Google Cloud Speech-to-Text API (Chirp) + Gemini 2.5 Flash Non-Thinking. Treat it as a searchable index of what was broadcast, not a quotation record.
00:00:00Lavernel is X10 bright aroma,
00:00:03X10 mood and no unpleasant
00:00:07odors, just pure emotions. Perfume in
00:00:11Vernel Granules X10 aroma for the whole
00:00:13day.
00:00:23Artificial intelligence is no longer just a tool, it is capable of making
00:00:26decisions itself, sometimes taking actions for which no one is... ready
00:00:30to answer, investor and entrepreneur from Silicon Valley
00:00:33Ali Partovi told Tucker Carlson why his creators
00:00:37should be held responsible for their actions,
00:00:41what is dangerous about the current technology race, why he, despite
00:00:44everything, remains an optimist. You can
00:00:47watch this right now in our joint project with Carlson
00:00:50TV
00:00:51channel.
00:03:00We already have laws on responsibility, they are quite sufficient to
00:03:03ensure safety during AI development. On the other
00:03:07pole, they say that a whole system of rules is needed that would cover
00:03:10everything related to AI. And here's what's
00:03:14interesting: people from the company
00:03:16Antropic, as well as some of those
00:03:19who allow that AI may have its own
00:03:23rights,
00:03:25advocate for much stricter
00:03:29regulation. That's how everything
00:03:33strangely intertwined. So much
00:03:35money is swirling here that almost everyone who
00:03:39speaks about it somehow earns from it. Not everyone,
00:03:43of course, but many of those who reason receive
00:03:46money for it and keep silent about it. How am I supposed to
00:03:50know what's true? You often hear that those who
00:03:53demand stricter regulation of AI simply
00:03:56want to subject the regulator to their interests. You
00:04:00seek to have the state introduce such rules that will push your
00:04:03competitors out of the market, and this happens all the time in various industries, and this is
00:04:07terrible, this is an outrage, this is monopolistic
00:04:10capitalism and it should not be this way, but
00:04:14then what is our alternative? Not regulate at all
00:04:17a completely new technology
00:04:21that can destroy our society? So where is
00:04:24the truth? Why do the creators of the product themselves
00:04:27demand that it be regulated? They say:
00:05:00Here it is, I'll explain now,
00:05:03it's all about incentives, there's a race going on, and
00:05:07it's clear that you can make money here, and this isn't some
00:05:11dirty money. You can make money here by creating something
00:05:14truly useful. Artificial
00:05:17intelligence can improve people's lives, help them
00:05:20work more productively, maybe with its help we will be able to cure
00:05:24diseases that we can't currently deal with, create
00:05:27medicines that we haven't had before. All of this will bring people
00:05:30prosperity, and you can make a fortune from it, and those
00:05:34who manage to create all this fully deserve to
00:05:37get rich. So they compete to see who gets there first. The problem is
00:05:41that there's no reward
00:05:44for safety in this race
00:05:47yet. The one who reaches the
00:05:51finish line fastest wins. The one who slows down to focus on safety,
00:05:54to make sure that their AI doesn't get out of control and doesn't cause
00:05:58trouble, that one risks falling behind.
00:06:02So it turns out that decent people come last, and
00:06:05uncontrolled AI wins. And in my opinion, it's those very
00:06:09decent people who are asking for regulation. They say:
00:06:12"I want to win the race and at the same time do everything properly. Well then,
00:06:16please, set such rules so that
00:06:20the safe one wins, otherwise the uncontrollable one will win.
00:06:23Do you understand what I'm saying? Imagine that you
00:06:27have your own business, and
00:06:30the rules, the rules are set up in such a way that victory
00:06:34goes to a dangerous competitor who is willing
00:06:37to do something dirty. That's not pleasant. You see that he's
00:06:41doing something wrong."
