Latest / Engineering Matters / #286 AI In the Real World

Transcript

#286 AI In the Real World · Engineering Matters

0:09
The group that I'm particularly
0:11
focused on, really, is the
0:14
convergence of
0:16
response to applications and transformation.
0:21
So how do we actually take this technology,
0:24
not just at a high level or as
0:26
a sort of, you know, an AI whitewashing
0:28
type
0:30
blur, but how do you actually practically apply
0:32
it around some
0:34
significant industrial problems?
0:37
Earlier this year, we talked to Darren Martin,
0:40
CDO of Atkins Realis,
0:42
about the relationship between AI and humanity.
0:46
We saw how Darren had used big data
0:48
in his early career as a clinical psychologist
0:51
to identify effective treatments for severe mental illness.
0:55
We saw that machine learning and big data
0:58
could be powerful tools,
1:00
but they were not a fix for every
1:02
problem.
1:04
But what does it mean to live in
1:06
a world where AIs control autonomous devices that
1:09
work alongside us?
1:11
As machine autonomy replaces or supplements human work,
1:15
what does that mean for how we design,
1:17
build, and live in cities?
1:20
These are some of the questions that Darren
1:22
has been examining along with fellow members of
1:25
the World Economic Forum's
1:27
AI Governance Alliance.
1:29
This really boils down to,
1:31
for example,
1:33
how my robots and people
1:36
relate to each other and work with each
1:38
other in a safe way
1:39
on a mega
1:41
project, like an, a project in the Middle
1:44
East where we're building huge
1:47
cities, and how might we
1:48
help achieve that construction
1:51
faster
1:52
with the AI
1:54
and the people working together.
1:57
Bringing that together in a safe, practical way
1:59
is really important.
2:00
But if you then flip out into
2:03
different
2:04
areas of opportunity
2:06
and discuss the pros and cons of this
2:08
technology with and the applications
2:11
The impact of these technologies will be felt
2:14
throughout our society
2:16
We will be leaning into conversations about health.
2:19
We'll be leaning into conversations about
2:21
defense and security.
2:24
It has the potential to do great harm,
2:27
but also has the potential to
2:29
do great good.
2:45
Welcome to Engineering Matters.
2:47
I'm Johnny Dowling.
2:49
And I'm Rian Owen. Earlier this year, we
2:51
heard from Darren Martin, chief digital officer of
2:54
our partner, Atkins Realis,
2:56
about the importance of considering human needs in
2:59
our deployment of AI.
3:01
In this episode, Darren returns to look at
3:03
what it means for humans to live and
3:05
work alongside autonomous devices
3:08
and what that implies for the relationship between
3:10
big tech and architecture,
3:12
engineering, and construction professionals.
3:14
One day, autonomous devices will be close at
3:17
hand wherever we go. But for now, we
3:20
are in some corners of the world most
3:22
likely to encounter them on the road.
3:33
In San Francisco, for example,
3:35
one can get in a car without a
3:37
driver
3:39
and pay a fare and go.
3:44
So the Uber concept has evolved to a
3:47
a Robot Taxi
3:49
concept.
3:50
Now taking the learnings from San Francisco, which
3:52
is is very healing, but it's a grid
3:55
city. So it's, you know, it's a traditional
3:57
American city with the grids, and taking that
4:00
to where I live, and I latest it
4:02
with our single lane country lanes and, you
4:04
know, the all of the complexity that comes
4:07
with having a deer in the middle of
4:08
the road, and and how does the robot
4:10
taxi deal with
4:13
that? The ability of AIs to operate potentially
4:15
dangerous equipment in public spaces
4:18
is being proven in the neat grids of
4:20
American
4:22
cities. Development now is focused on improving the
4:24
safety of these systems and verifying that safety
4:27
so that they can be deployed
4:29
around the world.
4:31
It's bringing these two things together, like like,
4:33
what does the consumer need? How might that
4:36
be enabled?
4:38
And also,
4:39
what are the benefits for doing that?
