Bradley Tusk on Anthropic IPO: Why add public market pressure if safety is your top priority?

IT'S JUST ABOVE 90. PRESIDENT TRUMP SAYS THE U.S. WILL NOT ATTACK IRAN BEFORE THE MIDTERM ELECTIONS. >> OKAY. MEANTIME, ANTHROPIC REPORTEDLY EYEING AN IPO IN NOVEMBER. JUST A SAFETY. CONCERNS ABOUT AI ARE HEATING UP AROUND THE WORLD. VENTURE CAPITALIST AND ENTREPRENEUR BRADLEY TUSK IS HERE. TUSK VENTURES FOUNDER AND CEO, ARGUES THAT SAFETY IS AT ODDS WITH AN IPO. EXPLAIN. >> SO DARIO MODY IS VERY GOOD ABOUT TALKING ABOUT SAFETY. HIS WHOLE THING IS WE HAVE TO REALLY BE MINDFUL OF THE RISKS. WE NEED REGULATION. HE EVEN SAYS THINGS LIKE, AI MIGHT KILL US ALL, WRITE THEIR OWN. S-1 SAYS THAT THEIR PRODUCT PROPOSES POSES AN EXISTENTIAL RISK TO HUMANITY. SO IF YOU'RE REALLY THAT WORRIED ABOUT SAFETY, IS TAKING ON ALL THE PRESSURE OF THE PUBLIC MARKETS. ON TOP OF THAT, A GOOD IDEA BY DEFINITION EVERY QUARTER. AND YOU GUYS SPEND HALF OF YOUR SHOWS TALKING ABOUT THIS, YOU HAVE AN EARNINGS CALL AND YOU WANT TO NOT DISAPPOINT THE ANALYSTS IN THE MARKET BECAUSE THE SHARE PRICE IS A REPORT CARD ON YOUR PERFORMANCE AND IS A MAJOR PART OF YOUR COMPENSATION. AND IF YOU HAVE ALL THIS PRESSURE TO MEET NEW REVENUE TARGETS, THE WAY YOU DO THAT, IF YOU'RE ANTHROPIC, IS YOU KEEP RELEASING MORE AND MORE SOPHISTICATED MODELS. BUT IF A MODY BELIEVES THAT THOSE MODELS MIGHT KILL US ALL, WHY WOULD YOU WANT TO ADD THE PRESSURE OF THE PUBLIC MARKETS IF YOU THINK THAT SAFETY. >> SO YOU'RE MAKING YOU'RE MAKING THE SAM ALTMAN ARGUMENT. UNLESS YOU BELIEVE THAT SAM ALTMAN IS MAKING THAT ARGUMENT BECAUSE THEY'RE NOT READY TO GO PUBLIC. >> LOOK, IT IS HARD FOR ME TO BELIEVE THAT SAM ALTMAN CARES ABOUT ANYTHING OTHER THAN SAM ALTMAN FROM EVERYTHING THAT WE'VE EVER SEEN FROM THE GUY. SO I'M NOT WILLING TO GRANT HIM ALL THAT MUCH. BUT A MODY AT LEAST POSITS HIMSELF AS LUKE SKYWALKER TO ALTMAN'S. >> PUSHBACK ON ONE THING ABOUT THE MODY ARGUMENT THAT I THINK IS LARGELY MISUNDERSTOOD. PEOPLE SAY, OH MY GOODNESS, IT'S SO HYPOCRITICAL TO SAY, I'M GOING TO HAVE AN IPO AT THE SAME TIME THAT I'M GOING TO RUN FULL STEAM AHEAD. AT THE SAME TIME THAT I'M ARGUING, WE NEED LOTS OF REGULATION. CORRECT? HOW COULD YOU BE DOING ALL THE SAME? YES. I THINK HAVING NOW COVERED THIS GUY FOR A LONG TIME, AND ALL OF THEM, WHEN THEY SAY THEY WANT GUARDRAILS OR THEY THEY WANT THE INDUSTRY TO HAVE GUARDRAILS. THE REASON THAT THEY'RE