this something will be

enabling and then it's to modify it with

some signal coming from either humans or

maybe machines to to modify the model so

really the modification part is super

important for differentiation and we're

taking this approach there's a full open

source approach which I think is very

valid as well for science in which you

disclose your data set you disclose

everything that I think that's that's

something that we would strive toward at

some point but obviously it's super

competitive and the data set part is

very hard to to obtain it's also very

Capital intensive you need a lot of gpus

so right now we're taking a balanced

approach in between what we uh opens

what the open ways we provide the things

we keep for ourselves to to get a

competitive uh Edge and this is going to

be a dynamic play and we expect it to to

evolve with time and with technology

okay and then does the does the open

weight approach help with other

challenges like biases and control yeah

so it helps with basically two things

the first thing is that you can modify

the the biases you can have like a

strong and fine uh modification

capabilities on the editorial tone on

the orientation and alignment of the

model so we allow alignment of your own

models to your own values and those can

slightly differs um so like fine control

of biases goes through fine deep access

to to models that's the first thing the

second thing it allows and we've seen it

with active engagement of the AI safety

community in particular around open

Source models it allows to have better

interpretability because you can see the

inner activations of the of the models

and and that tells you things about

what's happening uh about why the model

is taking a decision and not another so

why is it outputting award and not

another and so in the interpretability

world it's also super useful it's also

and I guess the last thing is that it's

very useful to do red teaming because

you have a deep access to to the model

and so you can try to verify the the

part of it which are a bit failing or

behaving unexpectedly and these are

things that you can then correct very

similarly to uh what we've been doing in

the open source software for security

cyber security and the like okay and

then what I mean what is sort of what do

you view as at stake here you know why

is this is this in other words is this a

business Advantage for Mel or is it

something more fundamental that you see

as almost a

responsibility so it's both a business

Advantage because we allow further

customization and differentiations and

it's a very mature market and we expect

that on the application space the one

actors the actors that are going to

survive and create some value are the

one that will be able to strongly

differentiate themselves and so they

would need deep access to models so

that's a business differentiator then

there's a bit of an ideological

differentiators in the sense that I've

been contributing to open source for 10

years G as well we really think that AI

has been accelerated by open science by

the circulation of knowledge and that's

how we went in 10 years from something

very very uh well interesting but that

would just detect Tech boats and

something that actually uh will speak

the human language so this has been

allowed because you had big tech labs

you had the Academia as well that was

all of them were communicating at

conferences every every year and and

information would circulate and that

accelerated things and suddenly in 2020

open I decided to stop publishing and it

was followed by its competitors uh very

closely after and so ever since 2022 we

haven't seen like major advances in llm

publicly announced and so we've seen

currently there's like new architectures

that are used internally by our

competitors and that are not available

out there this is something we will

correct very soon okay great um so I

want to shift Focus now talk about

something you mentioned earlier which is

regulation and it's a topic you kind of

can't avoid I think you've thinking

about AI um a lot of focus within Europe

and and in the UK um and I think you at

the safety Summit in the the AI safety

Summit um last month there's a lot of

ideas out there and I think um you know

curious to hear your view is to what

should be the priority how should

regulation be prioritized and

instrumented yeah so I think it's quite

interest it's a very interesting topic

for me and and we've been uh yeah we've

been contributing IDs the one thing that

I would start with is that we've been

talking about regulation and safety and

mixing Concepts very heavily so there's

a matter of product safety which is

answering the question of you deploy a

diagnosis assistant in the hospital you

want it to be safe you want to be able

to measure whether the decision it's

making is actually sound is actually

correct so that's that's what we call

product safety that's something you have

when you buy a car you have product

safety of your car and it should very

much be similar for applications that's

one thing and AI to some extent creates

new problems because you have models

that are not deterministic and so they

behave in a potentially an expected way

so it's useful to refine the hard lws

that we have around uh product safety

regulation now there's another topic

that came up which is National Security

so the question of whether the llms that

we're training the LM that everyone is

training is spreading too much knowledge

so when you have access to llm you're

effectively able to educate yourself on

many topics and this is something that

is a concern for different actors

because you could have like small groups

that are deemed bad that could use this

knowledge to do bad things so this is

this has been at the a central topic

especially in the US um we're still

lacking a lot of there's absolutely no

public evidence that llms are

facilitating anything so we're really we

we've been advocating for for some

empirical grounding of the discussion

and this is something that's currently

very much lacking and then there's a

third thing which is kind of mixed with

all this with with the two first which

is existential risk so knowing whether

the technology we're making is

effectively on an unbounded exponential

that will end up destroying us because

as every exponential it kind of breaks

the limits at some point and and that's

well it becomes IL defined as we say in

mathematics so this is something that

for us is very much science fiction

that's empirical evidences so what we've

been saying is that we should really

focus on the first topic which is

imminent is something that is we do need

to have product safety on AI because

it's it's going to to otherwise it's

going to break trust in the technology

we're making and so we want to enable

that on the second part we are lacking

empirical evidence but I think this is

something that we should monitor closely

knowledge historically knowledge the

spreading of knowledge has always had

more benefits than uh than than

drawbacks and we AI is not different in

that respect but still it's something

that that could do with monitoring

because it's really new technology on

the third aspect of AGI and and and the

like and and the fact that that you

could have an autonomous system that

would go out of control this is