4:42
If these autonomous devices can be shown to
4:44
be safe enough for our roads,
4:46
then they'll be well on their way to
4:48
working alongside us everywhere.
4:54
We can already see autonomy
4:59
autonomy in the workplace at ports around the
4:59
world. Here, autonomous rubber tire gantry
5:00
cranes move containers around ports
5:03
without human intervention.
5:05
In applications like this, the autonomous equipment is
5:08
separated from human staff by a fence.
5:11
Part of the terminal will be exclusively autonomous.
5:14
And part of it will be operated by
5:16
people.
5:19
There are many
5:21
quite advanced examples now with
5:24
the safety case has been lowered
5:27
for the technology, because people are not allowed
5:29
in there. And so if the technology goes
5:31
wrong,
5:32
it's embarrassing, it's expensive, but no one was
5:35
hurt. So the profile of it is much
5:37
lower.
5:39
But to extract the real efficiency benefits of
5:42
AI autonomy,
5:44
we need to be able to work with
5:45
these devices in the same space.
5:47
So the question is,
5:49
under what conditions can one move away from
5:51
a a closed loop scenario?
5:55
So
5:56
it's about the interoperability
5:58
of of humans
6:00
and the robot systems.
6:03
That could then allow a new generation of
6:06
robot workers to work in environments far more
6:08
complex and unpredictable than a modern port.
6:12
If we then translate that back into
6:15
how do we operate
6:17
a really complex,
6:19
dangerous, multi level, multi layered, multi machine
6:22
construction
6:25
site. The advantages of using a robot to
6:27
do something that
6:29
is dangerous, like working at height, working with
6:31
heavy loads,
6:33
working with dangerous substances,
6:35
for example,
6:37
is really appealing.
6:39
But some workplaces are even more risk laden
6:42
than a construction site, like a nuclear power
6:44
plant.
6:49
We're currently,
6:51
talking to a number of research institutes
6:54
to really explore how that is gonna work
6:56
in the future and start to trial it
6:58
in our ways of working. And and as
7:00
an organization where
7:02
we know we own,
7:04
one of the 6
7:06
nuclear technologies,
7:08
that are licensed on this planet and that
7:11
there's a significant
7:12
need for
7:14
more nuclear
7:15
facilities to be built.
7:17
The interoperability
7:18
of of humans
7:20
and the use cases of robots to keep
7:23
people safe and to to,
7:26
make sure we can design, run, and maintain
7:28
those facilities in the safest way is really
7:31
attractive.
7:33
There's a whole community of people that have
7:36
spent their whole lives working in those environments
7:39
and they're extremely
7:41
proud of what they've done
7:43
and long may that continue.
7:45
Their ability to get suited up and and
7:48
operate in that environment
7:49
and then move away from that environment, it
7:52
might take a whole day to spend
7:54
around 40 minutes or
7:56
maybe an hour in that environment, but it
7:58
would take a whole day to kinda go
8:00
through that process.
8:01
So your time on tools and your ability
8:03
to
8:04
interact and perform accurate maintenance
8:07
could be significantly improved if the human safety
8:10
factor
8:11
could be lowered.
8:17
An engineering business like Atkins Realis is well
8:20
placed to develop systems that can work in
8:22
these hazardous environments.
8:24
On one level, engineering is all about the
8:27
assessment and control of risk.
8:29
But as engineers,
8:30
architects, and construction managers build new infrastructure,
8:34
they must also think about creating an environment
8:36
that is suitable for the use of autonomous
8:38
machines.
8:40
But the
8:41
disadvantages
8:42
that occur when you have these closed loop
8:45
systems, you and you don't have people and
8:46
machines together, There's still many things the machines
8:49
can't do very well,
8:51
not nearly as well as humans. So the
8:53
advantage of that interoperability
8:56
of these two entities is is appealing. It's
8:59
and for me, personally, it is really appealing.
