Aerospace & AI - Airbus’ responsible and human-centric approach to AI | #aidatasummit24

Summary

Der Vortrag diskutiert die Verwendung von KI bei Airbus, einem führenden Hersteller von Flugzeugen und Verteidigungssystemen. Es wird auf die Anwendung von KI in täglichem Geschäftsbetrieb, operativer Unterstützung und game changing Anwendungen eingegangen, wie z.B. Flugzeugdesign. Es wird betont, dass KI nicht nur Compliance, sondern auch Verantwortungsbewusstsein und Werte der Firma fördert. Die Herausforderungen der KI-Regulierung und der Ansatz zur Umsetzung von KI im Einklang mit den Firmenwerten werden ebenso thematisiert wie die Bedeutung des menschzentrierten Ansatzes und der ethischen Gestaltung von KI-Systemen.

Takeaways

• 🚀 Airbus ist bekannt für die Herstellung von Flugzeugen, Verteidigungssystemen, Satelliten und ist der größte Hubschraubenerzeuger der Welt.
• 🤖 Sie setzen künstliche Intelligenz (KI) in verschiedenen Bereichen wie täglichem Geschäftsbetrieb, operativer Unterstützung und gamechanging KI-Anwendungen ein.
• 🔍 Generative KI wird besonders in der Informationssuche und -verarbeitung in technischen und dokumentationsintensiven Prozessen genutzt.
• 🛠 KI wird zur Verbesserung der Kernoperationen eingesetzt, um Effizienz zu steigern, Planungen zu verbessern und Muster zu verstehen.
• 🛩️ KI hat auch das Potenzial, den Flugzeugentwurf und die Produktion von Flugzeugen grundlegend zu verändern, indem sie digitale Zwillinge erstellt und Prozesse automatisiert.
• 🌐 KI wird auch in der Kundeneffizienz unterstützt, beispielsweise durch die Vorhersage von Wartungsbedarf und die Verbesserung des Kundenerlebnisses.
• 🤝 KI-Anwendungen in verschiedenen Abteilungen von Airbus umfassen Satellitenbildanalyse, automatisierte Landungen und Taxifahrten sowie KI-gestützte Testhubschrauber.
• 🏭 Auf der Produktionsebene wird KI bereits genutzt, um Arbeit zu priorisieren, Qualitätsmerkmale zu analysieren und die Produktion effizienter zu gestalten.
• 📚 KI wird auch in der Ingenieurarbeit genutzt, um den Entwurf von Flugzeugen zu verbessern und digitale Zwillinge zu erstellen.
• 📖 Die KI-Akt-Richtlinie von Airbus betont die Verantwortung für KI-Systeme, Transparenz, Datenschutz, Fairness und menschliche Kontrolle und Überwachung.
• 🌟 Eine humanzentrierte Herangehensweise und die Einhaltung von Regelungen und gesellschaftlichen Werten sind zentrale Aspekte der KI-Strategie von Airbus.

Q & A

Wie viele Airbus-Flugzeuge sind täglich im Einsatz?

Rund 14.000 Airbus-Flugzeuge sind täglich im Einsatz.

In welchen Bereichen ist Airbus tätig?

Airbus ist im Bereich der Luftfahrt, Verteidigung, Raumfahrt und ist auch der größte Hubschrauberhersteller der Welt tätig.

Was versteht Airbus unter verantwortungsvoller Nutzung von KI?

Airbus versteht unter verantwortungsvollem Umgang mit KI nicht nur die Einhaltung von Mindestanforderungen, sondern auch die Integration von KI in den Unternehmenswert und eine verantwortungsvolle, ganzheitliche Herangehensweise.

Inwiefern wird KI bei Airbus zur Verbesserung des täglichen Geschäfts eingesetzt?

KI wird bei Airbus eingesetzt, um den täglichen Geschäftsablauf zu verbessern, z. B. durch die Nutzung generativer KI-Funktionen, um Informationen schneller zu finden und Prozesse effizienter zu gestalten.

Wie wird KI bei Airbus in der Produktion und im Betrieb unterstützt?

KI unterstützt bei Airbus die Kernoperationen durch maschinelles Lernen, Computervision und Optimierung, um Prozesse effizienter zu gestalten und Planungs- und Vorhersageprozesse zu verbessern.

