- Tubelator AI
- >
- Videos
- >
- Science & Technology
- >
- Exploring the Future of Generative AI & Computer Vision with MIT Professor: Part 2
Exploring the Future of Generative AI & Computer Vision with MIT Professor: Part 2
Discover the optimistic perspective of a MIT professor on integrating AI with human-like intelligence into society in a positive manner. Learn about the potential for powerful AI systems to coexist and interact with us in a valuable way.
Install Tubelator On Chrome
Video Summary & Chapters
No chapters for this video generated yet.
Video Transcript
(logo whooshes)
(text chitters)
- So next we have, "How do
you weigh the social balances
of equipping AI with
human-like intelligence
and how does your work aim
to integrate AI into
society in a positive way?"
Ultimately, I'm an optimist.
I think that there is a positive future
for AI integrating into society.
I have had the dream since
childhood of, you know, robots
that are human-like that interact with us.
I really think there can be a positive
and valuable future
with powerful AI systems
living alongside us.
However, it is definitely true
that there are a lot of ways
that this could go wrong.
So my own current approach
to dealing with that kind
of trade off is to say,
I wanna focus on
understanding these systems,
not necessarily being the
one who's going to deploy
and make the big systems.
So I get to kind of step back
and just act as a scientist
and try to understand,
what are the principles of intelligence?
How do current AI systems work?
What are they good at,
what are they not good at?
And that's a position that I quite enjoy
and I'm quite comfortable with
because I think that
understanding these things
and developing the science of these things
is just going to pay off for society.
It will allow us to make good choices
about how to integrate
them into the world.
But that is going to be
a hard policy decision
later on in the future.
What do you see as the key challenges
for generative video models
and what other emerging trends
in generative AI pique my interest?
Yeah, generative video models
are really, really cool.
They're the current
thing that's taking off.
So there's maybe two big challenges here.
The first challenge is
that videos are really big.
They are a lot of frames.
So imagine if it's hard
to make a single image,
well, a video is like thousands of frames,
thousands of images.
So the amount of compute
and the amount of memory
and resources required to generate a video
is just a lot more than to
generate a single photo.