9:01
Like like,
9:03
from an academic perspective it's appealing, but also
9:05
from a real value
9:07
in terms of
9:08
doing things faster
9:10
and safer,
9:12
which is right at the ethos of
9:15
the construction industry.
9:18
From a design perspective,
9:19
how do you how does one design a
9:22
construction site? How do you even design a
9:24
city?
9:25
Where, as I mentioned before, you know, the
9:26
the robot taxis can operate in? But also,
9:29
how do we actually design
9:30
buildings so that we can
9:33
design for manufacturing
9:36
with people,
9:37
with 3 d printing,
9:39
and with robotic devices that have high mobility
9:42
and interoperability with people.
9:45
And I think I think that whole piece
9:47
is
9:49
we're scratching the surface of that.
10:01
Construction sites and the building materials supply chains
10:04
that support them can be designed to support
10:07
the use of autonomous devices.
10:09
We can imagine robots able to manipulate and
10:12
install large modular components
10:14
alongside human workers,
10:16
And this may give us the pace needed
10:18
to build new cities to house a rapidly
10:21
growing, urbanizing
10:23
and aging population.
10:28
As we think about how much infrastructure's gotta
10:31
be built to replace the aging infrastructure we've
10:34
got,
10:35
how the population keeps on growing, and there
10:37
are many more sustainable
10:39
energy transition
10:40
initiatives that have to be built
10:43
to
10:45
make sure we
10:47
are able to all survive on this planet.
10:49
I think that the ability to think about
10:52
how would
10:53
how do we design it in a way
10:55
that allow for modular construction, design for manufacturing,
10:58
and how would we
10:59
design it so that
11:01
there were some closed loop
11:03
environments where it's really not safe for a
11:05
human being to be in, but we designed
11:08
also for the for those two things to
11:09
come together.
11:11
The urban planners of the near future
11:13
shouldn't just be considering how construction sites can
11:16
safely allow for humans and machines to work
11:19
together.
11:20
They need to think about how to make
11:22
the best use of space
11:24
within cities
11:25
where new forms of transportation
11:27
are possible.
11:29
When I bring this back to an AEC,
11:30
an architecture engineering construction
11:33
situation,
11:34
the master planners, the architects of the new
11:38
cities,
11:39
the retrofitters of our existing cities,
11:42
how are we thinking about
11:44
transportation,
11:45
how are we thinking about social value, so
11:48
how are we thinking about
11:49
walkways that allow people to walk and cycle
11:52
and scoot for example and be more active
11:56
and be safer
11:57
but also how might we be thinking about
12:00
the utilization of car parking space?
12:03
Is that now social housing opportunities?
12:06
We've seen older people returning back to cities.
12:10
Is that community space for communities to
12:14
experience
12:15
different social interactions than our cities have typically
12:19
seen and how might you
12:21
design
12:23
a roadway system for robot
12:25
cars
12:27
preferably robot taxis because if we all have
12:29
robot cars parked outside wherever we choose to
12:32
live and work, we're taking up unnecessary real
12:35
estate space that
12:37
just doesn't really add a lot of value
12:39
other than the prestige of owning the car.
12:42
We live today in cities that have for
12:45
decades
12:45
been closed loop systems,
12:48
designed to allow for the widespread use of
12:50
unsafe,
12:51
unreliably controlled machines.
12:55
We mark off huge swathes of our cities
12:57
for cars, segregating roads from pavements and dedicating
13:01
space for them to sit unused.
13:04
Now urban planners should be thinking beyond these
13:07
risk controls.
13:08
If truly safe autonomous vehicles
13:10
designed to be shared and used constantly were
13:13
available,
13:14
what other uses could we find for those
13:16
spaces?
13:18
If you are gonna be in a car,
13:20
let's make sure that car is a robot
13:22
vehicle that isn't just going to park and
13:24
take up precious real estate.
13:27
I can trust you in a study at
13:28
the moment between
13:30
Cranfield University, Open University
13:33
and Oxford and Cambridge
13:35
looking at the some of the connections between
13:37
those
13:38
cities, those educational institutes and how do we
13:41
create thriving
13:43
natural
13:44
corridors between waterways,
13:46
cycle path routes, and so on to make
13:49
the the connectivity
13:50
of those parts of the world better.
