>>> NOW, IT'S BEEN CALLED THE SPACE RACE OF THE 21st CENTURY.

BUT THIS TIME IT'S ALL ABOUT ARTIFICIAL INTELLIGENCE.

PRESERVE'S AMERICA'S LEAD OVER CHINA HAS BECOME A TOP PRIORITY OF PRESIDENT TRUMP'S SECOND TERM.

BUT AS THE TWO NATIONS VIE FOR INFLUENCE, THE QUESTION ISN'T JUST WHO GETS THERE FIRST BUT WHAT KIND OF FUTURE THEY CREATE IN THE PROCESS.

FORMER NATIONAL SECURITY ADVISER JAKE SULLIVAN JOINS WALTER ISAACSON TO BREAK DOWN EXACTLY HOW AMERICA CAN KEEP COMPETING WHILE MANAGING THE RISKS.

>> THANK YOU, BIANNA.

AND JAKE SULLIVAN, WELCOME BACK TO THE SHOW.

>> THANK YOU FOR HAVING ME, WALTER.

IT'S GOOD TO SEE YOU.

>> UP AT HARVARD THERE YOU TEACH A COURSE IN THE GEOPOLITICS AND THE AGE OF ARTIFICIAL INTELLIGENCE, AND YOU WROTE SOMETHING FOR "FOREIGN AFFAIRS" WITH THAT TITLE.

WITH TAL FELDMAN, WHO WORKED ON AI WHEN YOU WERE IN THE BIDEN ADMINISTRATION.

IT HAS EIGHT SCENARIOS OF HOW WE'RE GOING TO COMPETE WITH CHINA IN THE AGE OF ARTIFICIAL INTELLIGENCE.

BUT LET ME START WITH WHAT WOULD BE MAYBE THE MOST LIKELY BUT ALSO THE MOST WORRISOME ONE.

WHICH IS WHAT IF AI REALLY IS GOING TO GET US TO ARTIFICIAL GENERAL INTELLIGENCE AND WHAT IF IT'S HARD TO CATCH UP IF SOMEBODY ELSE BEATS YOU THERE AND CHINA'S GOING HELLBENT FOR IT?

WHAT SHOULD OUR POLICY BE?

>> WELL, THAT'S THE FIRST WORLD WE LAY OUT, THAT IN FACT WE ARE JUST A SHORT TIME AWAY FROM A TRUE SUPERINTELLIGENCE THAT IS A CAPACITY OF VERY POWERFUL AI THAT HAS SOMETHING CALLED RECURSIVE SELF-IMPROVEMENT WHERE THE AI SYSTEM ITSELF IS INVENTING NEW THINGS, DISCOVERING NEW CAPABILITIES, UNLOCKING NEW CAPACITIES IN THE ECONOMIC AND NATIONAL SECURITY DOMAIN THAT THE FIRST COUNTRY THAT GETS THERE IN A HARD TO CATCH UP SCENARIO WOULD HAVE AN ENDURING ADVANTAGE AND THAT THE U.S.

AND CHINA ARE BOTH PELLMEL IN A RACE TO DO IT.

THAT'S SCENARIO ONE.

THERE ARE A LOT OF PEOPLE INCLUDING THE HEADS OF MAJOR AMERICAN AI LABS WHO WOULD SAY THIS IS THE WORLD WE'RE LIVING IN.

IN THAT WORLD YOU DON'T QUITE HAVE WINNER TAKE ALL BUT YOU WOULD HAVE WINNER TAKE A HECK OF A LOT.

BOTH IN TERMS OF THE ABILITY OF A COUNTRY TO GET REAL NATIONAL SECURITY ADVANTAGE IN TERMS OF THEIR MILITARY AND INTELLIGENCE ABILITIES, REAL ECONOMIC ADVANTAGE IN TERMS OF THE ECONOMIC GROWTH AND PRODUCTIVITY THEY WOULD UNLOCK, AND REAL SCIENTIFIC AND TECHNOLOGY ADVANTAGE BECAUSE AI INVENTIONS CAN PRODUCE INVENTIONS IN EVERY OTHER FIELD, WHETHER IT'S BIOTECHNOLOGY OR CLEAN ENERGY OR YOU NAME IT.