ASKING FOR THAT IS BECAUSE THEY KNOW THAT THE TRUTH IS THAT IF THEY EVEN IF THEY GO FULL SPEED AHEAD, OR ACTUALLY, RATHER, IF THEY WERE TO PULL BACK THE WAY YOU MIGHT BE SUGGESTING, YOU MIGHT SAY, OH, IF THEY'RE SO RESPONSIBLE, THEY'LL PULL BACK. WHAT THEY'RE SAYING IS THE CAT'S OUT OF THE BAG ANYWAY. IT DOESN'T. WE COULD PULL BACK. OPENAI COULD PULL BACK. IT DOESN'T MATTER BECAUSE THE OPEN WEIGHT MODEL GUYS IN CHINA, THEY'RE GOING TO KEEP GOING ANYWAY. SOMEBODY ELSE IS GOING TO KEEP GOING ANYWAY. THAT THE ONLY TRUE ANSWER TO THIS EXISTENTIAL THREAT, IF THAT IS IF THAT IS REAL, IS THAT YOU EFFECTIVELY NEED REGULATION FOR EVERYBODY OR THAT THIS WHOLE THING DOESN'T WORK. SO THE IDEA THAT SOMEHOW THEY'RE GOING TO SOMEHOW SLOW THEMSELVES DOWN OR SLOW DOWN DOESN'T REALLY MATTER IN THE IN THE GRAND SCHEME OF THE UNIVERSE. RIGHT. >> BUT UNDER THAT THEORY, YOU WOULD SAY PEOPLE CAN GET FOOD POISONING ANYWAY. SO LET'S JUST GET RID OF THE FDA. SO YES, I AGREE WITH YOU. I THINK THE REASON THEY WANT REGULATION IS LESS ABOUT EVEN SAFETY AND IT'S MORE ABOUT, HEY, IF WE CAN REGULATE TO PROTECT US FROM OURSELVES, IF WE'RE ALL PLAYING BY THE SAME RULES. >> DO YOU THINK IT'S A REGULATORY CAPTURE STORY? >> SURE. BUT AT THE SAME TIME, JUST LIKE A MODY, WHEN HE WAS IN THE WHITE HOUSE LAST WEEK WITH TRUMP DIDN'T STAND UP AND SAY, SELF-REGULATION IS WRONG. WE NEED ACTUAL GOVERNMENT REGULATION. WHY? BECAUSE WHEN YOU'RE SITTING IN THE WHITE HOUSE, THE MOST POWERFUL PERSON IN THE WORLD IS SAYING THIS, AND YOUR PEERS ARE NODDING AND DOING THIS. IT'S REALLY HARD TO. >> OBJECT BECAUSE THERE'S A SILENCING EFFECT. >> CORRECT. HE'S GOT THE SAME PROBLEM WITH THE PUBLIC MARKETS. HE'S NOT GOING TO WANT TO STAND UP AND SAY, NO, WE'RE NOT GOING TO RELEASE THESE NEW MODELS THAT WILL GENERATE MORE REVENUE BECAUSE IT'S THE WRONG THING TO DO BECAUSE IT'S UNSAFE. HE'S GOING TO WANT TO APPEASE THE ANALYSTS AND THE MARKETS, JUST LIKE HE WANTED TO APPEASE THE PRESIDENT. AND THAT IS PUTTING HIM IN A REALLY DIFFICULT. >> LET ME ASK YOU EVEN MORE COMPLICATED QUESTION. SURE, IT IS POSSIBLE THAT MAYBE AN ANTHROPIC SHOULDN'T GO PUBLIC YET, AS YOU MIGHT SUGGEST, IF ANTHROPIC WERE TO CHOOSE NOT TO GO PUBLIC RIGHT NOW, WHAT DO YOU ACTUALLY THINK WOULD HAPPEN TO THE STOCK MARKET IN TOTAL? AND I ASK BECAUSE YOU SAW WHAT HAPPENED JUST YESTERDAY, WHEN IT SEEMED LIKE OPENAI'S REVENUE MIGHT BE LESS THAN EXPECTED. YOU