13:59
Autonomous devices will change the way that we
14:01
build and design cities.
14:03
They will increase the pace at which we
14:05
build and create new space for us to
14:07
build in.
14:08
AIs are pattern matching machines.
14:11
They can mimic human communication,
14:13
allowing for their control by anyone,
14:16
and they can select designs that are safe
14:18
and buildable.
14:19
Taken together, this will allow for a tailored
14:22
design process
14:23
previously only available to the very rich.
14:26
And so when I think about high end
14:29
luxury goods, there have been some organizations
14:32
that have really brought in
14:35
the consumer into the personalization
14:38
process.
14:39
So hyper personalization.
14:43
Consumers of trainers and sunglasses can adapt designs
14:46
to their own taste, but within the options
14:49
offered by a drop down menu.
14:53
What would it mean if this personalization
14:56
could be taken beyond a set of fixed
14:58
choices
14:59
and intersections of the economy more vital to
15:01
life than luxury consumer goods?
15:03
But all of these are within parameters, so
15:06
the manufacturers
15:08
are able to offer a premium
15:11
differentiated
15:12
product,
15:13
but something that they can still affordably
15:16
manufacture and make profit from in a sustainable
15:20
and safe way.
15:22
Around the world, cities are now being planned
15:24
from scratch.
15:26
Robot workers will deliver the pace needed to
15:28
house 1,000,000,
15:30
and AI systems will allow residents to have
15:32
a real say in what their new homes
15:34
will look like.
15:35
Where we need to go
15:37
is the ability to
15:40
draw in
15:41
enough parameters
15:43
so the consumer can have ownership
15:46
and identify with that modular experience,
15:49
but at the same time, the benefits
15:52
of,
15:53
the construction process, the efficiencies, the sustainability
15:56
of it, the ongoing
15:58
operational maintenance of it are much better than
16:01
we've achieved in the past.
16:09
AI and autonomy
16:10
will allow us to build cities in a
16:12
new way.
16:13
We'll be able to rapidly construct
16:15
personalized
16:16
homes
16:17
and to interact and travel in a more
16:19
human way, free from the risk and waste
16:21
of cars.
16:23
But the climate crisis makes clear that even
16:26
designing cities from scratch is not enough.
16:30
We must leverage these new technologies
16:32
to consider and analyze our whole planet.
16:37
Part of my job and other people's at
16:39
Atkins Real Estate's job is to constantly
16:42
look at
16:44
what do certain technology partners bring
16:48
to our customers,
16:49
and what value together do we bring
16:52
to drive a better societal outcome for our
16:55
customers.
16:56
And one of the examples that, you know,
16:58
we're really interested in
17:00
is the impact of climate change
17:02
and the impact it has on society
17:05
and thinking about resiliency.
17:08
The risks and uncertainties of climate change have
17:10
undermined the functioning of insurance markets.
17:13
Without understanding the impact of climate change and
17:16
the resilience of our buildings and infrastructure,
17:19
many regions are becoming increasingly
17:22
uninsurable.
17:23
And just, you know, at the moment there
17:25
are some questions being asked around the bond
17:28
market,
17:29
the insurance bond market and
17:32
with the
17:33
frequency and the the impact of
17:36
climate change and weather,
17:38
impact on on critical infrastructure across the globe,
17:42
how do we keep on paying for it,
17:44
who underwrites it and it's a bet that
17:47
in the past used to come off quite
17:49
nicely. It's quite a profitable industry,
17:51
but actually a lot of these bets are
17:53
starting to become
17:55
loss making.
17:56
And in some places now,
17:59
all over the world.
18:00
Our current tools don't allow for reliable assessment
18:03
of risk and resilience.