Welche potenziellen Veränderungen durch KI bei der Flugzeugentwicklung werden genannt?

Generierte KI-Modelle wie Surrogate Modeling haben ein großes Potenzial in der Flugzeugentwicklung, insbesondere bei der Erstellung von digitalen Zwillingen und der Gestaltung von Flugzeugplattformen.

Wie plant Airbus die Nutzung von KI für die Kundenzufriedenheit?

Airbus plant die Nutzung von KI für die Kundenzufriedenheit durch Predictive Maintenance und die Bereitstellung von maßgeschneiderten Dienstleistungen, um die Kundenerfahrung zu verbessern.

Welche KI-Anwendungen werden bei Airbus in der Produktion und im Engineering genutzt?

Bei Airbus werden KI-Anwendungen wie Computervision für automatisierte Landungen und Taxifahrten sowie KI zur Unterstützung von Piloten in stressigen Situationen in der Produktion und im Engineering genutzt.

Wie schätzt Airbus die Herausforderungen der KI-Regulierung?

Airbus sieht KI-Regulierung als Herausforderung, die man sorgfältig und in enger Zusammenarbeit mit Regulierungsbehörden wie der EASA anzugehen hat, um sicherzustellen, dass AI-Anwendungen sicher und verantwortungsvoll eingesetzt werden.

Wie plant Airbus die Einhaltung der KI-Akt-Richtlinien?

Airbus plant, die KI-Akt-Richtlinien einzuhalten, indem sie eine zentrale AI-Expertise hat, die Projekte leitet, aber auch die Organisation dezentralisiert und Experimente in einem einwandfreien Weg ermöglicht.

Was ist die Rolle der zentralen AI-Gruppe bei Airbus?

Die zentrale AI-Gruppe bei Airbus ist für den Zugang zu cloud-bereiten Plattformen verantwortlich, die für die Produktion vorgesehen sind, aber auch für die Schaffung von sicheren Sandbox-Umgebungen, in denen Experimente stattfinden können.

Outlines

00:00

🚀 Einleitung in die AI-Anwendungen bei Airbus

Der Sprecher stellt sich und Airbus vor, wobei Airbus bekannt ist für die Herstellung von Flugzeugen, Verteidigungssystemen und Satelliten. Es wird erwähnt, dass Airbus auch der größte Hubschraubenerzeuger der Welt ist. Der Schwerpunkt liegt auf der Verantwortungsbewussten Nutzung von KI, die sowohl den gesetzlichen Rahmenbedingungen als auch den Unternehmenswerten entspricht. Es wird auf die verschiedenen Anwendungsbereiche von KI eingegangen, wie der täglichen Geschäftsprozesse, der operativen Unterstützung und der generativen KI für die Informationssuche. Außerdem werden potenzielle gamechanging KI-Anwendungen in der工程设计 und im Bereich der Kundeneffizienz diskutiert.

05:03

🤖 KI in der Flugzeugproduktion und -operation

Der zweite Absatz konzentriert sich auf die Anwendung von KI in der Flugzeugproduktion und -operation. Es wird über die Verwendung von KI in der automatisierten Landung und Steuerung von Flugzeugen und Hubschraubern gesprochen, die die Sicherheit erhöhen sollen. Ebenso wird die KI im Bereich der Produktionsplanung und -überwachung thematisiert, wobei die KI dazu beiträgt, Prozesse effizienter zu gestalten und Muster zu verstehen. Die Herausforderungen im Bereich der KI-Regulierung werden angesprochen, insbesondere im Hinblick auf die verschiedenen Sensibilitätsbereiche innerhalb des Unternehmens.

10:06

🌐 KI-Regulierung und menschzentrierte Ansätze

In diesem Absatz wird die Rolle der KI-Regulierung und der menschzentrierten Ansätze bei der Entwicklung und dem Einsatz von KI in Airbus besprochen. Es wird erwähnt, dass die KI-Aktivitäten im Unternehmen registriert, eine Risikoklassifizierung vorgenommen und die Governance im Laufe der Zeit gestärkt wird. Es wird auf die Notwendigkeit hingewiesen, frühzeitig Exploration zu ermöglichen und gleichzeitig sicherzustellen, dass dies im Einklang mit den gesetzlichen Rahmenbedingungen und den Unternehmenswerten geschieht. Die KI-Aktivitäten sollen demnach in einem verantwortungsvollen und koordinierten Umfeld stattfinden.