SO THE STAKES ARE INCREDIBLY HIGH IN THAT WORLD.

AND IT PLACES A PREMIUM ON MAKING THE INVESTMENTS IN COMPUTING POWER AND TALENT AND ELECTRICITY TO POWER DATA CENTERS BUT IT ALSO PLACE S A HIGH PREMIUM ON SAFETY AND A CONVERSATION BETWEEN THE U.S.

AND CHINA TO MAKE SURE THIS TECHNOLOGY DOESN'T GET AWAY FROM US IN WAYS THAT HARM HUMANS IN BOTH OUR COUNTRIES AND IN EVERY OTHER COUNTRY.

>> SO YOU TALK ABOUT A CONVERSATION ON SAFETY THAT SEEMS TO HARKEN BACK ALMOST 80 YEARS TO WHEN NUCLEAR WEAPONS CAME AND WE SAID WE'RE GOING TO HAVE TO HAVE NUCLEAR WEAPONS ARMS TALKS.

DO YOU THINK WE'LL HAVE TO DO THAT IN THIS WORLD WITH CHINA ON AI?

>> WE ARE ABSOLUTELY GOING TO HAVE TALKS THAT ARE AS SERIOUS, AS TECHNICAL, AS SUSTAINED AS WE DID IN THE COLD WAR EXCEPT FOR THE STAKES ARE EVEN HIGHER BECAUSE THE IMPACT GOES BEYOND JUST ONE CLASS OF WEAPONS AND THE PROBLEM IS EVEN HARDER.

BECAUSE IT'S MUCH EASIER TO COUNT MISSILES AND COUNT WARHEADS THAN IT IS TO DETERMINE AI CAPABILITY.

AND IT IS MUCH EASIER TO VERIFY THROUGH INSPECTIONS AND OVERFLIGHTS AND THE OTHER METHODS WE BUILT UP THAT THE OTHER COUNTRY IS IN COMPLIANCE WITH WHATEVER AGREEMENT THEY REACH.

BUT I'LL TELL YOU, WALTER, IN THE BIDEN ADMINISTRATION I HAD A SERIES OF VERY HIGH-LEVEL ENGAGEMENTS WITH MY CHINESE COUNTERPART.

AND TOGETHER WE TEED UP FOR OUR TWO PRESIDENTS A PROPOSAL THAT THE TWO PRESIDENTS, XI JINPING AND JOE BIDEN, ACTUALLY DIRECT TALKS ON AI RISK AND SAFETY MANAGEMENT BY THE U.S.

AND CHINA.

WE HAD A FIRST SESSION IN GENEVA, THE PLACE OF MANY OF THOSE COLD WAR DISCUSSIONS ON NUCLEAR WEAPONS IN THE SUMMER OF 2024.

AND I WOULD ARGUE THAT DONALD TRUMP AND XI JINPING SHOULD REACH A SIMILAR AGREEMENT THIS YEAR AND GET THEIR TEAMS WORKING ON THIS ISSUE.

>> WHAT DID YOU TEE UP AND HOW COULD THAT BE APPLIED DURING THE TRUMP ADMINISTRATION?

>> IN THEIR LAST MEETING XI JINPING AND JOE BIDEN MADE A SIMPLE AGREEMENT.

THEY AGREED THAT THE U.S.

AND CHINA AS A MATTER OF POLICY WOULD ENSURE THAT THE DECISION TO USE NUCLEAR WEAPONS WOULD NEVER BE HANDED OVER TO AN AI, THAT IT WOULD REMAIN IN THE HANDS OF HUMANS.

NOW, AT ONE LEVEL THAT SEEMS PRETTY DAMN STRAIGHT FORWARD.

THAT SHOULD DEFINITELY BE THE CASE, SHOULD BE EASY TO AGREE TO.

ON ANOTHER LEVEL THAT TOOK US NEARLY A YEAR TO NEGOTIATE BECAUSE IT'S THE FIRST TIME THAT CHINA HAS ACTUALLY MADE AN AGREEMENT WITH THE UNITED STATES ON AI NUCLEAR WEAPONS IN OUR HISTORY.

AND FURTHERMORE, IT SHOWS YOU THAT THESE KINDS OF AGREEMENTS ARE POSSIBLE.

