Latest / Connor with Honor: AI, Real Estate and Real Life / Why AI Can Never Settle Our Biggest Fights: Truth vs. Values

Transcript

Why AI Can Never Settle Our Biggest Fights: Truth vs. Values · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: Here's a question people are starting to ask out loud.
• 0:03
  SPEAKER_00: We find about everything religions, guns, race, who has money and who doesn't, which country is allowed to have the most dangerous weapons on the planet.
• 0:12
  SPEAKER_00: We never agree and we never stop.
• 0:15
  SPEAKER_00: So somebody says, fine, let the machine decide.
• 0:18
  SPEAKER_00: Artificial intelligence, the smartest thing we've ever built, it has no skin in the game, it's not on the left, it's not on the right, it's not black, it's not white, doesn't go to church, and it doesn't skip church.
• 0:30
  SPEAKER_00: Just feed it the facts and let it hand down the answer.
• 0:33
  SPEAKER_00: Neutral, fair, done.
• 0:35
  SPEAKER_00: And that sounds amazing.
• 0:37
  SPEAKER_00: And also understand that AI is controlled by human beings, not autonomous, not conscious, not self-aware, at least from what we know about what's happening behind the scenes, which probably isn't much.
• 0:49
  SPEAKER_00: But we're talking about, as it is currently, being controlled by humans, humans in charge, at least to some level.
• 0:56
  SPEAKER_00: AI, it's also the most dangerous idea of the whole era if we want to give it control.
• 1:02
  SPEAKER_00: And by the end of this, I think you're going to see exactly why.
• 1:05
  SPEAKER_00: Step by step, nothing left on the table.
• 1:07
  SPEAKER_00: So let me start by clearing up what artificial intelligence even is, because most people picture the wrong thing.
• 1:14
  SPEAKER_00: In the 50s, at a conference, Dartmouth conference, AI was a marketing term.
• 1:20
  SPEAKER_00: Somebody came up with it and it stuck.
• 1:22
  SPEAKER_00: Artificial intelligence to words.
• 1:24
  SPEAKER_00: Artificial means man-made, not natural.
• 1:28
  SPEAKER_00: Intelligence means the ability to figure things out.
• 1:31
  SPEAKER_00: Put them together and you've got a man-made system that figures things out.
• 1:34
  SPEAKER_00: People shorten it to two letters, AI.
• 1:36
  SPEAKER_00: When you hear AI, that's all it means.
• 1:39
  SPEAKER_00: Man-made thinking.
• 1:40
  SPEAKER_00: And it also carries a connotation to not being as good as the original, which would mean it probably can't achieve what humans can achieve.
• 1:49
  SPEAKER_00: And that could be a dangerous, a dangerous thing.
• 1:52
  SPEAKER_00: But let me get into this.
• 1:53
  SPEAKER_00: Now here's the part nobody tells you.
• 1:55
  SPEAKER_00: The kind of AI everybody's talking about right now learned to do what it does by reading.
• 2:01
  SPEAKER_00: Read a huge chunk of everything humans have ever written books, websites, arguments, seen every movie, listened to every tape, every video, every CD, comment sections, all of it.
• 2:11
  SPEAKER_00: Then it learned to predict what words come next based on all that human writing.
• 2:16
  SPEAKER_00: It learned from us.
• 2:17
  SPEAKER_00: Every word it knows it got from a human first.
• 2:20
  SPEAKER_00: The fact is going to matter more than anything else I say today.
• 2:25
  SPEAKER_00: Now let's talk about two kinds of problems in the world.
• 2:28
  SPEAKER_00: Because this is the whole ballgame and almost nobody separates them.
• 2:31
  SPEAKER_00: The first type and the kind I'll call the truth problem.
• 2:35
  SPEAKER_00: A truth problem has a real answer sitting out there in reality, whether we like it or not.
• 2:39
  SPEAKER_00: How far is the moon?
• 2:41
  SPEAKER_00: What's two plus two?
• 2:42
  SPEAKER_00: What kills this specific cancer cell?
• 2:45
  SPEAKER_00: There is a correct answer.
• 2:46
  SPEAKER_00: We might not know it yet, but it exists.
• 2:48
  SPEAKER_00: Reality already decided.
• 2:50
  SPEAKER_00: We're just trying to find it.
• 2:52
  SPEAKER_00: For truth problems, being smarter helps a lot.
• 2:56
  SPEAKER_00: A smarter machine can search faster, spot patterns we miss, test more possibilities.
• 3:02
  SPEAKER_00: If the answer is really out there, more brain power gets you closer to it, and that's real.
• 3:06
  SPEAKER_00: The second kind, it's called a values problem, and this is completely different.
• 3:10
  SPEAKER_00: The values problem is not about what's true, it's about what matters, what's fair, what's sacred, what's worth more than something else.
• 3:18
  SPEAKER_00: And here's the thing that breaks people's brains.
• 3:20
  SPEAKER_00: There's no correct answer sitting out in reality waiting to be found, because the answer depends on what you care about.
• 3:27
  SPEAKER_00: And people care about different things.