15:07

🛠️ KI-Prinzipien und Anpassungsfähige Governance

Der vierte Absatz beschäftigt sich mit den KI-Prinzipien von Airbus und der Anpassungsfähigen Governance. Es wird betont, dass die KI-Prinzipien auf europäischen Werten und den Unternehmenswerten beruhen und wie diese in den企业文化 und die KI-Implementierung einfließen. Es wird auf die Bedeutung von Sicherheit, Nachhaltigkeit, Verantwortlichkeit, Transparenz, Datenschutz, Fairness und menschlicher Agency in den KI-Systemen hingewiesen. Schließlich wird die Bedeutung der fruchtbaren Zusammenarbeit und des Austauschs mit anderen Unternehmen und Behörden im Hinblick auf die zukünftige Entwicklung von KI und deren Regulierung hervorgehoben.

Keywords

💡AI

Künstliche Intelligenz (AI) ist eine Technologie, die es Computern ermöglicht, Aufgaben zu erlernen und auszuführen, die normalerweise nur von Menschen ausgeführt werden können. Im Video wird die Anwendung von AI in verschiedenen Bereichen der Luft- und Raumfahrtindustrie diskutiert, wie zum Beispiel in der Verbesserung des täglichen Geschäfts, der operativen Unterstützung und der generativen KI.

💡generative AI

Generative AI ist eine Form der künstlichen Intelligenz, die neue, originale Inhalte erzeugen kann, anstatt nur vorhandene zu analysieren. Im Kontext des Videos wird sie für die Verbesserung der täglichen Arbeitsaufgaben genutzt, wie zum Beispiel die Suche nach relevanten Informationen in technischen Dokumenten und Prozessen.

💡Regulierung

Regulierung bezieht sich auf die Vorschriften und Gesetze, die von Regulierungsbehörden entwickelt werden, um bestimmte Industrien oder Aktivitäten zu überwachen und zu kontrollieren. Im Video wird die Herausforderung der Einhaltung von Regulierungen in verschiedenen Bereichen der Luftfahrtindustrie hervorgehoben, wie zum Beispiel bei der Qualitätsprüfung von Flugzeugen.

💡Digitale Zwillinge

Digitale Zwillinge sind virtuelle Repliken physischer Objekte oder Systeme, die es ermöglichen, diese in der virtuellen Welt zu analysieren und zu optimieren. Im Video wird die Verwendung von digitalen Zwillingen in der Flugzeugentwicklung diskutiert, um die Effizienz und Sicherheit zu erhöhen.

💡Prädiktive Wartung

Prädiktive Wartung ist ein Verfahren, bei dem KI und maschinelles Lernen verwendet werden, um vorherzusagen, wann ein Teil oder eine Maschine instand gesetzt oder ersetzt werden muss. Im Video wird dies als Anwendung der KI in der Luftfahrtindustrie erwähnt, um den Kundensupport und die Kundenerfahrung zu verbessern.

💡Ethik durch Entwurf

Ethik durch Entwurf ist ein Ansatz, bei dem ethische Überlegungen von Anfang an in den Entwicklungsprozess von Technologie integriert werden. Im Video wird dies als Bestandteil der AI-Strategie der Firma beschrieben, um sicherzustellen, dass die entwickelten Systeme den ethischen Werten und Prinzipien der Gesellschaft entsprechen.

💡KI-Prinzipien

KI-Prinzipien sind Leitlinien oder Richtlinien, die eine Organisation entwickelt, um die Verantwortungsvolle Nutzung von künstlicher Intelligenz zu gewährleisten. Im Video wird die Entwicklung von KI-Prinzipien als Teil der Unternehmenskultur und der Umsetzung von KI in der gesamten Gruppe betont.

💡Regulierungssandkasten

Ein Regulierungssandkasten ist ein Rahmen, der von Regulierungsbehörden eingerichtet wird, um Unternehmen die Möglichkeit zu geben, neue Technologien und Geschäftsmodelle in einem kontrollierten Umfeld zu testen. Im Video wird die Möglichkeit erwähnt, dass solche Sandkästen von der Luftfahrtbehörde oder der nationalen Regierung unterstützt werden könnten.