SO EVEN THOUGH THAT WAS RELATIVELY LOW-HANGING FRUIT MY ARGUMENT IS TAKE THAT AND BUILD ON IT AND WE WILL SEE IN 2026 MULTIPLE SUMMIT MEETINGS BETWEEN DONALD TRUMP AND XI JINPING AND I BELIEVE THAT NEAR THE TOP OF THE AGENDA SHOULD BE FURTHER STEPS TO REDUCE THE OVERALL LEVEL OF AI RISK BY HAVING THE TWO COUNTRIES WHO, YOU KNOW, HAVE THE MOST CAPABILITY WITH THIS TECHNOLOGY MAKING FURTHER AGREEMENTS.

>> AMONG THE EIGHT SCENARIOS YOU PUT IN YOUR "FOREIGN AFFAIRS" PIECE THERE'S ONE THAT'S A LITTLE BIT LESS SCARY FOR US WHICH IS THAT IF SOMEBODY GETS AHEAD IN THE AI RACE IT WOULDN'T BE REALLY THAT HARD FOR THE OTHER COUNTRY TO CATCH UP.

IT'S NOT SORT OF AN EXPONENTIAL THING.

DID THE ADVENT OF DEEP SEEK IN CHINA, WAS THAT A DATA POINT THAT THAT MIGHT BE THE CASE FOR THE WORLD?

>> SO FOR YOUR VIEWERS, YOU KNOW, DEEP SEEK AT THE BEGINNING OF LAST YEAR CAME OUT WITH A MODEL THAT REALLY MOVED CHINA'S CAPABILITIES FORWARD.

AND IT CAUGHT --THE WORLD TO A CONSIDERABLE EXTENT BY SURPRISE BECAUSE PEOPLE THOUGHT CHINA WAS MUCH FURTHER BEHIND.

THE REALITY, WALTER, IS THAT INSIDE THE U.S.

GOVERNMENT WE ACTUALLY HAD BEEN FOLLOWING DEEP SEEK FOR WELL MORE THAN A YEAR BEFORE THAT MODEL CAME OUT.

MY TEAM WAS PRODUCING MEMOS FOR ME ON A REGULAR BASIS SAYING HEY, WATCH THIS REALLY INTERESTING COMPANY IN CHINA.

THEY'RE DOING SOME SERIOUS AND CREDIBLE WORK TO ADVANCE THE CAPABILITIES OF THEIR MODELS.

NOW, THE REALITY IS U.S.

MODELS ARE STILL AHEAD OF CHINESE MODELS.

BUT IF YOU GO TALK TO THE AMERICAN AI LABS, THEY WILL TELL YOU THAT A BIG PROBLEM IS SOMETHING CALLED DISTILLATION, WHICH IS BASICALLY A FANCY WAY OF SAYING THAT ONCE A U.S.

COMPANY HAS PUT A MODEL OUT THERE IN THE WORLD CHINESE COMPANIES CAN TAKE A LOOK AT THAT, RUN SOME TESTS ON IT AND PRETTY RAPIDLY FOLLOW WITH SOMETHING FAIRLY CLOSE IN CAPABILITY.

THAT'S NOT JUST TRUE OF DEEP SEEK.

IT'S TRUE OF OTHER CHINESE LABS AS WELL.

AND IF THAT REMAINS THE PATTERN GOING FORWARD, IF ANY ADVANCE BY ONE COMPANY CAN BE RAPIDLY FOLLOWED BY ANOTHER COMPANY, WHETHER IT'S A U.S.

COMPANY OR CHINESE COMPANY, THEN WE'RE IN MUCH LESS OF A WINNER-TAKE-ALL WORLD.

IT ALSO HAS IMPLICATIONS FOR HOW MUCH YOU INVEST IN THE MASSIVE TRAINING RUNS AT THE FRONT END.

AND IT'S SOMETHING THAT I THINK THE AMERICAN AI COMPANIES ARE GOING TO HAVE TO LOOK VERY CAREFULLY AT BECAUSE THIS IS A TOTALLY CREDIBLE SCENARIO THAT IS MUCH EASIER TO FOLLOW ONCE AN INNOVATION HAPPENS THAN MANY PEOPLE EXPECT.