something that we are not at heas

discussing because we really think that

as scientists we are lacking evidence of

any existential risk and we think that

it pollutes the discussion on the first

aspect which is super important yeah and

so if I just kind of make sure I

understand this right the view is that

the application layer is probably the

one that has the most responsibility in

terms of safety at least to consumers or

end users whoever that is businesses but

that perhaps the models could provide

that as a feature or functionality but

it's not the responsibility of the model

to ensure that the ultimate data

transmitted is itself safe exactly so we

think that the correct way of putting

some pressure on the model providers uh

like us is to effectively say that any

application which is deployed and that

includes the application that we deploy

uh should be should meet a certain

number of safety standards so they

should do what they're expected to do

and if you do that then that means that

the application providers will be

looking at model providers that are

controllable enough that can give some

form of guarantees that can give some

evaluation Tools around the fact that

they're controllable and that they do

what they're expected to do so you have

some form of second order pressure that

is put you put pressure on the

application layer and that puts a market

pressure on the foundational model

developers and that's the correct way of

making a healthy competition in making

the most controllable models in making

the best evaluation tools and making the

best Guard railing tools and we think

that it's a much better way of doing it

than applying directly a pressure on the

foundational model layer because if you

do that well you're you're in a IL

defined territory because you're trying

to control something which is by Design

super multi-purpose very akin to a

programming language so you can't really

regulate the programming language

because you can do anything with it and

so really there's a problem of

definition and then there's an

operational problem of the fact that if

you put some heavy pressure on that

layer you're effectively um favoring the

big actors that have a lot of compliance

capabilities and you're you're making it

harder for startups with Innovative IDs

to to come up and compete and

so this like foundational models is a

bad proxy for a market capture and so we

believe that applying the regulation

pressure on the application layer is the

one thing to do because that's going to

Foster competition and provide a safer

world do you think that there's a role

for an or you know an iaea kind of like

organization to exist that helps to

enforce or provide this guidance

regulation so yes I think um this kind

of Regulation if we need

to monitor I think we we do need to have

empirical evidence of what's happening

in the space and we need to monitor the

product side safety and one way of doing

it is to enforce that we have very very

independent uh organisms that actually

monitor these things and when I say

independent I mean that we should be

very cautious of of preventing pressure

and Regulatory capture of this things so

setting standards but ensuring that no

big actor is basically writing the

standard themselves so what that means

that if we are if we if we need to have

this this form of organisms they need to

be very well funded probably state

funded

and being completely screen from

pressure from the industry okay so now I

want to shift you know I think the

regulation debate is largely many of the

debates in AI are tend to sort of skew

somewhat negative so let's dream for a

second like how can AI make our lives

better what do you see as the utopian

future with

AI so I think the there's so there's

many vertical in which Ai and like

interact ING with machines with natural

language carry a lot of value uh so

Healthcare is going to be completely

changed by AI because you you will be

able to

interact uh with empathic beings uh that

are actually super well grounded on

statistics and that's really what you

were expecting from medicine so we

expect that AI is going to empower uh

Physicians to be much better at what

they're doing and to make better

decisions um education is also a super

interesting topic uh personalization of

Education we know that it's super

important to uh take the most most

potential of of human beings and having

some like your individual teacher being

an assistant this is going to change a

lot of things especially in the global

South um so that's two things generally

speaking this is going to change the way

we work so it's a way the fact that it

can interact with v structure knowledge

and that it can do well act as if well a

bit imita the boring task of of your

daily life this is going to enable more

space for creative thinking so we will

be able to think more creatively and

that's going to unleash I think a new

Society very soon and if you think about

some of the more existential risks we

face in the world like climate change do

you think that something like that can

be addressed or at least improved yes so

I think this is a frontier which which

hasn't been completely addressed yet but

this is really a promise of having

better models the fact that if you if

you have some ways of reasoning around a

pool of science well you can enable

scientists to come up with new ideas you

can potentially unlock very precise

things like create like in chemistry in

accelerating uh chemical reaction so

that you emit less CO2 for instance

these things like Material Science

chemistry Fus nuclear fusion as well all

of these locks that we have and that are

that we basically need to break in order

to address climate change well I mean

that's one of the way you can can

address climate change obviously the one

way is also to reduce consumption but

the the these things we we think that AI

is going to be an enabler of of of

breaking these lcks it's not going to be

an easy task uh there's still many

things to invent and we think that going

through the open science part uh

fostering the the the keeping fostering

the AI community that drove the field

forward for 10 years is is super

important to break these logs okay

that's great so I think I want to come

back to to Europe sort of for our last

questions we're we're out of time um how

I mean I think the fact that the company

is being built in Europe is very

important to you it was obvious to you

and your co-founders when we invested um

how important do you think it is for the

industry that we have a European

champion emerg in in the field of

AI so the Technologies is AI generative

AI is is really a wave you can it's

going to change society quite

significantly and in Europe we have a

choice of either being on top of the

wave and driving the technology forward

or just looking at it happening in the

US uh and in China and we think that in

order to shape the technology to our

values and to the way we think about

democracy about Society we need to have

very strong technological actors that

are able to drive the field forward make

proposals um both in term of policy and

in term of technology and so that's why

we believe it's super important that

actors

that we have strong actors in Europe

great thank you so much Arthur really

appreciate it

amazing