💡Human Centric

Ein humanzentrierter Ansatz bedeutet, dass Technologie und Prozesse darauf ausgerichtet sind, die Bedürfnisse und Erfahrungen von Menschen zu verbessern. Im Video wird dies als Teil der Unternehmenskultur und der KI-Strategie hervorgehoben, die auf die Wertschätzung der menschlichen Expertise und die Schaffung von Systemen um den Menschen herum basiert.

💡Autonomie

Autonomie bezieht sich auf die Fähigkeit von Systemen oder Maschinen, Entscheidungen und Aktionen ohne menschliche Eingabe oder Überwachung zu treffen. Im Video wird die Forschung und Entwicklung von autonomen Flugzeugen und Hubschraubern als Teil der langfristigen Vision der Luftfahrtindustrie beschrieben.

Highlights

Abus is known for manufacturing aircraft, with 14,000 operational every day.

Abus has divisions in defense, space, and is the largest helicopter manufacturer in the world.

Abus is focused on a responsible approach to AI, aligned with company values.

AI is applied to improve daily business, operations support, and game-changing applications.

Generative AI is used to find information efficiently within technical documentation.

AI is utilized in machine learning, computer vision, and optimization for core operations.

Surrogate modeling in engineering design has significant potential.

AI is used for customer efficiency, predictive maintenance, and improving the customer experience.

AI is being researched for autonomy in flight, single pilot operations, and safety improvements.

Abus operates its own satellite fleets for imagery analysis and various services.

AI features are tested on aircraft for automated landing and taxiing.

The company uses AI for complex operations on the shop floor, prioritizing work, and quality analysis.

AI is combined with robotics in aviation for human-machine augmentation.

AI is used in the engineering process to create digital twins for aircraft design.

Abus has different areas of sensitivity due to regulation in aircraft quality inspection.

The use of generative AI is seen across the company for chatbots and information retrieval.

Abus encourages a culture of open-source within the organization for AI innovation.

AI Act encompasses AI use cases and sectorial regulations, important for aircraft regulation.

A human-centric approach to AI is valued, with different levels of augmentation outlined.

Abus aligns with the human-centric approach and values human expertise.

The company follows a human-centric approach to developing new aircraft programs.

Abus focuses on responsible AI aspects and regulation, considering what is technically possible, sustainable, and preferable.

Ethics by Design is an approach that considers stakeholders and ethical values from the early stages of a project.

Abus has developed AI principles that align with European and company values.

Adaptive governance and AI principles are key for building safe and reliable AI systems.

Abus is looking forward to the interplay between AI regulation and innovation.