>> A LOT OF WHAT YOU TALK ABOUT IN THE ARTICLE TENDS TO BE DIGITAL AI, INCLUDING THINKING THINGS.

THERE'S ALSO REAL WORLD AI.

AUTONOMOUS ROBOTS, AUTONOMOUS CARS, AUTONOMOUS DRONES, PHYSICAL OBJECTS.

IS THAT A SEPARATE CATEGORY THAT WE HAVE TO TALK ABOUT?

>> WELL, IT DEPENDS ON WHO YOU ASK.

ACTUALLY, THERE'S A ROBUST DEBATE INSIDE THE AI COMMUNITY RIGHT NOW.

THERE ARE THOSE WHO BELIEVE THE LARGE LANGUAGE MODEL IS THE KIND OF KEY TO UNLOCKING EVERYTHING, MEANING THAT AS WE CONTINUE TO MAKE ADVANCES IN THE CURRENT ARCHITECTURE IT WILL UNLOCK NEW CAPABILITIES IN THE PHYSICAL WORLD.

WITH ROBOTICS.

WITH SELF- DRIVING CARS, WITH INDUSTRIAL PRODUCTION.

THERE ARE OTHERS WHO BELIEVE THAT THERE'S A NATURAL LIMIT TO HOW FAR LARGE LANGUAGE MODELS CAN GO AND WHAT WE REALLY NEED TO DO IS FOCUS NOW ON WORLD MODELS, THAT IS, MODELS THAT ARE LESS ABOUT PREDICTING THE NEXT WORD AND MORE ABOUT MAPPING THE WORLD AROUND US.

AND IT IS ONLY THROUGH THAT MECHANISM THAT WE'RE GOING TO PRODUCE THE BREAKTHROUGHS IN PHYSICAL MANIFESTATIONS OF AI.

I OBVIOUSLY AM NOT IN A POSITION TO ADJUDICATE BETWEEN BRILLIANT PEOPLE ON TWO SIDES OF THAT DEBATE.

BUT I THINK THAT FOR THE UNITED STATES AS A MATTER OF STRATEGY WE HAVE TO LAY CHIPS ON BOTH BETS.

WE NEED TO BE CONTINUING TO INVEST IN THE CUTTING EDGE OF LARGE LANGUAGE MODELS, BUT WE ALSO NEED TO BE INVESTING SUBSTANTIALLY IN THE DEVELOPMENT OF THESE WORLD MODELS AND OF THE CAPACITY TO BUILD ROBOTS AND AUTONOMOUS SYSTEMS AS SPEED AND SCALE.

AND I THINK THERE'S A LOT OF INTERESTING WORK BEING DONE ACROSS THOSE AREAS THAT GO BEYOND JUST THE OPEN AIs AND ANTHONY ROPPICS AND GOOGLES OF THE WORLD.

>> THE DRIVING FORCE AT LEAST IN THE UNITED STATES RIGHT NOW IS THE PRIVATE SECTOR.

WE HAVE FOUR OR FIVE VERY BIG COMPANIES PURSUING THIS LIKE CRAZY.

AND IN YOUR PIECE YOU SORT OF SAY THAT CAN BE A PROBLEM IF THE INCENTIVES OF THE PRIVATE SECTOR AREN'T ACTUALLY ALIGNS WITH OUR NATIONAL INTERESTS.

EXPLAIN HOW THAT COULD HAPPEN.

>> WELL, WHAT'S REALLY INTERESTING, WALTER, IS IF YOU THINK ABOUT THE TECHNOLOGIES IN THE LAST 100 YEARS WITH TRULY TRANSFORMATIVE NATIONAL SECURITY IMPLICATIONS, NUCLEAR WEAPONS, THE SPACE AGE, THE INTERNET, ALL OF THESE BASICALLY WERE BIRTHED BY GOVERNMENT AND THEN, YOU KNOW, ENDED UP IN SOME KIND OF PUBLIC-PRIVATE PARTNERSHIP.

IN THE NUCLEAR WEAPONS CASE IT WAS PURELY GOVERNMENT BUT NUCLEAR POWER THEN WENT TO PRIVATE INDUSTRY.

IN THE CASE OF LARGE LANGUAGE MODELS AND GENERALITYIVE AI THE GOVERNMENT IS ON THE SIDELINES.