• 3:30
  SPEAKER_00: Let me make that concrete.
• 3:33
  SPEAKER_00: It is more important to be free or to be safe.
• 3:37
  SPEAKER_00: That's not a fact question.
• 3:40
  SPEAKER_00: There's no number you can measure.
• 3:42
  SPEAKER_00: Some people would give up a little freedom for more safety.
• 3:46
  SPEAKER_00: Some people would rather die before they give up an inch of freedom.
• 3:51
  SPEAKER_00: Neither one is doing the math wrong.
• 3:53
  SPEAKER_00: They just want different things.
• 3:55
  SPEAKER_00: Now watch what happens when you point the smartest machine on earth at a values problem.
• 4:00
  SPEAKER_00: It can't solve it.
• 4:01
  SPEAKER_00: Not because it's not smart enough, but because there's nothing to solve.
• 4:05
  SPEAKER_00: There's no hidden right answer buried in the data.
• 4:08
  SPEAKER_00: When two people fight over what's fair, they're not both reacting or reaching for the same fact and fumbling with it.
• 4:15
  SPEAKER_00: They just want different things.
• 4:16
  SPEAKER_00: And no amount of smart, no amount of smart tells you which want is the correct want.
• 4:24
  SPEAKER_00: Being smarter doesn't help.
• 4:25
  SPEAKER_00: The same way being taller doesn't help you seeing any better.
• 4:28
  SPEAKER_00: It's the wrong tool for the thing.
• 4:31
  SPEAKER_00: That's the whole thesis.
• 4:33
  SPEAKER_00: Every fight I'm about to walk through is a values problem wearing a costume.
• 4:37
  SPEAKER_00: And a machine can't referee a values problem, no matter how smart it gets.
• 4:41
  SPEAKER_00: Let me prove it.
• 4:42
  SPEAKER_00: Here's five fights.
• 4:43
  SPEAKER_00: Let's watch this pattern emerge.
• 4:46
  SPEAKER_00: Fight one religion.
• 4:47
  SPEAKER_00: And of course I had to go there.
• 4:50
  SPEAKER_00: People have killed each other over and over for thousands of years for religion's sake.
• 4:57
  SPEAKER_00: So somebody says, let the machine settle it.
• 5:00
  SPEAKER_00: We'll just figure out which religion is correct.
• 5:03
  SPEAKER_00: Well, think about what that even asks.
• 5:05
  SPEAKER_00: What faith is true?
• 5:06
  SPEAKER_00: Is there a God?
• 5:07
  SPEAKER_00: What happens after you die?
• 5:09
  SPEAKER_00: What does God want from you?
• 5:11
  SPEAKER_00: Now, would where would the machine get an answer?
• 5:15
  SPEAKER_00: It learned from human writing, and human writing about religion is a billion people disagreeing across thousands of years in every direction.
• 5:22
  SPEAKER_00: The machine read all of it.
• 5:23
  SPEAKER_00: It doesn't contain the answer to which God is real.
• 5:26
  SPEAKER_00: Nobody's writing contains that.
• 5:28
  SPEAKER_00: It's a question of faith.
• 5:30
  SPEAKER_00: Faith by definition is a belief in something you cannot prove with facts.
• 5:35
  SPEAKER_00: You can't measure it.
• 5:36
  SPEAKER_00: You can't test it in a lab.
• 5:38
  SPEAKER_00: There's no experiment that ends the argument.
• 5:41
  SPEAKER_00: So when you ask the machine which religion is true, it can only do one of two things.
• 5:45
  SPEAKER_00: It can dodge the question or it can choose one.
• 5:49
  SPEAKER_00: And the second it chooses, it's not being neutral, it's taking a side.
• 5:53
  SPEAKER_00: On the deepest values question a human ever asked, a machine has no business handing anyone that answer.
• 6:00
  SPEAKER_00: And it's an answer it doesn't have to give in the first place.
• 6:04
  SPEAKER_00: I'll say where I stand because I'm not going to hide behind neutral and pretend I'm a robot too.
• 6:09
  SPEAKER_00: I'm a man of faith, and that's exactly why this one lands so hard for me.
• 6:13
  SPEAKER_00: My faith is the most personal thing I own.
• 6:16
  SPEAKER_00: The idea that a machine trained on the internet would hand me a verdict on it, no, not a chance.
• 6:23
  SPEAKER_00: And if it shouldn't do it to me, then it shouldn't do it to the other guy who believes the opposite of me either.
• 6:30
  SPEAKER_00: Fight two.
• 6:32
  SPEAKER_00: Countries wanting nuclear weapons.
• 6:35
  SPEAKER_00: Case in point, nuclear weapon.
• 6:37
  SPEAKER_00: Let me spell it out fully because people throw the word around.
• 6:40
  SPEAKER_00: A nuclear weapon is a bomb that gets its power by splitting or smashing together the tiniest building blocks of matter called atoms.
• 6:49
  SPEAKER_00: That release of energy is so huge that one bomb can erase a city.
• 6:53
  SPEAKER_00: That's what nuclear means.
