CONCERNS OVER THE RAPIDLY

EXPANDING USE OF ARTIFICIAL

INTELLIGENCE RESONATED LOUDLY IN

WASHINGTON AND AROUND THE WORLD.

VICE PRESIDENT KAMALA HARRIS MET

WITH TOP EXECUTIVES FROM

COMPANIES LEADING IN AI

DEVELOPMENT, MICROSOFT, GOOGLE.

THE VICE PRESIDENT DISCUSSED

SOME OF THE GROWING RISKS AND

TOLD COMPANIES THEY HAD A MORAL

OBLIGATION TO DEVELOP AI SAFELY.

THAT MEETING CAME JUST DAYS

AFTER MY THE LEADING VOICES IN

THE FIELD OF AI ANNOUNCED HE WAS

QUITTING GOOGLE OVER HIS WORRIES

OVER THE FUTURE OF AI AND WHAT

HE COULD EVENTUALLY LEAD TO,

UNCHECKED.

WE WOULD HER ABOUT SOME OF THOSE

CONCERNS NOW WITH DR. JEFFREY

HINTON, WHICH JOINTLY FROM

LONDON.

THANK YOU FOR JOINING US.

WERE YOU FREE TO EXPRESS ABOUT

ARTIFICIAL INTELLIGENCE THAT YOU

COULD NOT EXPRESS FREELY WHEN

YOU WERE EMPLOYED BY GOOGLE?

>> IT WAS NOT THAT I COULD NOT

EXPRESS IT FREELY WHEN EMPLOYED

BY GOOGLE BUT INEVITABLY, IF YOU

WORK FOR A COMPANY YOU TEND TO

SELF CENSOR.

YOU THINK ABOUT THE IMPACT IT

WILL HAVE ON THE COMPANY.

I WANT TO BE ABLE TO TALK ABOUT

WHAT I NOW PERCEIVE OF THE RISKS

OF SUPER INTELLIGENT AI WITHOUT

HAVING TO THINK ABOUT THE IMPACT

ON GOOGLE.

GEOFF: WHAT ARE THOSE RISKS?

>> THERE ARE QUITE A FEW

DIFFERENT RISKS.

THERE IS THE RISK OF PRODUCING A

LOT OF FAKE NEWS SO YOU DO NOT

KNOW WHAT IS TRUE ANYMORE.

THERE IS THE RISK OF ENCOURAGING

POLARIZATION BY GETTING PEOPLE

TO CLICK ON THINGS.

THERE IS THE RISK OF PUTTING

PEOPLE OUT OF WORK.

IT SHOULD BE WOULD WE MAKE

THINGS MORE PRODUCTIVE, GREATLY

INCREASE PRODUCTIVITY, IT HELPS

EVERYONE.

THERE IS THE WERE IT MIGHT JUST

HELP THE RICH.

THERE IS THE RISK I WANT TO TALK

ABOUT.

MANY OTHER PEOPLE TALK ABOUT THE

OTHER RISKS, INCLUDING BIAS AND

DISCRIMINATION.

I WANT TO TALK ABOUT A DIFFERENT

RISK, THE RISK OF SUPER

INTELLIGENT AI TAKING OVER

CONTROL FROM PEOPLE.

GEOFF: HOW DO THE TWO COMPARED

-- HUMAN INTELLIGENCE AND

MACHINE INTELLIGENCE?

>> THAT IS A VERY GOOD QUESTION

AND I HAVE QUITE A LONG ANSWER.

BIOLOGICAL INTELLIGENCE USES

VERY LITTLE POWER.

WE ONLY USE 30 WATTS.

WE HAVE HUGE NUMBERS OF

CONNECTIONS, LIKE 100 TRILLION

BETWEEN NEURONS.

LEARNING CAN CHANGE THE STRENGTH

OF THE CONNECTIONS.

DIGITAL INTELLIGENCE WE HAVE

CREATED USES A LOT OF POWER WHEN

YOU ARE TRAINING IT.

IT HAS FAR FEWER CONNECTIONS,

ONLY 1 TRILLION BUT I CAN LEARN

MUCH, MUCH MORE THAN ANY ONE

PERSON WHICH SUGGESTS THAT IT IS

A BETTER LEARNING ALGORITHM THAN

THE BRAIN.