THIS IS THE COMPANIES DOING IT.

AND IT'S A RELATIVELY SMALL NUMBER OF REALLY LARGE COMPANIES.

NOW, ON THE ONE HAND THAT'S A HUGE NATIONAL ASSET THAT THE UNITED STATES OF AMERICA, UNIQUE AMONG COUNTRIES, HAS THESE BIG PRIVATE COMPANIES DRIVING TECHNOLOGICAL INNOVATIONS WITH ALL OF THESE MASSIVE IMPLICATIONS.

IT'S ALSO A RISK.

BECAUSE IF YOU GET TOO MUCH CONCENTRATION IN THE HANDS OF TOO FEW PEOPLE WHO ARE LOOKING FOR BASICALLY RETURNS TO THEIR SHAREHOLDERS AND NOT NECESSARILY ENTIRELY TO THE OVERALL PUBLIC INTEREST, THAT'S A CHALLENGE.

THAT'S WHY THE GOVERNMENT HAS TO COME IN AND MANAGE THAT TENSION WITH A HEALTHY SET OF TOOLS TO ENSURE THAT ULTIMATELY THE VEMT OF THIS TECHNOLOGY WORKS FOR US RATHER THAN AGAINST US.

BUT HERE WE'RE IN SOMEWHAT UNCHARTERED TERRITORY.

BECAUSE UNLIKE IN PAST INNOVATIONS WHERE THE GOVERNMENT WAS IN THE DRIVER'S SEAT HERE WE'RE VERY MUCH IN THE PASSENGER'S SEAT AND IT'S GOING TO REQUIRE A NEW STYLE, A NEW METHOD OF GOVERNMENT PARTNERSHIP WITH THE PRIVATE SECTOR.

WE BEGAN THAT WORK IN THE BIDEN ADMINISTRATION.

THAT'S GOING TO HAVE TO CONTINUE AS WE GO FORWARD.

>> WHY NOT JUST GET OUT OF THE WAY IF THESE COMPANIES ARE GOING REAL FAST AT IT?

>> WELL, A COUPLE REASONS.

SO FIRST OF ALL, I DISAGREE WITH THE IDEA THAT THERE IS NECESSARILY AN ENTIRELY INVERSE LINK BETWEEN SPEED OF INNOVATION AND BASIC NEGOTIATIONS OF SAFETY.

LET'S JUST TAKE RAILROADS AS ONE EXAMPLE.

WHEN THE RAILROAD INDUSTRY GOT OFF AND RUNNING THAT WAS THE PRIVATE SECTOR DOING ITS THING.

BUT TRAINS WERE GOING SLOW BECAUSE THEY DIDN'T HAVE THE KIND OF SWITCHING AND SIGNALING NECESSARY.

THEY HAD DIFFERENT GAUGES FOR DIFFERENT TRACKS.

AND IT WAS ONLY AFTER THE GOVERNMENT WORKED WITH THE PRIVATE SECTOR TO CREATE A SET OF BASIC RULES OF THE ROAD THAT TRAINS COULD HAVE CONFIDENCE TO DRIVE REALLY FAST.

AND I THINK A SIMILAR DYNAMIC HAS TO TAKE HOLD HERE TOO.

WE SHOULD NOT BE OVERREGULATING OR HOLDING BACK THE PRIVATE SECTOR UNNECESSARILY.

BUT WE SHOULD ALSO BE ENSURING THAT THIS TECHNOLOGY IS NOT SUBJECT TO MISUSE BY BAD ACTORS OR DOES NOT BECOME MISALIGNED IN WAYS WHERE IT'S ULTIMATELY DEEPLY HARMFUL EITHER ON A NATIONAL SECURITY BASIS OR A HEALTH OR ECONOMIC BASIS.

SO IT'S ABOUT GETTING THE BALANCE RIGHT.

AND THAT MEANS THE GOVERNMENT DOES HAVE A ROLE BUT THAT ROLE HAS TO BE ONE OF HUMILITY AND IT HAS TO BE ONE OF DEEP PARTNERSHIP WITH A PRIVATE SECTOR THAT IS A MASSIVE NATIONAL ASSET IN DRIVING THIS TECHNOLOGY FORWARD.

