Transcriber: Reihaneh Iranmanesh
Reviewer: Elisabeth Buffard When most people think about AI, they picture a sci-fi dystopian future,
with man versus machine. Terminator, Skynet, Black Mirror,
Blade Runner, Westworld. But as someone who is working on the most ambitious
AI projects in the world, every day, I can tell you
that is far from reality. To me, it’s the contrary of that. AI enhances
and even supercharges humanity. Let me explain why. There are many reasons
why AI will never replace humans. AI always has,
and always will, rely on humans. That’s one of the reasons that I was actually inspired
to start an AI company. That and my background
have had a huge impact on me and why I started Scale. My parents were brilliant scientists
of Los Alamos, who accomplished a lot
in advancing their field. That inspired me to use
science and technology to have a real impact on the world. My dad was a physicist,
and my mom was an astrophysicist, both at the top of their field, who made meaningful contributions
to plasma fluid dynamics and the beginnings of the universe. Their scientific work will have
meaningful impact on how we understand
and perceive our world. And I wanted to work
on something as impactful or even more impactful than that. That’s why I decide
to become a programmer, so I can make a difference in the world. Growing up as a programmer, despite how powerful computers are, you quickly realize
how limited they are. In particular, they lack
judgment and intelligence. Programming is the art
of giving clear robotic instructions to computers to accomplish
simple objectives. It’s all black and white,
and there’s no gray area. As I learned about AI,
it was clearly transformational. It changed the game. It equipped computers with intelligence,
and I knew I wanted to be deeply involved. I was studying AI at MIT and slowly
became more and more excited about all the potential applications of AI
for solving more nuanced problems. For example, there was one class project where I worked on applying AI
to human emotions. The goal was to take picture
of human expressions and ultimately identify and understand
the emotion through very subtle signals in facial expressions. Using AI, we built an algorithm that was able to detect intent
with 80% accuracy and efficacy. We were extremely proud of that. It was the start of using AI to do
entirely new things using computers. That’s when I realize
the implications of AI and how it could tackle the gray areas
that involve judgment or intelligence. You see, AI needs humans
to teach it individual values, nudge it to find thoughtful outcomes, and ensure that human intentions
and values are aligned with the AI. It was a revelation. Before, coding was like
a black-and-white film versus watching in technicolor. What’s more, AI has the potential
to take away the repetition in our lives, meaning that new and fresh ideas
will matter more and ultimately enable us to be more human. But, to power AI,
you need powerful data, which was especially hard to come by
at that time, in 2016, while I was at MIT. I realized that nobody was building
anything with AI outside of school. It’s unusual for MIT students
to not be building something. Mechanical engineering majors
are building catapults in the lawn, electrical engineering majors
are building robots, and computer science majors are building
apps for their friends to use. But nobody was building
anything using AI. That’s when I discovered what
a bottleneck data can be to building meaningful
and powerful AI systems. You can't treat data as an afterthought. Bad data or lack of data
results in bad AI. I even realize this in my personal life. I put a camera inside my fridge
to gather data, to tell me when to refill my groceries
and what I needed to buy. That’s when I realized
just how much data I needed to actually be able
to successfully predict what to purchase. There’s no way I could create enough data to be successful
with the algorithms on my own. But it did help me discover
that my roommate was stealing my food. (Laughter) At that point, I realized that this was going to be
a pivotal problem for AI. Building large-scale,
high-quality datasets to power every single application. This was the impetus
behind starting Scale: quality data, to create
reliable AI outcomes, requires human insight and guidance. If you think about the core setup of AI,
the algorithms need data, and data needs humans. To ensure data is accurate, an expert human is often required. Only humans can understand
the context and nuance to properly annotate
the data to be fed to algorithms. Humans are the one who teach
the algorithms what to do. They’re the ones making the decisions,
they guide them. If something happens,
here’s what you should do. And AI learns from that
and replicates it. We are teaching the AI
our individual values and nudging the AI
to find thoughtful outcomes. Machines make mistakes. We have to teach them and
incentivize them to tell the truth. This is why teaching the AI human
intentions and values is so important. It’s through this process
that we will ensure that AI will have fair, ethical outcomes
in line with human values. It’s this alignment
that we must solve for. The constant alignment of AI to human
intentions will always require humans. and human ideas and creativity
can actually matter much more, with the power of AI behind them. The long tale of real-world problems, and the fact that there’s always
unknown unknowns means that humans
will never be fully removed from the AI development lifecycle. For example, I remember back in 2016 when chatbots were first starting
to become a big thing. It was right when we were starting Scale. We were all thinking there's no way
to build a fully automated system. There’re so many different conversations
that can have so many different pathways. It’s hard to build AI systems that can
