Welcome to the first VIP room for Boston.
Yeah, cheers.
There we go.
Our first speaker today is William Jones. He comes from the South Hampton, Utah.
Believe it or not, he loves someone half the house of strong men, so that's all really cool.
And he's recently taken up some bouldering to site really hurting someone the first time he ever did it.
I've got to say, I am very interested in this talk and please, will take it away.
Applause
Right here, right here, over here. I think I'm just trying to hear it.
Even louder.
No, that's even worse.
Over here, maybe sound, sound better. Here, let me see here.
I think it's better to be loud.
I just want to say, I'm really really interested in this talk.
I'm really curious.
I'm just wondering, but Ken is close to the door because people are coming in like so much more than he's already in his room.
I'm not sure if the door doesn't lock.
So if you were here first, you could drag and hit the little button to open it.
So if you want people to keep knocking on the door, it will lock you in.
So let's start to hear.
I think it's better to be loud.
No, I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just wondering.
I'm just curious?
I'm just wondering.
I'm just wondering.
I'm just wondering.
How well have you been talking to habenson recently.
Can you share your thoughts on y한테 Lancer Boiq announced?
Which cablebomber maybe?
I was concerned for the end of the day,
which was when Linda had talked about lived and lost.
So let's have a look at what we did today.
So we've got A, we've got schedule day at Torse,
there's a little gap between our Torse,
so possibly using it for me,
and then have a finish up,
taking a web page and just checking
if we're finding any of us involved.
We've tried very hard to keep this beginning better than we did today.
I think it's about just focusing on everybody
and all the other systems who we need to walk away with something to do.
And we also really want to keep it in the resource areas.
We've done just one or a bunch of tools,
which we like to use in AI, but we like to exhaust it.
That's the normal goal of today.
In terms of the design of the day,
we've done it the day after this is over three months.
We have some technology and technique talks at the start.
We have some of the sort of schooling talks in the middle,
slightly in there.
And we have some centre like community talks at the end.
So, that's what we're doing today.
That's five minutes from my 30th install.
So, we're the best discussion we've ever had to talk to,
said that we wouldn't be able to save
the three new productions to AI, which is very...
And I still have some innovations of why this is the one
that I want to do this.
And what I'm very interested in is the not start
with a set of innovations, something like AI,
which is a television, which is like that.
These people obviously have their own business,
but I just don't think it's much of a thing to do
or do anything at the time.
So, what I like to do is then,
or move the technical point of interest
through this, where this kind of thing is important at the moment.
So, to do that...
Yeah, well, I don't think it's this,
but I was just wondering,
so, how could we, from that to today,
look at AI and AI, which is a core of what we're doing
as we're doing this in the world?
It's not going to be what it's not going to be.
And what we're going to try to do is understand things.
But especially, what is it all about what we do
with AI and AI, which is a new giant,
that's about the large,
it's a data-all-like-large population data
from a small size of data.
And I don't think having to go to AI
is a more complicated thing to do in general,
if I give you the time,
because I want to understand how hard-working
all sorts of engineers are in the world, then.
That's quite an evil thing to do,
the population of software engineers is high,
that's quite mean.
Even if I could ask them how hard they go,
if I'm going to give them an answer,
I'll use the last thing.
So there is a fully-invented times
when we want to do this,
actually we might want to give it up now,
just a while.
It's understanding about large populations,
large size of data,
it's just a niche,
it's something we have to do in the most every case,
and it's still a problem that we're solving
with AI and AI.
And the motivation is
that sort of thing,
if you find that in the case of all the
sustainability problems of explosion,
and what this is about,
is about how
even data sets with not many innovations
not always exist in there,
will, because of the way these inventions
are defined in the total way,
and are made absolutely massive,
and they can be at a small scale.
So, let's do an example of that,
and let's see,
in the case where the man tonight
has an idea of where he needs to work on.
So, okay, I want to understand
how the artist is going to get to work on.