>> DOES CHINA HAVE CERTAIN SYSTEMIC ADVANTAGES, WHETHER IT'S BEING ABLE TO THROW MASSIVE RESOURCES AT ENERGY OR COMPANIES LIKE THAT, OR NOT HAVING A WHOLE LOT OF PRIVACY CONCERNS ABOUT BUILDING HUGE DATA BASES, OR DO WE HAVE A NATURAL ADVANTAGE?

>> CHINA HAS SOME ADVANTAGES FOR SURE.

THEIR CAPACITY TO PRODUCE THE ELECTRICITY TO POWER DATA CENTERS IS UNMATCHED IN THE WORLD AND IT OUTSTRIPS WHAT THE U.S.

CAN DO.

THEY HAVE PLENTY OF TALENT.

THEY HAVE PLENTY OF DATA.

WHAT THEY DO NOT HAVE THAT THE UNITED STATES HAS IS PLENTY OF COMPUTING POWER.

THOSE ADVANCED AI CHIPS THAT GO INTO THE DATA CENTERS THAT TRAIN THE MODELS AND RUN THE MODELS.

HERE THE UNITED STATES AND OUR DEMOCRATIC ALLIES HAVE A MASSIVE ADVANTAGE OVER CHINA.

AND WE PLACED IN THE BIDEN ADMINISTRATION CONTROLS ON THE SALE OF THE CHIPS AND THE EQUIPMENT THAT MAKES THOSE CHIPS TO CHINA SO THAT WE WOULD SUSTAIN THAT ADVANTAGE.

AND I WOULD SUBMIT TO YOU, WALTER, THAT THIS IS THE SINGLE MOST IMPORTANT INGREDIENT TO POWERING THE FUTURE OF THE AI REVOLUTION, COMPUTING POWER.

AND THAT WE NEED TO SUSTAIN OUR ADVANTAGE HERE.

THIS IS WHY I HAD CONCERNS ABOUT THE TRUMP ADMINISTRATION ACTUALLY DECIDING TO SHIP SOME OF THESE ADVANCED CHIPS TO CHINA.

THIS REDUCES THE AMERICAN COMPUTE ADVANTAGE.

IT HELPS CHINA SOLVE THE BIGGEST PROBLEM THAT IT HAS.

AND I DON'T THINK THAT THAT IS IN OUR INTERESTS.

AND I WOULD HOPE THAT THE TRUMP ADMINISTRATION CONTINUES TO SUPPORT A STRATEGY OF TECHNOLOGY CONTROLS ON THE MOST ADVANCED COMPUTING CHIPS TO CHINA.

>> WHEN YOU TALK ABOUT ARMS CONTROL BEING A PRECEDENT FOR AI TYPES OF TREATIES, WE USED TO 80 YEARS AGO CREATE GREAT INSTITUTIONS WHETHER IT BE NATO, THE WORLD BANK, TO DO THESE THINGS.

IS THERE SOME NEW GRAND INTERNATIONAL INSTITUTION LIKE A NATO OR A U.N.

THAT SHOULD COME ALONG FOR THE WISE MEN OF THIS GENERATION TO CREATE?

>> YOU KNOW, IT'S FREQUENT LY DISCUSSED IN THE CIRCLES OF PEOPLE WHO KIND OF FOLLOW AI AND NATIONAL SECURITY THAT WE SHOULD CREATE SOMETHING LIKE AN INTERNATIONAL ATOMIC ENERGY AGENCY EQUIVALENT FOR AI.

I THINK THAT ANALOGY DOESN'T QUITE HOLD IN THE SAME WAY BECAUSE THERE ARE REAL DIFFERENCES BETWEEN THIS GENERAL PURPOSE TECHNOLOGY THAT TOUCHES EVERYTHING AND NUCLEAR POWER, WHICH IS OBVIOUSLY TRANSFORMATIVE BUT NOT AS GENERAL PURPOSE AS AI.

HOWEVER, I BELIEVE THAT THERE SHOULD ULTIMATELY BE AN INTERNATIONAL INSTITUTION THAT LOOKS AT THE BASIC QUESTIONS OF SAFETY AND ALIGNMENT AND THE RISKS THAT AI POSES AND FINDING WAYS FOR THE WORLD COLLECTIVELY TO COLLABORATE TO MANAGE THOSE RISKS.

