(music)
- I'm Alison Berman. We're here live at the Global Summit.
I'm with Anita Schjoll Breda.
Anita is the CEO and Co-Founder of Iris AI.
It's a start-up that Fast Company
recently named one of the most innovative
Artificial Intelligence start-ups of 2017.
She's recently also been named faculty
at Singularity University in Denmark,
and she's an alumni of the 2015 Graduate Solutions Program.
Welcome.
- Thank you.
Let's talk about Iris AI.
It's been so exciting following the growth of the company.
Tell me a bit about the application
of the AI system.
- Our ultimate goal is to build an AI Researcher.
And we're of the core belief that if one human being
can sit down and read every single research paper
every single patent in the world,
just read them all in one go, connect the dots;
we'd be able to solve a lot of problems.
We have a lot of knowledge, it's just inaccessible.
Ultimately what we're doing is building an AI
that can read and understand and connect the dots,
and all of it, for us.
But obviously zooming that back in to today,
which is what really matters, right?
What is it that we're building now is a tool
for R&D, research institutes, entrepreneurs, with big hairy
problems to solve, where you need to apply research
and science to solve it.
And we're semi-automating the process of mapping out
what you should read to solve the problem
or to see what research you need to do to solve the problem.
Basically you start with a problem statement,
take that problem statement and give it to the tool to read,
you can write out in your own words--
- You're going to copy and paste it,
or write into the system.
- Exactly.
So in your own words,
"What is the problem you are trying to solve?"
Give that to the system to read, and we map out in a visual
format what research is relevant to the challenge
you're solving.
- It's amazing.
Iris AI in many ways is about transparency of research.
Right now in technology some times there's tension
of technology and transparency.
Sometimes it feel like oil and water,
other times, with open-source trends,
it feels very homogenous.
What are your thoughts on transparency in technology?
- I think we're in an interesting spot.
Because we're both in that tech, where you talk about
open-source and do you keep things a trade secret?
Do you patent your software or do you publish all the code?
Then we're also operating in the science field
where Paywall content and Open Access research papers,
are kind of in the same ...
What do you do?
Do you pay to publish it openly,
so that anyone can access it?
We very much fall on the side of openness and transparency,
we're firm believers in that.
Especially when it comes to science.
We believe that it should be open,
it should be publicly available, especially the science
that has been paid for by our tax money;
should be openly and freely available.
We are not in a position right now to do much about that,
but the least we can do, and what we're working on,
is making sure you can at least find the right research.
Then we'll see the Open Access movement is flourishing,
more and more papers are published to Open Access,
then you have Archive and everything
that is being pre-published there,
so you have this movement of openness.
Especially in areas of exponential tech,
where this is just more and more openness.
We're firm believers of that.
We think that if we want to get humanity to the next level,
we have to, as we say, and sorry for swearing,
"Science the shit out of it."
If you want to do that you have to have it openly.
It doesn't help to have lots of knowledge
if it's all hidden behind Paywalls.
- Completely.
This Open Access, was this part of the inspiration
of founding Iris AI?
- There were a number of things.
We sat down and looked at the ten to the ninth challenge.
How could we positively impact the world?
We ended up stumbling into the academic publishing industry,
and looked at it from a number of different angles.
There's so many things that are problematic with it.
Paywall content is one of them.
The system for scoring points when you publish papers
is one of them.
It's better to split up a research project in two papers,
rather than publish it all in one,
because you'll score more points.
There's just a number of things.
The citation system is one the things where we saw that
most existing tools to navigate the research world,
or search engines if you like,
is based on the citation system that has some merit,
but when it comes to finding solutions to your problems,
the citation system works more as a popularity index,
and I know that's simplify it, but--
- Yeah
So we just believe that by ...
There's a number of issues with the whole industry,
and Open Access was one of the trends that we saw.
So as more and more research is becoming open,
how do we find it?
Right?
- Yeah.
Is there an example of positive impact that the system
has been able to make or just an interesting use case
that you love talking about?
- We have a couple.
One of them, which is very much tied to the industry,
we're focusing in on material science to start with,
just a good field to begin with.
It's cross-disciplinary by nature.
One of the challenges we've worked on
with one of our partners is, "Can you build a reusable
rocket out of composite materials?"
This is one of the ways we proved that the tool worked.
We had multiple teams compete against each other
to solve that challenge.
Can we do that?
And one of the teams was using an old school search engine,
their conclusion at the end of five hours was,
"Nope, we can't do this, technology isn't there."
The team that used our tool, they concluded
that, "It was possible."
They outlined three key papers on how it could be done,
and said it was going to be really expensive,
but it was possible.
That was very exciting.