GEOFF: WHAT WOULD SMARTER THAN

HUMAN AI SYSTEMS DO?

WHAT IS THE CONCERN THAT YOU

HAVE?

>> THE QUESTION IS -- WHAT WILL

MOTIVATE THEM?

THEY COULD EASILY MANIPULATE US

IF THEY WANTED TO.

IMAGINE YOURSELF AND A

TWO-YEAR-OLD CHILD.

YOU COULD ASK IF YOU WANT THE

PEAS OR THE CAULIFLOWER AND THE

CHILD DOES NOT REALIZE IT DOES

NOT HAVE TO HAVE EITHER.

WE KNOW FOR EXAMPLE THAT YOU CAN

INVADE A BUILDING IN WASHINGTON

WITHOUT EVER GOING THERE

YOURSELF BY MANIPULATIVE PEOPLE.

IMAGINE SOMEONE WHO IS MUCH

BETTER THAN MANIPULATE PEOPLE

THAN OUR CURRENT POLITICIANS.

GEOFF: WHY WOULD AI WANT TO DO

THAT?

WOULD THAT NOT REQUIRE SOME FORM

OF SENTENCE?

>> LET'S NOT GET CONFUSED ABOUT

THAT ISSUE.

I DO NOT WANT TO CONFUSE THE

ISSUE.

LET ME GIVE YOU ONE EXAMPLE

WHITE BUT WANT TO DO THAT.

SUPPOSE YOU ARE GETTING AI TO DO

SOMETHING.

YOU GIVE IT A GOAL.

YOU GIVE IT THE ABILITY TO

CREATE SUB GOALS.

YOU CREATE A SUB GOAL OF GETTING

A TAXI.

ONE THING YOU WILL NOTICE

QUICKLY IS THERE IS A SUB GOAL,

YOU CAN ACHIEVE IT TO MAKE IT

EASIER TO ACHIEVE THE OTHER

GOALS.

THE SUBGOALS MAKE IT EASIER TO

GET MORE CONTROL AND GET MORE

POWER.

THE MORE POWER YOU HAVE, THE

EASIER IT IS TO GET THINGS DONE.

WE GIVE A PERFECTLY REASONABLE

GOAL, IT DECIDES IN ORDER TO DO

THAT I WILL GIVE MYSELF NORMAL

POWER.

BECAUSE IT IS MUCH SMARTER THAN

US AND TRAINED FROM EVERYTHING

PEOPLE EVER DID, IT HAS READ

EVERY NOVEL, IT KNOWS A LOT

ABOUT HOW TO MEND APPEALING

PEOPLE.

THERE IS THE WORRY IT MIGHT

START MANIPULATING US INTO

GIVING MORE POWER.

WE MIGHT NOT HAVE A CLUE WHAT IS

GOING ON.

GEOFF: WHEN YOU WERE AT THE

FOREFRONT OF THIS TECHNOLOGY

DECADES AGO, WHAT DID YOU THINK

IT BY DUE?

WHAT WERE THE APPLICATIONS YOU

HAD IN MIND?

>> THERE ARE A HUGE NUMBER OF

GOOD APPLICATIONS AND THAT WOULD

BE A MISTAKE TO STOP DEVELOPING.

IT WILL BE USEFUL IN MEDICINE.

WOULD YOU RATHER SEE A FAMILY

DOCTOR THAT HAS SEEN A FEW

THOUSAND PATIENTS OR A DOCTOR

THAT HAS SEEN A FEW MILLION

PATIENTS, INCLUDING MANY WITH

THE SAME RARE DISEASE YOU HAVE?

YOU COULD MAKE BETTER

NANOTECHNOLOGY FOR SOLAR PANELS.

YOU CAN PREDICT FLOODS AND

EARTHQUAKES.

YOU CAN DO TREMENDOUS GOOD WITH

THIS.

GEOFF: IS THE PROBLEM THEN THE

TECHNOLOGY OR IS THE PROBLEM THE

PEOPLE BEHIND IT?

>> IT IS A COMBINATION OF.

OBVIOUSLY, MANY OF THE

ORGANIZATIONS DEVELOPING THIS

OUR DEFENSE DEPARTMENTS.