Transcripts

for having me um and thanks for joining

the session and the talk um so yes J

Schon back thanks for the introduction

um just a word about Abus so you

probably know us from from the aircraft

you you fly and that we manufacture

there's about I think 14,000 abis

aircraft operational every day flying

taking off every two seconds um so

that's what we're most known for we also

uh have larger divisions in defense and

space for example the EUR fighter you

may know um as well as large satellite

programs we are also the largest

helicopter manufacturer in the world uh

so that's you know broad divisions uh we

try to be quite coordinated on the AI

front um and yes something that's really

keen and sort of important to us is not

just sort of the compliance of the

minimal approach to AI but really how do

we U merge a take on AI um with our

company values and a responsible

approach uh companywide so happy share

with you about that um starting um with

an overall picture of where AI is

relevant where do we employ it um some

examples of use cases maybe focus on gen

um I'll show you a bit on uh the

regulation challenges we have in

different areas of the company and how

that leads us to really think very

closely around regulation but also sort

of what is our way how do we you know

responsibly handle the

technology uh oh sorry can I go back

that's that one right yeah okay so first

off um the different application Fields

the value streams where do we apply AI

um on the one side it's improving sort

of the Daily Business this can be you

know as part of your your workspace

tooling that that you know is quite

important just to have like available to

you um in terms of you know uh

generative AI features it will also come

through you know features coming through

applications we all use um but something

we see quite a bit is is especially

around generative AI just the sort of

daily tasks of finding information so

you know we're we have very technical

language we have tons of documentation

processes references so a lot of jobs in

abis and Engineering or customer support

that really rely on finding information

at the right point in time that's

relevant and take that taking a lot of

time out of their job so here things

like gen are really really interesting

um in terms of operation support so also

when I look at sort of traditional AI

maybe machine learning computer vision

uh optimization there's a lot of fields

here where um you know they support our

core operations help us bring more

efficiency to processes be better at

planning at uh uh forecasting

understanding patterns and so on so um

this is really important also when you

think further on sort of there's the

operational improvements but there's

also sort of gamechanging AI

applications when I look at how we how

engineering design is done so things

like surrogate modeling have really big

potential already U utilized in the

design processes um if you look at for

example you know how we design an

aircraft you know that's a very very

heavy process and we're sort of just

preparing to launch a new aircraft

program meaning we'll design the whole

aircraft platform again from scratch so

having endtoend

um AI type you know capabilities data

continuity and means to create true

digital twins is going to be a true

GameChanger in for example the whole

design of an aircraft we have customer

efficiency so this is supporting our

customers uh with a flying product so

things like you know predictive

maintenance really having that

understanding and of the aircraft's

operation um in the fleet you know you

can imagine it's quite different if

you're uh operating an aircraft in the

desert versus you know in in northern

Alaska so um I can help really give

customized services and really improve

the customer experience but it can also

be implemented in the products s so the

aircraft the helicopters and so on um

where you know that's uh when you look

towards autonomy of flight single pilot

operations augmentation of pilots and

improving safety of the aircraft

themselves so that's a longer sort of

research but is very actively looked at

I can tell you as

well and some examples from the

different divisions here uh so on the

left side you see uh satellite imagery

analytics services from defense and

space so we operate our own satellite

fleets and have satellite uh imagery

analysis widely available since quite a

long time this can be for example object

detection of assets uh for intelligence

purposes but also for sort of you know

climate monitoring uh Climate Services

analysis of n natural catastrophes or

Insurance Services even um on the

products themselves for the aircraft in

the middle you see um essentially one of

a350

of our own test aircraft that utilizes

AI features with the computer vision for

example to help with automated Landing

automated taxiing um basically

augmenting as well the pilot we're

testing single pilot operations so to

increase the safety of aircraft that the

pilot is incapacitated can the aircraft

still land can AI support a pilot in

very stressful situations so these are

things that are being researched that

will take a longer time and similar on

on the helicopter we have our test Fleet

helicopters equipped here as well we

call this the pyer lab where um it also

supports I believe a 360 sort of view

with object detection and sort of seen

understanding yeah and of course in the

process in the company itself so um many

areas that are like really interesting

is on the shop floor itself uh very

complex operations that require a lot of

information you have to be very careful

to do the right steps at the right time

a lot of validation there but here as

well I mean workers already utilize AI

capability is a lot when it comes to

sort of prioritizing the right you know

outstanding work um to make sure you

know you don't disturb the production to

uh analyze quality occurrences across

the you know 10 different final assembly

lines we have for aircraft so that's

already working quite a bit um AI for

example in combination with

robotics Aviation is still a very manual

labor in most of the cases um when

robots come into play it's usually not

sort of you know a single production

with you know thousands of of aircraft a

day is more sort of human machine

augmentation where the robot needs to

understand where it is really needs to