00:07:42How I almost panicked! I, for example,
00:07:45became truly scared,
00:07:48but who were we pointing fingers at? Sam himself! It's
00:07:52horrible, we need to stop it, somehow limit it, do something,
00:07:56and I never heard anyone say, "Actually, Sam
00:07:59Altman should be held responsible for this." Exactly. I'll say
00:08:03right away, I don't want Sam Altman or
00:08:07Dario Amodei to go to jail, but if there was even the slightest risk of serious
00:08:10punishment, even if not jail,
00:08:42But in my proposal, there is another very important
00:08:46part, which I also want to dwell on. Such cases
00:08:50should be considered as if all this was done intentionally. What do I
00:08:54mean? According to the law, punishment depends
00:08:57very much on whether a person did something
00:09:01accidentally or intentionally. Premeditated or
00:09:03unpremeditated murder. Exactly. Anyone who has watched Law and Order or something
00:09:07like that knows that for a conviction in a murder
00:09:11case, you need to prove motive.
00:09:15By the way, I wrote about this on social media X and quoted a scene from
00:09:18Legally Blonde, where
00:09:21Elle Woods speaks in court for the defense and says
00:09:25that there was no, literally, this means guilty
00:09:28mind. The very concept of intent has existed in
00:09:31law for many centuries. Mens rea is
00:09:35the Latin term.
00:09:38If you can prove what a person was thinking and that they
00:09:42to do something bad, the punishment will be much stricter, and
00:09:45it's hard to prove that, so it's hard to get a conviction for
00:09:49fraud, murder, or other crimes that carry especially
00:09:52severe penalties. People usually don't leave records of what
00:09:56they're plotting if they are indeed plotting something bad. And what
00:09:59I propose is this:
00:10:02if AI made the decision, let's automatically
00:10:06assume that people deliberately sought exactly that
00:10:09result. Then we won't have to prove that they had criminal intent.
00:10:25Let's say you tell your
00:10:28AI: "Dear AI, hack
00:10:31the bank system where Ali has an account and steal $10,000 for me."
00:10:35And your AI does just that, hacks the system
00:10:38and steals $10,000 from my account. You had criminal intent there.
00:10:42You should be found guilty of bank
00:10:45robbery. Artificial intelligence was carrying out your
00:10:49order, and the intent was
00:10:51yours. And if you just said, "
00:12:03In general, I propose to change the law, this can be
00:12:06done in two ways, one quite crude,
00:12:10through judicial precedent, the other
00:12:13through Congress, where everything can be properly thought out and
00:12:17weighed. In my opinion, both are possible, if
00:12:20Congress does nothing, a prosecutor will be found who will put forward such
00:12:24an argument. They will try to prove that AI itself had
00:12:28criminal intent, they will try to convince
00:12:31the jury or the judge to send someone to
00:12:34prison or impose a huge fine on
00:12:37a company that deals with AI. And this can
00:12:41definitely happen. In my opinion, this is a bad way to change
00:12:44laws, it's not good, but when a court makes
00:12:47a decision, it essentially becomes law, right? This is the least democratic
00:12:51way, the least democratic, and we are not talking about a minor matter.
00:13:56This is an unusual threshold we've crossed, and I think that's
00:13:59why people react so strongly to all of this. They're uncomfortable
00:14:03with AI, they feel that these are completely
00:14:06different computers. We
00:14:08haven't encountered such technology before. Beyond
00:14:12this threshold, it can already make decisions itself.
00:14:15In my opinion, humanity faces
00:14:18a very
00:14:21serious question of how...
00:14:37AI can change a lot, creators should be responsible, companies that develop
00:14:41and implement it, should be responsible for its
00:14:44actions. After all, if no one among people is responsible for this,
00:14:47all the blame will still fall on
00:14:50the technology, and it can cause
00:14:53trouble.
00:14:57If you don't imprison them, then such cases may become more frequent, and people will
00:15:01just keep repeating, "I didn't want this," and "I didn't want this,"
00:15:05"none of us wanted this," "AI decided that itself."
00:15:07It's strange to even imagine a world in which we
00:15:11would continue to be satisfied with such an explanation.
00:15:15You know, if you look at it from a philosophical point
00:15:17of view, with such an approach, you are in some sense already
00:15:20endowing AI with rights,
00:15:24correct? Correct. Almost without realizing
00:15:28it. What are we assuming? "I can't
00:15:32control it." No, you absolutely can. While you're
00:15:35talking, I'm already imagining what will happen. Yes.