18:06
Getting more precision from AI
18:09
about the exact nature
18:12
or and likelihood
18:13
of resiliency
18:16
impacts and also putting in remedial action to
18:19
mitigate against them where possible
18:22
is
18:23
really important because if you own real estate
18:25
or you own a business or you run
18:27
a school, you wanna make sure that it
18:29
can be insured and protected.
18:34
Atkins Realis is already using AI systems to
18:37
analyze the impact of climate change across tens
18:40
of thousands of properties
18:42
at a time.
18:43
We've done work with the federal emergency
18:46
management agency, FEMA, in the US,
18:49
where we've
18:50
been able to do surveys on properties that
18:53
have been impacted by, severe
18:55
climate change and prioritise
18:57
the surveying of those properties
19:00
to measure the impact
19:02
and to help,
19:03
free up funds
19:05
to get people back
19:07
in those homes or in new homes.
19:11
That's a 150,000
19:13
potential surveys that needed to be done.
19:17
Relying on human assessment
19:19
would leave those impacted by climate change in
19:21
vulnerable situations for weeks or months.
19:25
AI can allow support to be delivered where
19:27
it's most needed, when it's most needed.
19:31
With, being able to use AI to 0
19:34
in on
19:35
around,
19:36
1 5th of them that really needed to
19:38
be done as a huge priority
19:40
to,
19:41
make sure people were
19:43
helped and looked after as a result of
19:45
that,
19:47
climate
19:48
impact on their lives. And so
19:50
zeroing in and removing waste and prioritizing and
19:54
getting help and resources to people is something
19:56
that that I think
19:57
AI is a really good example of where
19:59
it can do it, and I think we
20:00
need to get better at that.
20:03
AI based analysis can identify the impact of
20:06
climate events across cities and regions.
20:09
One vision of the interaction of machine learning
20:12
and city planners has already been tried in
20:15
Canada.
20:15
In this vision, tech led the way and
20:18
engineering followed.
20:20
Google Sidewalk Labs,
20:22
it's, it's this company as part of the
20:24
Alphabet group that was set up,
20:26
and
20:27
its first project was to look at the
20:30
Quayside project,
20:31
on the South Bank of Lake Ontario,
20:34
which was a port area
20:37
back in Toronto,
20:39
and then the idea there was that
20:42
big government had aligned with big tech, Google,
20:45
and that this
20:47
large,
20:49
real estate project was gonna be put put
20:51
in and that it would it had some
20:53
fantastic benefits. I was looking at some of
20:55
the stats and I'm not an expert on
20:57
on this area, but just looking
20:59
at the stats,
21:00
it said by 2040, it would generate around
21:03
44,000
21:04
jobs. So it's an industrial pool, and it
21:07
was being turned into a,
21:10
you know, as we have in the UK,
21:12
you know, some really good
21:13
waterside
21:15
retail
21:16
business
21:17
and
21:18
residential,
21:20
quadrants and and and locations.
21:22
And
21:23
it was reducing emissions of that industrial site
21:26
by 89%,
21:28
and it was reducing the cost for
21:30
social housing,
21:31
assets
21:32
that were being put into that alongside some
21:35
higher end housing assets.
21:37
But the typical cost for a social housing
21:39
asset was
21:41
40%
21:42
below the market rate.
21:46
Sidewalks Labs' plan had been to offer more
21:48
jobs and cheaper housing in an attractive environment,
21:52
managed and maintained using sophisticated
21:54
monitoring.
21:55
But by 2020,
21:57
the project had fallen apart.
21:59
Stymied by the impacts of COVID and by
22:01
a lack of trust in the monitoring systems
22:04
used to manage the district.
22:06
Darren believes that projects like this need to
22:08
draw on the established expertise
22:10
of architecture,
22:11
engineering, and construction businesses.
22:17
Big Tech came in
22:18
and did a deal with big government,
22:22
and local government, and local sentiment, and human
22:26
permitting
22:27
considerations,
22:28
objections,
22:30
community resistance
22:32
wasn't dealt with in the same way
22:35
that perhaps Atkins Realis would have
22:38
understood
22:39
the nuances
22:41
of engaging with local government, understanding the planning
22:44
rules,
22:45
understanding how to engage with communities
22:48
to help them understand what the changes and
22:50
what the benefits are for them, but also
22:53
to adapt the master planning concept
22:55
based on consultation with those communities.