- It's exciting to see how Artificial Intelligence systems
can extend human Intelligence.
- Definitely.
- I want to ask you, zooming out, looking at the industry
and the technology of Artificial Intelligence,
in many ways it's under heat.
Hype and heat.
- Very much so.
- What kind of responsibility do you think AI researchers,
people developing it, have to ensure they're making systems
that are going to propel humanity forward for the better?
Because we see these news headlines that feel
extremely dystopian.
- Yes.
- What are your thoughts?
- I think it's important to see that the big difference
between 25 plus years into the future, and then zooming
back into today.
You see a lot of start-ups that have these ...
And we do the same.
We have this big mission,
we're going to build an AI researcher.
But that is still 10 years into the future.
So what are we doing today?
I think it's the same when it comes
to the ethical responsibility.
Sure, we can have the discussions on utopia, dystopia.
Are we building Skynet?
Which is the simplified version of it.
I think that's too much of hype headline
that is really easy to latch on to.
I think the more pressing issues is, already today,
what data sets are you using to build your algorithms?
And you have a ton of examples on beauty pageants online,
which uses all pictures of white women.
White skinny women to judge beauty.
And then if anyone who doesn't look like that
uploads their picture to see if they're beautiful,
they're not.
Because the data set you picked isn't the right one.
You have a police department in Florida that did racial
profiling in their algorithm to assess.
I think, today, that is where a lot
of the responsibility lies.
What data are we using?
How are we making sure that we don't build-in our own biases
into the system?
I'm far more concerned about that, as more and more
automated systems comes into our everyday lives,
how do we make sure that we don't keep ...
Because we are living in a society that is incredibly
discriminating against a number of different minorities.
How can we make sure that we don't
build that into our systems?
Because suddenly also, we remove ourselves.
It wasn't me, it was the computer system.
- I didn't do that.
- I didn't do it.
But we do, because we build it into our systems.
- It's amazing hearing that, and your methodology,
because we saw the huge article that came out,
Artificial Intelligence Has A White Guy Problem,
about how it's being built.
- Exactly.
- So this is critical to be addressing it now.
- Yes.
- With Iris AI right now, it works in tandem with someone--
- Yes.
- Inputting a question.
Do you one day see Iris AI totally autonomous,
as just a machine not working in collaboration with a human?
- More and more so.
Today it's very much in collaboration,
it's an iterative process going from a problem statement,
zooming out to find a bunch of research.
Next steps that we're launching this fall
is focusing back in to figure out ...
The geeky term is the semi-automation
of the systematic landscape mapping.
But anyway, it's focusing in and that's very much
an iterative process.
Iris makes some assumptions, asks the users about
the assumptions, and we build it together.
If you look further into the future, there's going to be
more and more autonomy.
Iris can extract a hypothesis from a paper,
see all of the hypothesis'
in connection in a similarity graph, build new hypothesis'
on the top of existing, and the actually go test them
in a simulation environment, or robotic lab.
At that point you're looking at more autonomy.
So yes.
But on the other hand, it will never, never say never,
our goal isn't to press play and then Iris solves
all of the problems in the world.
Humanity in the world is complex.
I think we're always going to need
some level of human involvement.
Although, if you talk way beyond what we give
as a standard pitch, when Iris is able to figure out science
and find science, find the right theories,
extract the hypothesis, et cetera.
We can actually connect Iris to other AI's,
and teach them science.
And at that point we're starting to see
less and less human involvement.
As with anything, right now it requires
a lot of manual time.
The next version will reduce the manual labor
with about 90% for that part of the process.
So yes.
We do fall in to the category which will in fact
reduce labor time.
- Yeah.
When you talk about Iris AI being able to one day speak
with other AI systems, I get the HER image,
where Samantha starts communicating with all
the other systems.
- Right.
I don't think Iris will ever be friendly and pleasant.
It's a researcher.
Get the job done.
- What questions are hot on your mind about AI research,
or AI in general?
- Right now it's about the hype.
Are we doing ourselves ...
And it's a personal question too, we started the company
two years ago and we present ourselves as an AI company.
Iris.AI it's in our domain and our name.
The question is are we doing ourselves and the world
a disservice by positioning everything as AI?
AI for dog walking.
AI for this, AI for that.
Are we hyping it too much
so that we end up over-hyping it?
Because people are very excited about AI these days,
and I get that and there's plenty of things we can do
that are super exciting,
but then there's also the fact
that we're not quite there yet.
There's still a lot of development.
We can do the little things really well,
but the big vision, the crazy future is still years away.
So I think that's one of my concerns
is that we're over-hyping it.
And is started more and more ...