understand its context and on the right

side is the engineering what I mentioned

the sort of you know the digital twin

the whole design process um that's a

really large field as

well um and a bit of a focus on some of

the challenges that lead to where I'm

going to end um if you look at you know

those proses I showed before you can

imagine there's quite different areas of

sensitivity some of the processes we

have are under regulation so of course

the Quality Inspection of an aircraft is

very related to safety and air

worthiness so not just do we have ai act

to care about what we also have the

asasa which is the European Aviation

sort of safety administration that

regulates the process that we have so

they also have a say and if we basically

augment our quality process with AI they

also so are part of the regulation and

sort of validating um how we change the

process and use the processes so that

can be quite different than when you're

in like a support function um and just

supporting an administrative task one

area we see what I mentioned in the

beginning is with Gen um that's really

interesting where we see a lot of value

in the short term all across the company

is utilizing gen for sort of types of

chat Bots that understand your data

scope your reference documentation quite

well and can just help you and find

really information really well um we do

a lot of prototyping across the company

in many different fields and processes

and find that that is something that you

know brings a lot of value across all of

the functions just to name a few

examples that can in include for example

for procurement or legal analyzing uh

contracts authoring contracts

identifying for example liability and

Legacy contracts uh quite a complex task

for you know a huge Legacy amount of

contracts you have engineering support

answering technical customer inquiries

so this can be from well we have an app

for example where the active Fleet the

airlines themselves raise technical

questions to us and we have hundreds of

Engineers on our side analyzing and

making sure they give the right answers

that can be referencing to manuals that

can be analyzing from previous requests

that this asked before so they take

quite a long time to analyze these

topics we see that with gen they can

very easily find similar questions from

before you know directly find the

reference to the right manual and really

speed up that whole process so it's

quite it's quite promising

overall um and just a word on on our

approach of that is as well because like

I'm from the central team for AI we have

a lot of technical experts we run like

the key projects um but it's such a

large company you have to democratize

and Federate the organization and allow

for this experimentation to happen

across the company in a compliant way um

so we as the center of excellence really

tried to um give access to cloud ready

platforms that are already pretty much

in the architectural environment you

would put things into production but the

really sort of secured sandbo spaces um

that we can open to any business team

that's you know technically well vered

we have a large population of very well

technically vers people who've upskilled

themselves in coding analytics Ai and so

on so that's been a huge Ena but it

raises all these questions so right if I

even increase the exploration all across

the company you want to make sure you do

that in the right way and with AI Act of

course you need to be aware of all

initiatives running have everything

registered do the right risk

classification and so I think it's

really an opportunity to think of how do

you allow early exploration and a

compliant sort of safe way and that over

time the governance that you also impose

um sort of really strengthens so when

you put things into production you do it

very meticulously very well so these are

the questions that really raise and you

I think you as a central team you really

have to find aners to that um one word

is also sort of the whole culture around

that so from uh building a kind of you

know open source culture within the

confines of of an organization is quite

enabling here as

well um but yeah so you you still have

to ensure that basically all these

imposing aspects of you know

regulation um aligning that with your

company values is is all in line and

still happening um and you do it with

the right Spirit of course there's a lot

of fear in all areas of the company um

around what will I take my job away do

we do I trust my employer to handle that

responsibly uh so um there's a lot of

these questions coming in we think it's

quite important to to address us the

right way just to give you an idea um on

how this works with the AI act I

mentioned it a bit before already so AI

act overall encompasses of course you

know AI type use cases and for the

sectorial regulations so anything that

is under regulated processes on the

aircraft itself we'll have the ASA

regulation as well um so quite

interesting to see how this will play

out in terms of like how do we manage um

regulatory sandboxes will this come from

measa or the German governments the

national governments how will that be

supported so something we also

appreciate I think the exchange with

bitcom um and the peers here quite a bit

to see that you know the right approach

the right guidance

um yes and on the Outlook when it comes

to Human Centric approach so as's Vision

as well aligns with that quite nicely

with us so they when they lay out how

this will work um they basically see it

there different levels of augmentation

um with the level one being assistance

to the human so this is basically human

using AI assistance but still in full

control of the workflow level two sort

of human machine pairing where it's like

a coordination and the human can still

uh basically um you know has control

over the decisions and level three is

when you know the systems are already

kind of autonomously start to make

decisions take take actions but but are

still overridden by the human there will

possibly be a level four for like sort

of full autonomy um but as you can see

aasa is taking a very sort of human

Centric approach to this and I feel I

mean we feel like this is well aligned

with our company culture as well we

really value the the human Centric

approach we value the highly the human

expertise in everything we do so um we