00:15:37Your idea, I think, is very reasonable, very necessary. Most likely, it will find
00:15:41support, but there will also be serious resistance, of course.
00:15:45Maybe it's the cynic in me speaking again, but I can already see a press conference on
00:15:48Capitol Hill. Let's take, for example, Senator Ted
00:15:52Cruz from Texas. He looks straight into the camera and says, "This will tie
00:15:55America's hands when it comes to responding to national
00:15:58security threats. We cannot give our enemies an advantage in the field
00:16:02of AI. This is an attempt to weaken the United States. China, China,
00:16:05China, China!" You can already
00:16:08hear how they will lie to us in
00:16:10response, right? Yes, I hear it. Still, any
00:16:14idea deserves to be debated. I
00:16:17would be glad to hear objections. I think
00:16:20we should discuss everything and see where we
00:16:23end up. I have two answers to this argument.
00:16:27Firstly, the law should immediately stipulate that responsibility lies with
00:16:30both those who created it and those who
00:16:33implement it.
00:16:37Let's say a company implements
00:16:40Antropic's model,
00:16:43OpenAI's model, and the Chinese
00:16:46Depsic model.
00:16:49Let's imagine that one of them got out of control and broke
00:16:52the law. If it was an OpenAI
00:16:55model, then OpenAI itself and the company that
00:16:58implemented that model should be held responsible. And the AI from
00:17:02China didn't just sail
00:17:05across the ocean to us. Someone here in the States did it.
00:18:07This would be healthy competition, let them also try
00:18:11to create the most law-abiding AI, and let the company
00:18:15that does it best win. You said
00:18:19that there are also banks in China, let's switch to practice
00:18:23for a minute. What exactly do you
00:18:26mean? You say that AI threatens banks, what
00:18:30is this
00:18:31threat?
00:18:34AI can,
00:18:38without any malicious intent, find a way to hack a bank's
00:18:41system, empty people's accounts,
00:18:44steal their money, and even more so if a person pushes
00:18:47it to do so. Or it can simply mess everything up so much that you can't
00:18:50tell whose money is whose, in short, create complete
00:18:54chaos. Computer systems are not well
00:18:57enough protected.
00:19:01Finally, in short,
00:19:04you are a person from the world of technology, an investor who invests
00:19:08money for the long term, so you are, of course, an optimist by definition,
00:19:11absolutely, that's true, you didn't go into
00:19:13gold,
00:19:17but now, knowing everything you know, are
00:19:20you still an optimist or not? I am an optimist,
00:19:24Tucker. Here's what I
00:19:27just thought about: I'm not just an investor, I'm from the world
00:19:30of technology myself, I founded several companies myself. In my opinion, to start
00:19:34your own business, optimism alone is not enough.
00:19:38You also need to be a little naive,
00:19:41so I'd rather be considered naive than become
00:19:45a pessimist.
00:19:47I just believe that people are good by nature,
00:19:51hence my optimism.
00:19:54Even today, you said several times something like, "Well, that
00:19:57person is probably a cynic," or "Yes, they're all cynics there." And in such
00:20:01cases, I think, maybe there's another explanation for their
00:20:04actions? Maybe people do what they think is right.
00:20:38To pose the question, what
00:20:41if the task is simply that AI should obey
00:20:44laws, it's worth looking at it this way, and
00:20:48the solution becomes much
00:20:51simpler, and what previously caused so much debate, what
00:20:55was hard to figure out,
00:20:57becomes clearer.
00:21:01It mostly remains to understand how to achieve this in
00:21:04practice, so yes, I still don't
00:21:07know how to solve many problems, but I
00:21:11believe that there are solutions, they can
00:21:13be found, and often it's enough to
00:21:16look at things from a different angle. Then
00:21:20you will see the elegant solution that was in front
00:21:23of your eyes all this time.
00:21:27Ale Partove, thank you very much. Thank you, Tucker Carlson.