That's something that's unsaid
and it's difficult to measure,
because I'm going to tell you honestly.
But, what I can do is that
I need to know how you can do that
in a way that's, okay,
there's a lot of people on the internet
who have been there,
which is because they work with art,
so maybe what you do to do it
how hard the solution is to work,
is to see if you want to see those
sort of a new job,
and, okay, I can propose this
by the digital age, okay.
But, then, I need to understand
how the solution works,
and I could either guess that
as something I could do,
or I could do, say,
I'm a digital learning,
which I'm learning how the solution is.
And, if I could do that,
that's okay, that's something I could do.
I probably want to make it go on,
but then I have to do power
to get the solution up something.
So, I want to make it go on,
but if I propose this,
you can literally do a,
and how hard it would work
if I at least had something
that was not made,
or I made a problem
with what kind of a band
maybe it's something that's been
made by me,
or maybe you know what I'm saying.
That's, and all of these things
would be perfectly possible,
and they would be a huge commitment.
However, when I unfortunately go away,
and I find that shockingly
actually the most important thing
that people find is that
you're not working hard,
so you're not doing this
in a relationship to your job.
And, this is a whole line
of explaining, but I still want to
continue to invite those to do this one way.
And I'm adding in a second minute
something that might be a little bit different again,
and I've seen a lot of what the people say,
and if you do that,
I do hope that you can see
the sound of the band,
although I won't mention it,
but I think it is a different thing.
But I'm genuinely
not working with this sound of the band,
so I think you get that
through maybe the slow doing
in something that it takes time.
And the problem I might do this is that
I suddenly don't need to
find the point of doing this,
because you can't find the power
of the music in my head,
you know what I mean?
I don't just need to see
the people who are actually
using this division
and the single-boxing combinations.
I'm not doing that with the sound of the band,
which that sound is different,
but for any other music
I don't need to put the sound of the S.
So I'm going to be doing this
with people that have not been
made to perform in ten minutes,
because I don't need ten people
having a hundred people.
And I'm just going to put the ten minutes
soon on the speaker,
and I'll have to do that
with the speaker and the sound of the band.
I'm just going to put the sound of the band
in the studio.
And in my house in
the village today,
I just suddenly like,
I notice a lot of something
that's something that's changed,
something that's changed
what you think it is.
And I notice that any minute
I notice how my mother,
the things that we get,
what we get,
I never just see
98.74,
98.74,
98.74,
98.74,
and this is
what we're going to do
with a
music that
is something that we've got on the music outside.
And I think that this problem is
that
the music that's on the music outside
is something that we should
get very, very good.
And it's really remarkable.
It's really useful
to look at the music
outside the music outside the music
because the only thing we can do
is understand about the life
and the existence of the music
outside the music.
And then I notice that the music is so large
that we have to see what it is
in the art of all innovation.
And the music is so large
that we have to see what it is.
And then I notice that the music is so big.
We know
and this is something that
allows even more of our
exciting
modern music
and other modern music
that we learn about
and we use for all of us
what we do
and we use for all of us
the amount of things that we can do
to do that.
And we're going to be able to
make this music outside the music
outside the music.
And we're going to be able to make
this music outside the music
outside the music.
And we're just waiting
for the music to finish.
So, we'll just stop here.
Okay, wind, wind, wind.
This is
Yeah, and this is why
I'm issuing
this.
Suddenly so popular because
it's a massive cliché but we are in
an increasingly data-driven world
and we need the way of managing that data
and by just
the argument I've just given you, there's only one way to do it
and that is to use things like AI,
machine learning and statistics.
We also have
a nice little
a side benefit from all of this
is that AI and machine learning do allow us
to automate a set of tasks which are otherwise
very, very difficult to do.