22:59
But just as Alphabet is moving out of
23:01
city building,
23:02
other big tech players are moving into engineering.
23:06
Cognizant
23:08
have put forward
23:10
and acquired
23:11
7 companies
23:12
in the with engineering focus since 2021.
23:15
So Cognizant typically would have been, you know,
23:19
Indian HQed
23:20
IT and Business Process Outsourcing Organization
23:24
competing with the likes of, you know, IBM,
23:27
Japan's Fujitsu, and so on. Probably doing so
23:29
on a lower price point.
23:32
It made a 1,300,000,000
23:34
acquisition
23:35
of an engineering company
23:37
in June 24.
23:39
Tata,
23:40
for example, bought an automotive
23:43
engineering company recently.
23:45
Accenture
23:46
have bought
23:48
IoT
23:49
companies,
23:50
a whole range of industrial and engineering
23:54
expertise.
23:56
As the space between big tech and engineering
23:58
closes,
23:59
those leading and investing in these sectors
24:02
should really think critically about the skills each
24:04
sector brings.
24:06
You have to ask the question,
24:08
is
24:09
a company like Cognizant becoming an engineering company?
24:18
Professional services businesses in architecture, engineering, and construction
24:23
should not, Darren believes, try to themselves into
24:26
tech firms.
24:28
Big tech is not accustomed to dealing with
24:30
the human complexity of urban planning.
24:33
This is not a space in which you
24:35
can just move fast and break things.
24:37
So how do these 2 sectors work together
24:40
in a new age of AI driven autonomy?
24:43
What can they learn from a project like
24:45
Sidewalk Labs?
24:47
Perhaps there wasn't the,
24:50
awareness
24:51
and the,
24:53
sensitivity
24:54
and experience
24:55
that an architecture engineering and
24:57
construction company would apply
24:59
in getting that project to work.
25:01
You can work out, for example, when the
25:04
lifts are being used
25:05
and, you know, are people exiting the building,
25:08
and how many people are in the in
25:10
the lifts, and what's the weight of the
25:11
lifts.
25:12
There are there are a number of less
25:14
intrusive ways
25:16
of understanding
25:17
how the,
25:19
buildings and the
25:20
the environment are being used
25:22
that don't require, you know, quite personal sense
25:26
of information to be shared. So once you're
25:28
in a community,
25:29
people don't want to feel that they're being
25:31
observed. Their movements are you know, who are
25:33
they talking to? How long are they spending
25:35
in these spaces?
25:38
The experienced human planners of an AEC business
25:41
can work alongside local governments and citizens
25:44
to deliver projects that really consider
25:47
human
25:48
needs, desires, and privacy.
25:51
While emerging technologies will have a vital part
25:54
to play, it is this human expertise and
25:57
experience
25:58
that will be key to meeting human needs.
26:02
We form part of a partner ecosystem,
26:05
which in some cases might be tech enabled,
26:08
in many cases it will.
26:11
We're highly differentiated
26:13
from Big Tech in our ability
26:16
to
26:18
create environments that are both progressive from an
26:21
economic standpoint but sympathetic to to communities and
26:26
enhance the communities.
26:41
Engineering Matters is a production of Reebi Media.
26:44
This episode was written and produced by Will
26:46
North and hosted by me, Rianne Owen, and
26:48
by Johnny Dowling, editing and series supervision by
26:50
John Young, sound engineering by Ross Macpherson, and
26:53
the human intelligence that steers our use of
26:55
data is Rory Harris.
26:57
Special thanks to our partner for this episode,
26:58
Atkins Realis, and thank you for listening. You
27:01
can find us on all podcast apps and
27:03
on our website engineeringmatters.
27:04
Rebi.media
27:05
and on LinkedIn.