SAW A STEP CHANGE, RIGHT? >> SO THAT'S ONE ARGUMENT, RIGHT? WHICH IS THAT IN ORDER TO KEEP THE TRAINS MOVING OR THE OR THE CHAINS MOVING FROM THE FOOTBALL ANALOGY, YOU JUST GOT TO KEEP GOING HERE WOULD BE THE OTHER WAY TO LOOK AT IT, WHICH IS IF THERE IS AN AI BUBBLE, I THINK THERE'S A PRETTY GOOD CHANCE THAT THERE IS. ONCE YOU REALLY EXPOSE IT, THAT'S WHEN YOU'RE REALLY RUNNING THE RISK. >> SO THIS IS THE CANARY IN THE COAL MINE RIGHT HERE. >> SO ARGUABLY ANTHROPIC RIGHT NOW, IF YOU TAKE THE MOST GENEROUS REVENUE ESTIMATES AND EXCLUDE THE 518 BILLION IN CAPEX COMMITMENT, IT'S STILL A 27 X MULTIPLE. THAT'S PROBABLY NOT SUSTAINABLE UNLESS YOU'RE PRODUCING THE KIND OF REVENUE TO COMPETE AGAINST IT. SO ACTUALLY HIDING THEM FROM THE STOCK MARKET MIGHT BE THE BEST WAY TO PROTECT THE STOCK MARKET. >> THAT'S INTERESTING. SO, OKAY, WHAT IS YOUR GAMBLE THAT THIS IPO ACTUALLY HAPPENS? BECAUSE THERE IS A VIEW THAT THEY'LL FLIP THAT EITHER THEY NEVER FLIP THE THE PROSPECTUS OR THEY FLIP THE PROSPECTUS AND THE MARKET EFFECTIVELY, OR INVESTORS SAY, YOU KNOW, WE'RE NOT DOING THIS AT A $2 TRILLION VALUATION. AND IF THEY HAVE TO GO AT A $1 TRILLION VALUATION, I'VE HEARD PEOPLE SAY IT'S A BETTER ARGUMENT. YOU KNOW, THEY STILL SHOULD. I KNOW OTHER PEOPLE WHO SAY THEY SHOULDN'T. >> I TELL PORTFOLIO COMPANIES ALL THE TIME. WE DON'T ALWAYS NEED TO GRAB THE HIGHEST VALUATION IN THE NEXT ROUND. WE NEED TO GET A NUMBER. THAT IS, HOW MUCH MONEY DO YOU REALLY NEED? WHAT'S EVALUATION THAT WE CAN ACTUALLY JUSTIFY? BECAUSE THE GOAL ISN'T TO HAVE A REALLY GOOD STORY ON SQUAWK BOX. THE GOAL IS TO ULTIMATELY HAVE A GOOD EXIT. >> BUT THE. >> SAME THING IS TRUE HERE. >> THE PROBLEM IN THIS INSTANCE IS THAT THESE COMPANIES ARE SO CAPITAL INTENSIVE THAT THEY NEED TO CAPTURE A. THERE'S TWO ISSUES. THEY NEED TO CAPTURE A VALUATION WHERE THEY CAN CAPTURE A HUGE CHUNK OF CASH. THEY NEED THE CAPITAL, THEY NEED TO GET AS MUCH CAPITAL AS THEY CAN RIGHT NOW BECAUSE OTHERWISE, RIGHT. THAT'S, THAT'S PART OF IT. AND THEN THE SECONDARY ISSUE IS SO MANY OF THE OTHER HYPERSCALERS, WHETHER IT'S AN AMAZON OR THINK ABOUT A MICROSOFT OR NVIDIA OR YOU NAME IT, ALL HAVE STAKES ALREADY IN THESE COMPANIES AND ARE CARRYING THEM ON THEIR BALANCE SHEET. RIGHT. SO IF IN FACT, ANTHROPIC VALUATION IS HALF OF WHAT YOU THINK IT IS, OPENAI'S VALUATION, HALF WHAT YOU THINK IT IS, IT IS A STEP CHANGE FOR EVERYBODY. >> YEAH. BUT