WHETHER IT'S IN CYBER AS I MENTIONED BEFORE, THE INTERSECTION BETWEEN AI AND BIOWEAPONS.

WHETHER IT'S TO DO WITH AI SYSTEMS THAT COULD GO ROGUE IN SOME WAY AND ACT CONTRARY TO THE INTEREST OF HUMANITY.

OR WHETHER IT'S JUST SIMPLY HOW TO MANAGE FOR YOU WILL AOF THE OTHER IMPLICATIONS OF ARTIFICIAL INTELLIGENCE INCLUDING ECONOMIC DISRUPTIONS, FINANCIAL RISKS AND WHAT HAVE YOU.

SO YES, THE BOTTOM LINE IS I DO THINK WE NEED A NEW INSTITUTION.

IT PROBABLY HAS TO LOOK AND FEEL DIFFERENT THAN ANYTHING WE SAW DURING THE COLD WAR.

AND IT ALMOST CERTAINLY HAS TO BE NOT JUST AN INSTITUTION OF GOVERNMENTS BUT AN INSTITUTION THAT INVOLVES THE PRIVATE SECTOR AND CIVIL SOCIETY AS WELL BECAUSE OF THE CROSS-CUTTING NATURE OF THIS TECHNOLOGY.

>> YOU KNOW, YOU WRITE IN YOUR PIECE THAT THE U.S.

SHOULD USE AI TO STRENGTHEN DEMOCRATIC VALUES BOTH AT HOME AND AROUND THE WORLD.

HOW COULD WE DO THAT AND ESPECIALLY HOW COULD WE DO THAT WHEN OTHER NATIONS MIGHT THINK WE'RE WEAPONIZING IT?

>> FIRST OF ALL, THIS COMES DOWN TO GUARDING AGAINST SOME OF THE RISKS TO DEMOCRACY FROM ARTIFICIAL INTELLIGENCE.

THE ABILITY TO ENGAGE IN MASS SURVEILLANCE AND REPRESSION.

THE ABILITY TO TAKE THE PERSONAL DATA OF CITIZENS AND USE IT FOR NEFARIOUS PURPOSES BY GOVERNMENTS.

THE ABILITY TO DRIVE PROPAGANDA, MISINFORMATION AND DISINFORMATION.

SO PART OF IT IS ABOUT HAVING A DOMINANT POSITION WITH RESPECT TO THE TECHNOLOGY SO THAT YOU CAN WARD OFF THOSE VERY HARMFUL RISKS.

AND THE PROOF IS IN THE PUDDING AS TO WHETHER THE U.S.

IS PREPARED TO WORK WITH ITS PRIVATE COMPANIES TO BE ABLE TO DO THAT.

>> BUT IF THE WORLD WAS BUILT ON CHINESE AI, THE RISKS OF THOSE THINGS ARE MUCH HIGHER BECAUSE WE SEE MANY OF THOSE THINGS PLAYING OUT RIGHT NOW ACTUALLY IN CHINA.

THAT'S ONE THING.

THE SECOND THING IS IT GOES TO YOUR LAST QUESTION, WHICH IS THE U.S.

HAS TO BE INVOLVED IN A COLLABORATIVE EFFORT WITH OTHER COUNTRIES AROUND SETTING THE RULES, NORMS AND STANDARDS WHEN IT COMES TO AI.

AND THAT SHOULD BE DONE THROUGH MULTIPLE DIFFERENT VERTICALS.

IT SHOULD BE DONE THROUGH DEMOCRATIC ALLIES AND PARTNERS AND IT SHOULD BE DONE THROUGH THE UNITED NATIONS.

AND I'M PROUD THAT THE BIDEN ADMINISTRATION SUPPORTED THE EFFICIENCY GENERAL ASSEMBLY RESOLUTION ON ARTIFICIAL INTELLIGENCE TO TRY TO LAY DOWN SOME BASIC PRECEPTS ABOUT HOW THIS TECHNOLOGY SHOULD BE USED CONSISTENT WITH THE VALUES THAT WE HOLD DEAR.

>> JAKE SCHK ULLIVAN, THANKS SO MUCH FOR JOINING US AGAIN.

APPRECIATE IT.

>> THANK YOU FOR HAVING ME.