properly handle all these possibilities. For chatbots to work, there’re humans
behind it who make the decisions once, and from there, the chatbots
can replicate that over and over again. That’s again why it’s impossible
for AI to improve or change without human input. Let’s take you to the front lines of AI. The things that AI automates first
are not what you might expect. An unintuitive example is the weather. Humans have tried for many millennia to crack the code
of how to predict the weather. It’s especially hard for meteorologists because there are
so many different small things that can cause massive impacts
on the weather. It's the butterfly effect. Different elements react to one another
in unexpected ways. There’re so many inputs
that affect the weather, way more data than any person
would be able to comprehend on their own. That’s why we need AI
to analyze the vast oceans of data and provide more accurate, nuanced,
and comprehensive analysis. At the moment, AI can already provide extremely accurate
short-term predictions, including for critical storms and floods. So, it’s not what humans perceive
to be the simplest task that AI will automate first,
but rather where we have the most data. The use cases the brightest minds
are focusing on are much more positive
than what you might think. Much more so than Terminator or Westworld. That’s again why I think AI
will be a supercharger for humanity. Unlike the movies, AI developers
aren’t focusing their attention on building replacements for humans. They’re building tools
to help free up our time and energy to focus on what human
can uniquely solve. A good example about how AI
can be used in practice is health care. According to the Association
of American Medical Colleges, the United States could see
an estimated shortage of between 38,000
and 124,000 physicians by 2034. AI could save doctors’ time
with rogue tasks and ultimately enable them to serve
more patients and help more people. Health care is full of repetitive tasks
which are right for AI. When a patient is sick,
they go through all kinds of tests which produce all sorts of data:
blood tests, imagery, lab results, X-rays, etc. Doctors then analyze all this data
to make decisions about a case. AI can analyze all this data proactively and go through a list of possibilities by cross-referencing
against all prior data in cases. It can identify when something isn’t right
long before a physician can and flag it to a physician,
if it requires more attention. With AI,
doctors are still integral to the process, but it takes less time to get a diagnosis. You have to wait several weeks for your case to go
from one doctor to another. The AI will supercharge,
finding a diagnosis faster. Similarly, in the field of drug discovery,
it’s all about using complex data: experiment data, patient data, protein simulations and far more to guide a more efficient process of solving diseas
through new drugs and compounds. Recent advancements in AI have dramatically sped up
the scientific process by allowing us to process and make us
of more data than ever before. Another good example, and potentially
more concrete, is the Russia-Ukraine war. We've all seen the images of tanks
lining up ready to enter Kiev. AI can help assess satellite imagery
with superhuman speed and precision, so Ukrainian forces can respond faster. At Scale, we’re using our platform
to do damage assessment in key areas affected by the war. We’ve rapidly analyzed
over 2,000 square kilometers of Ukraine, identifying over 370,000 structures, including thousands not previously
available via other datasets. We focused on Kiev, Kharkiv and Dnipro, in which we provided some data
directly to government and users. We also made the data publicly available
to the broader AI community via Scale. We can also use this data
to maximize allocations of resources, people or commodities. It’s clear satellite data
can be extremely useful in these types of situations. Thanks to satellite data,
AI can analyze if planes or tanks have been moved from one place to another. This is called change detection. Algorithms can constantly be monitoring
for this kind of data, and if it notices a change or movement, it will alert a human
to further investigate, otherwise known as predictive modeling. AI can also help us understand
the economic impacts of war. We can use AI to track
farmland in Ukraine and measure the agricultural damage
in real time. Ukraine is a major food supplier
for much of the world. Understanding these impacts
is absolutely critical. In conclusion, AI is not
something to be feared, but it’s a tool that can be used
to better understand… that needs to be better understood, and has the potential
to transform our lives for the better. AI enables us to make use and sense
of massive amounts of data that has historically been
beyond human capacity. It allows us to add
intelligence and nuance to automated systems
that will dramatically improve humanity. Areas like health care and agriculture. This then allows humans
to do what they do best. Take this information,
put it into context with sensitivity, to strategize and act in a timely manner. AI is a supercharger for humanity. When AI is better than humans, it makes humans better. AI will automate repetitive tasks that don’t require
constant human judgment or creativity, which frees us up to explore
and focus on fresher, newer ideas. AI will enable us to be
even more creative and more idea-driven, which I personally find
incredibly exciting. It allows us to embrace
the generative aspects of human nature, so we can run faster with ideas and build better
and more powerful solutions to the world’s biggest problems. That’s why I believe
that human-led AI is the path forward, and I’m proud to usher all of us
into a future with human-led AI. Thank you. (Applause)