align with that quite well to give you

an idea on how this looks like in

aviation thinking about all these

processes on sort of production and on

the aircraft we're looking around it's a

bit hard to see here but uh level one

sort of the first use cases will be

validated by asasa probably next year

the level two guidance uh the guidance

itself um is already available and we

expect first use cases to be deployed 20

2035 and then 2050 plus is like the the

later almost at ionomy so far away for

the aircraft itself but you can see

basically if we think around developing

the new aircraft program this will

really come into question if we will see

this some but even if it's not autonomy

it's really also making the aircraft and

the whole system

safer so when we really think about how

do we how we go about all that

responsible aspects and the the

regulation um when you see it bring it

sort of out with the you know the

furthest is you're trying to you

whatever is feasible you approach um if

it's technically possible

the next question is is it sustainable

um so um does this make sense on the

long term is it acceptable in the sense

of compliance so do we you know fulfill

all the rules to deploy system but then

we really want to focus on the

preferable so essentially allowing

Innovation making use of innovation but

doing things the right way that we

believe are in line with our values with

you know European core values um with

the company overall we don't think that

this necessarily has to mean

um that you know this slows you down in

terms of innovation but it's really more

a ways of aligning things with your

corporate culture and making sure that

um making sure that you take these

concerns early on and are basically

developing even safer systems because

you anticipate much

better so what we try to approach is

that what we call it towards ethics by

Design so this can be as simple as very

early on in your project basically uh

you know giving room to understand who

are your stakeholders who are your

actual customers and users of the system

and just asking them about their

concerns and basically just making sure

that you have principles you know

ethical values that you adhere to and

you make that you know voicing of

concern part of the conversation and

include that in your risk management we

believe it basically Fosters trust with

the employees with the teams that you uh

try to assist with AI we believe it

creates Innovation and value you you

still focus on valuable use cases um and

it contributes to risk reduction because

you're much more anticipative of um

basically in a structured way being able

to voice concerns structurally as well

as sort of just by giving the

space so what we've done is we've been

aligning what we call these AI

principles it's really you know coming

from the core of European values but

also from our company uh values that

really merge and this is something we've

and lined on as sort of a culture to

implement AI across the whole group it

starts with safety first so that's at

the heart of everything we do we look at

saem systems from the safety perspective

um especially in the sort of production

related areas I've talked about while

being in sustainability we Implement AI

for societal good for the good of our

employees as well in a sustainable

manner we the systems that we build that

we integrate um uh we ensure that we

take accountability for them and we sure

transparency for them uh privacy data

protection of course always an important

Point

fairness I mean monitoring for buys you

know fairness equality is always a core

thing and what I mentioned the human

agency and oversight in any system that

we build we value the the human Centric

expertise and build systems around the

employee and the

human to give an idea what the this

looks like in this Innovation cycle in

this Federated uh network of teams

implementing AI it can be as simple as

you know having like a lab stage an

early stage where you have a sort of

preliminary ethics assessment along with

your qualification of a use case uh this

area where you raise concerns you may

have sort of a checklist uh that you

later on bring in as well just to make

sure that you're you know uh having good

intent with your system that you know

stakeholders and users of a system are

wellard

um but that still is lightweight so

ideally if you have like a lowrisk use

case that you know doesn't e concerns

you can still sort of you know smooth

sale through the first phase and there's

really not much of sort of heavy

governance that You' be approaching um

if of course you're really going in a

system that's sort of in ethical

challenging territory you would have

these kind of questions pop up early on

and even avoid such a use case being

addressed um or be very diligent and go

through like like an expert Board of

really looking in detail um around the

challenges you want to be very much

aware about um and anything that gets

deployed goes through that sort of

meticulous process for like a real

ethical

assessment that sounds very fluff but

how can you actually translate that into

sort of credible sort of technical

aspects and sort of associated

mitigation measures and strategies just

to take you a few examples so what I've

listed on the left side I like our

principles around this and they can lead

to identifying specific Harms that can

lead to basically identifying the

associated security threats and defining

technical mitigation measures so this

can be things like you know implementing

guard rails bias checks and outputs um

safety filters and output filtering um

so we believe that basically by

approaching projects like this in sort

of an

Adaptive um governance is a good way to

build safe and reliable AI

systems so in terms of key takeaways so

we all um very much interested in

looking forward to the interplay and the

regulation coming from the AI office but

also from author um regulation

authorities democratizing an i for us

basically is a scaling and wide

Innovation um but adaptive governance is

really important and AI principles I

would recommend everybody to sort of

identify their corporate culture blend

that with the value is important to you

and make that sort of the mindset in

your company and for us ethics is lot of

steering it doesn't have to be a break

but it's actually a steering wheel in

your AI approaches thank you

[Applause]