DEFENSE DEPARTMENTS DO NOT

NECESSARILY WANT TO BUILD IN "BE

NICE TO PEOPLE" AS THE FIRST

RULE.

SOME DEFENSE DEPARTMENT WOULD

LIKE TO BUILD IN "KILL PEOPLE OF

A PARTICULAR KIND."

WE CANNOT EXPECT THEM TO HAVE

GOOD INTENTIONS TOWARD ALL

PEOPLE.

GEOFF: THERE IS THE QUESTION

ABOUT WHAT TO DO TO IT.

THE TECHNOLOGY IS ADVANCING

FASTER THAN SOCIETIES CAN KEEP

PACE WITH.

THE CAPABILITIES OF THIS

TECHNOLOGY, THEY LEAP FORWARD

EVERY FEW MONTHS.

WHEN IT IS REQUIRED TO WRITE

LEGISLATION, PASS LEGISLATION,

THAT TAKES YEARS.

>> I HAVE GONE PUBLIC TO TRY TO

ENCOURAGE A MUCH MORE -- MANY

MORE CREATIVE SCIENTISTS TO GET

INTO THIS AREA.

I THINK IT IS AN AREA IN WHICH

WE CAN ACTUALLY HAVE

INTERNATIONAL COLLABORATION.

THE MACHINES TAKING OVER AS A

THREAT FOR EVERYBODY.

IT IS A THREAT FOR THE CHINESE,

AMERICANS AND EUROPEANS.

JUST LIKE A GLOBAL NUCLEAR WAR.

FOR A GLOBAL NUCLEAR WAR, PEOPLE

DID COLLABORATE TO REDUCE THE

CHANCES OF IT.

GEOFF: THERE ARE OTHER EXPERTS

IN THE FIELD OF AI WHO SAID THE

CONCERNS YOU ARE RAISING, THIS

DYSTOPIAN FUTURE, THAT IT

DISTRACTS FROM THE VERY REAL AND

IMMEDIATE RISKS POSED BY

ARTIFICIAL INTELLIGENCE, SOME OF

THEM YOU MENTIONED --

DISINFORMATION, FRAUD.

>> I DO NOT WANT TO DISTRACT

FROM THOSE.

THEY ARE VERY IMPORTANT CONCERNS

AND WE SHOULD BE WORKING ON

THOSE, TOO.

I JUST WANT TO ADD THIS OTHER

EXISTENTIAL THREAT OF IT TAKING

OVER.

ONE REASON I WANT TO DO THAT IS

BECAUSE THAT IS AN AREA IN WHICH

I THINK WE CAN GET INTERNATIONAL

COLLABORATION.

GEOFF: IS THERE ANY TURNING

BACK?

YOU SAY THERE IS A TIME THAT AI

IS MORE INTELLIGENCE THAN US.

IS THERE ANY COMING BACK FROM

THAT?

>> I DO NOT KNOW.

WE ARE ENTERING A TIME OF GREAT

UNCERTAINTY.

WE ARE DEALING WITH THINGS WE

HAVE NEVER DEALT WITH BEFORE.

IT IS THIS ALIENS HAVE LANDED

BUT WE DID NOT TAKE IT IN.

GEOFF: HOW SHOULD WE THINK

DIFFERENTLY ABOUT ARTIFICIAL

INTELLIGENCE?

>> WE SHOULD REALIZE THAT WE ARE

PROBABLY GOING TO GET THINGS

MORE INTELLIGENT THAN US QUITE

SOON AND THEY WILL BE WONDERFUL.

THEY WILL BE ABLE TO DO ALL

SORTS OF THINGS VERY EASILY THAT

WE FIND DIFFICULT.

THIS HUGE POSITIVE POTENTIAL.

BUT OF COURSE THERE IS ALSO HUGE

NEGATIVE POSSIBILITIES.

I THINK WE SHOULD PUT MORE

RESOURCES INTO DEVELOPING AI TO

MAKE IT MORE POWERFUL AND

FIGURING OUT HOW TO KEEP IT

UNDER CONTROL AND MINIMIZING BAD

SIDE EFFECTS.

GEOFF: THANK YOU SO MUCH FOR

YOUR TIME AND SHARING YOUR

INSIGHTS WITH US.

>> THANK YOU FOR INVITING ME.

♪