YOU KNOW WHAT, FIRST OF ALL. THE RESPONSIBILITY IS NOT THEM. BUT TWO, THERE IS AN ARGUMENT THAT A MODY WHO, YOU KNOW, AS YOU HAVE THE KEVIN ROOSE BOOK THERE, MODY GETS CREDITED WITH THE GUY THAT THOUGHT ABOUT SCALE AS THE WAY TO PROPEL ALL OF THIS. BUT THE REALITY IS, IF YOU LOOK AT AT MUSE AND YOU LOOK AT KIND OF THE META MODELS, 99% OF THE PEOPLE WATCHING THIS SHOW, INCLUDING ME, DON'T NEED THE CAPABILITIES OF ANTHROPIC OR OPENAI AND OPEN SOURCE INFERENCE MODEL, EVEN IF IT'S FAR LESS POWERFUL, MIGHT ACTUALLY BE THE FAR BETTER PLAY. BECAUSE IF I HAVE A CHOICE OF EITHER USING LLAMA FOR FREE AND THEN GETTING AGENTS ON TOP OF THAT FOR FREE, OR PAYING ANTHROPIC 100 BUCKS A MONTH, IT MAY BE THAT A MODY IS UNDERLYING ECONOMIC BET WAS WRONG. THAT'S NOT OUR PROBLEM. >> BUT PART OF HIS SCALE ARGUMENT IS A SLIGHTLY DIFFERENT ONE, AND IT GETS TO THE AGI OF IT ALL. PRESIDENT TRUMP WOULD SAY IT'S SGI NOW. SURE, WHICH IS THAT I BELIEVE HE THINKS THAT THERE'S A MOMENT IF YOU CAN GENUINELY GET TO TRUE RECURSIVE LEARNING, WHERE THE MODELS ARE TRAINING THEMSELVES, THAT THERE'S A HOCKEY STICK EFFECT AND EFFECTIVELY WHERE YOU COULD BE GOING LIKE THIS. AND THE SECOND YOU CAN START LEARNING YOURSELF, LIKE BY YOURSELF, LET'S SAY YOU'RE 3 OR 4 MONTHS AHEAD OF THE NEXT GUY, YOU COULD ALL OF A SUDDEN BECOME TWO AND THREE YEARS AHEAD OF THE NEXT GUY AND MAYBE BE THAT WAY FOREVER. TO WHAT DO YOU THINK? >> THAT'S TRUE. BUT TO WHAT END, I THINK IS THE QUESTION I WOULD HAVE, WHICH IS I DO BELIEVE THAT YOU'VE GOT, YOU KNOW, SCIENTISTS IN LABS TRYING TO CURE CANCER OR FIGURE OUT HOW TO REMOVE CARBON FROM THE ATMOSPHERE OR THINGS LIKE THAT. WE'RE ABSOLUTELY EVERY EXTRA PIECE OF, OF KNOWLEDGE IS HELPFUL. BUT IF MOST OF US WHAT AM I DOING FOR MY SUBSTACK? I'M LIKE FACT CHECKING THINGS OR I'M TRYING TO DO SOME MODELING FOR MY FUND. I DON'T REALLY NEED THAT. AND YOUR POINT ABOUT THE RECURSIVE LEARNING IS ALSO THE ITERATIVE DEPLOYMENT, BY THE WAY, THAT. >> YOU'RE TALKING ABOUT IS ALREADY AN OPEN WAY. MODELS YOU COULD PROBABLY GET FOR FREE AND RUN ON YOUR LAPTOP, CORRECT. >> SO OFTENTIMES WHEN WE KEEP TALKING ABOUT THE POTENTIAL, WHAT NO ONE SEEMS TO EXPLAIN TO ME IS A WHY EXACTLY. THE VAST MAJORITY OF THE ECONOMY NEEDS THAT, AND B, PEOPLE JUST THROW OUT, WELL, CHINA, WHAT DOES THAT MEAN? YOU KNOW, WE'RE NOT NO ONE IS SAYING THAT THE US MILITARY SHOULDN'T HAVE THE CAPABILITIES THAT IT NEEDS. NO ONE'S