There are many tasks like
for example the ones I just gave
where it's extremely difficult to come up
with a solution directly. For example,
if I wanted to
look at a picture
and tell you whether it has a cat or a dog in it,
that's not something like
if you put me in a room and said I couldn't leave until I'd done that
you'd come back to a skeleton 10 years later.
It's just not possible to solve that problem directly.
But to solve that problem
indirectly by
learning a solution from a set of
examples, that's something I could do.
And in fact, modern tooling is so good, I could probably do that
in five minutes recently.
It's really easy these days.
So,
where open source fits into this?
So,
we have to do our AI somehow.
We've obviously
sort of got to do it with software. It's not really a problem
that can be solved at other scales. So we have
open source software as thankfully actually
these days a very big part of AI.
And
that's good, that's great, that's
something we all know quite a lot about.
But alongside this
we also have
thank you,
we also have open source
data and open source
models. And this is
another great thing that we can share when we're doing
AI and machine learning.
We can share not just the software we use
to do things,
we can share the models that we use.
So, I posted earlier this
slightly silly
but I think it's a great thing.
So, we can share the
relationship between
what I say,
income or pay,
beard length
and seniority
and how hard people work.
And
yeah,
perfect.
Yay, cool.
Thank you.
Thank you.
Okay.
What was I saying?
Open source data.
Open source data
and open source models.
Yep, sorry.
So, as well as open source software
we have open source data and open source models.
These are,
this is another lovely extra thing we can do.
I think
Michelle is going to be talking
quite a lot about this later.
And
yeah, another
lovely, other thing we can do.
Again, it's another quite a big
cliche but obviously with all these
benefits that we get from AI and machine learning,
we get a lot of challenges as well.
We have some
big challenges at the moment.
Explainability is one of these.
We have,
I think, quite a lot of difficulty
getting our more
advanced models,
particularly to tell us
about why they work,
how they're coming to the solutions
that they do.
And this is a big
problem in
applications which directly relate to
people's lives like automotive
or healthcare
not being able to understand
why we've just run that poor, old
lady over is a big issue.
And this is a real issue.
The automotive industry at the moment has
these
wonderful self-driving cars
that are in many ways and by many
metrics are better than humans.
But
the regulators
aren't letting any of this
happen because there's no
accountability for them. There's no way of
having to behave well and there's no
way of proving
that they do what they should do.
We also
have issues of bias.
AI and machine learning models
are ultimately a reflection of the data
that they were trained on and
that can be
a problem when the data
has been
scraped from the internet.
The internet
is a very wonderful thing
but I think we all know that not all of the concepts on the internet is great and
Models that are trained on this come to reflect the biases inside of it
this has been sort of very publicly the case in a few examples and with
Hiring AI is where people have trained these on data sets
of their historical hiring practices and
This has turned out not particularly well and being quite biased for example against the women
that was quite a high profile case in the last few years and
Yeah, this is this is not a great issue either
Privacy is also a final difficult issue
With a demand for AI and machine learning models comes a demand for data and this data doesn't always respect people's privacy
as it should and
Even when it does models themselves often leak things about the data they were trained on either intentionally or unintentionally
Anyway, that is the end
I think I did it quite early which should give us plenty of time to fix things and switch over
Do we have any questions?
If you can raise your hand
I
So it was a bit difficult for me to hear that
What
I'd not to pass the buck here, but I actually think our next speaker's got quite a lot to say about that might be
Really more
Again it was a little difficult to hear that was that security
Yeah, it's it's a tricky thing I
Think the main difficulty here is that we do know how to do security well in some respects with things like differential privacy
but it's what cost we put what price we pay to do this and
Doing security and also having things run at the speed they want to we want them to go out
That's that's a tricky thing. Just answer your question
It's just that I see a lot of focus on
Yeah, no, that's I mean, I think that's a perfectly fair comment
Honestly people do really focus on these three things and it's really tempting to well like I just did
Focus on things like differential privacy
Information leakage because they're interesting, but yeah basic cyber security really
Should be the first port of call for these things