Stop talking about us as an AI company only,
but we're a company that solves important problems for R&D.
- Yeah.
And you guys just won here the Global Grand Challenge
for learning, congratulations.
- Thank you.
That's very exciting.
- I can see how you are a learning organization as well,
and so that would be one way also to position the platform?
- Yes, exactly.
We do fall in to the learning and tech space,
obviously with people who are ...
Not necessarily highly educated.
One of the effects we're seeing actually from our platform
is that it does, to a certain degree, de-skill the users,
or the requirements of the users.
You don't need to be a professor to map out the science.
And in some instances, if you do this manually,
you have to have at least an associate professor degree
or level to be able to do the full, rigorous manual process.
While with our tool, we de-skill it.
But still we're not a kindergarten tool.
You do need to know a little bit about science
or research or the field you're working in.
- But it's exciting.
Image a teacher putting it to work with their class
on their research paper.
I remember when we were in school
going through encyclopedia pages
and how much it slowed things down.
- Yes.
- So you are a female founder.
There's also a bunch of hype, misconceptions, stories
about this experience.
What have been any misconceptions that you have encountered
as a female founder?
- I think for me the thing that kind of messes with my head,
is I don't think of myself as a female founder.
I'm a founder.
I have a company to run, I have new technology to build,
we have a product to sell.
My day to day life isn't about being female.
My day to day life is about running a start-up company
and succeeding.
That's why whenever I'm ...
I won award for Inspiring 50: Women in Tech,
and I'm like, "Oh right, I'm a woman in tech.
"Right, I forgot about that."
We don't go around thinking about the fact our bodies
our genetics are the way ...
It just is.
I just happen to be female.
For me that's the biggest misconception.
This is not something that's on my mind.
While having, of course, been in situations
where I've gotten the older male engineer
patting me on the head and laughing at me.
- Thank you.
- I'm like (groans).
Yes, I've been there.
But that's not my day to day business.
- Yeah.
And it's not stopping your game either.
You are ...
- No and of course it is sad to see what's going on
in Silicon Valley and the limited amount of female partners
I meet, and the VC's I pitch to.
Of course, there is the question in the back of my head,
"Am I being judged unfairly?"
Again, when we talked about the biases in our algorithms,
it's the same thing.
And we're not aware of our biases, right?
So there is this little voice every now and then that goes,
"Is this affecting our fundraising?
"Should I put one of my male co-founders?"
Probably not, because I am the CEO of this company,
that's who they want to see.
Still, there's always the question,
"Is this impacting our fundraising?"
On the other hand, again, I have a company to run.
- Yeah.
What is fueling you often with pushing all of your work
forward, personally and with Iris AI?
- That's a good question.
I just really like what I do.
I have a lot of energy, I have a lot of passion,
I just really want to make something that matters.
I love seeing examples of our technology put to good use.
I mentioned early, we had a couple of different case studies
that are really like ...
Another one is this tiny little chocolate factory,
West Coast, U.S.
He wanted to build a sustainable product line.
New product line, sustainable, healthier chocolate.
But he's like, "But I don't have an R&D department."
And he stumbled across our tool and used that to build
a new product line.
And he was basically R&D enabling himself.
Not being highly skilled, not being a researcher--
- Not having tons of resources.
- Exactly but using our free tool that is available
on our website to do an R&D process.
And basically R&D enabling himself.
Things like that really gets me going, when we're doing
something that actually matters to people.
That people are excited about.
That's kind of what keeps me going.
And then, I don't know, I never had a real job.
I don't know what that is.
I don't know what it looks like.
I'm just really enjoying the freedom, and flexibility,
and the hard work and seeing things grow
from literally nothing, and then seeing what we've built.
(electronic music)
For more infomation >> Colorado Christian University Bans National Anthem Kneeling - Duration: 0:36. 
For more infomation >> Photography at Staffordshire University: Connect to your future - Duration: 2:19. 
For more infomation >> Student Spotlight: Strathclyde University - Duration: 1:16. 
For more infomation >> Student Spotlight: University of Aberdeen - Duration: 1:23. 
For more infomation >> Each kick for Avila University's football team raises money for childhood cancer research - Duration: 1:33. 
For more infomation >> Student Spotlight: St Andrews University - Duration: 2:07. 
For more infomation >> University of Edinburgh Coaches - Duration: 1:28. 
For more infomation >> Student Spotlight: Glasgow Caledonian University - Duration: 1:27. 
For more infomation >> Student Spotlight: University of Edinburgh - Duration: 1:52. 
For more infomation >> Student Spotlight: Glasgow University - Duration: 2:05. 
Không có nhận xét nào:
Đăng nhận xét