• 6:55
  SPEAKER_00: City ending power in a single weapon.
• 6:57
  SPEAKER_00: Right now a handful of countries have them.
• 6:59
  SPEAKER_00: Other countries want them.
• 7:01
  SPEAKER_00: The ones who have them tell the ones who don't, you can't have these.
• 7:05
  SPEAKER_00: Ask the machine to referee what?
• 7:07
  SPEAKER_00: Who's allowed to hold city ending power and who isn't?
• 7:12
  SPEAKER_00: That's not a truth problem.
• 7:13
  SPEAKER_00: There's no fact that says these specific countries are trustworthy and those who aren't.
• 7:19
  SPEAKER_00: It's a values fight tangled up in who has power and who's afraid of who.
• 7:24
  SPEAKER_00: The country that wants the weapon says you have one.
• 7:27
  SPEAKER_00: Why can't I?
• 7:28
  SPEAKER_00: That's not fair.
• 7:29
  SPEAKER_00: The country that has one says the world is safer, fewer people hold this.
• 7:34
  SPEAKER_00: Both are making a values argument, both think they're right.
• 7:37
  SPEAKER_00: A machine can lay off the risks, it can run the numbers on what happens if a bomb goes off.
• 7:43
  SPEAKER_00: That's the truth part, and it's very useful.
• 7:46
  SPEAKER_00: But the actual question, who deserves this power?
• 7:49
  SPEAKER_00: Who do we trust is a human values question soaked in fear and history.
• 7:54
  SPEAKER_00: No smart answer exists.
• 7:57
  SPEAKER_00: There's just what different people are willing to accept.
• 8:00
  SPEAKER_00: Fight number three, guns.
• 8:02
  SPEAKER_00: In America, we fight about guns harder than almost anything.
• 8:06
  SPEAKER_00: One side said a person has the right to own a firearm, period, to protect themselves and their family.
• 8:11
  SPEAKER_00: The other side says too many people are dying and we have to make it harder to get one.
• 8:17
  SPEAKER_00: So we'll ask the machine to settle it.
• 8:19
  SPEAKER_00: And here's the trap.
• 8:20
  SPEAKER_00: The machine can tell you facts, how many guns exist, how many people were hurt, what happened in other countries, when they changed their laws.
• 8:26
  SPEAKER_00: All of those are truth problems and they're all very useful.
• 8:29
  SPEAKER_00: So let's bring all those to the table, but the fight isn't about the facts.
• 8:33
  SPEAKER_00: The fight is about what matters more: your freedom to protect yourself, or the goal of fewer people getting hurt.
• 8:40
  SPEAKER_00: Both of those things people hold sacred.
• 8:43
  SPEAKER_00: When they collide, there's no formula that spits out the winner.
• 8:46
  SPEAKER_00: You're weighing freedom against safety, and how you weigh it depends on what you value.
• 8:51
  SPEAKER_00: The machine read millions of people screaming both sides, it doesn't contain the tiebreaker.
• 8:56
  SPEAKER_00: Nobody's writing does.
• 8:58
  SPEAKER_00: I'll wear my old job on this one because it's honest.
• 9:02
  SPEAKER_00: I spent a long career carrying a firearm as a peace officer.
• 9:05
  SPEAKER_00: I've seen what guns what guns does in the worst moment of somebody's life and about as close as a person can stand.
• 9:16
  SPEAKER_00: And I've also known good people who own firearms their whole lives and never heard a soul.
• 9:20
  SPEAKER_00: Both of those are true at the same time.
• 9:22
  SPEAKER_00: That's the whole point.
• 9:23
  SPEAKER_00: Even a man who's lived on both ends of it doesn't get a clean answer because there's isn't one to get.
• 9:29
  SPEAKER_00: In fact, it's not a fact I'm missing.
• 9:33
  SPEAKER_00: It's a value we're all weighing different.
• 9:37
  SPEAKER_00: Fight number four, race and the tension between people and the cops.
• 9:42
  SPEAKER_00: This one's raw, and I'm going to handle it carefully because it deserves to be handled with care.
• 9:47
  SPEAKER_00: People fight about race, about whether the system treats everybody the same, about police and the communities they work in, about history.
• 9:55
  SPEAKER_00: That's still sitting in the room.
• 9:57
  SPEAKER_00: And the machine to referee.
• 9:59
  SPEAKER_00: You ask it.
• 10:00
  SPEAKER_00: Now, remember the one fact I told you to hold on to.
• 10:03
  SPEAKER_00: The machine learned from human writing, and human writing about race carries every bias humans have, the kind we know we have and the kind we don't.
• 10:12
  SPEAKER_00: All of it went into the machine.
• 10:14
  SPEAKER_00: So when the machine talks about race, it is not clean, it's not neutral, it's a mirror.
• 10:19
  SPEAKER_00: It reflects back at us what we feed it.
• 10:22
  SPEAKER_00: And what we fed it was us, all of us, our fairness and our ugliness, everything mixed together.
• 10:29
  SPEAKER_00: So a machine that sounds calm and confident and says it's objective, talking about the most sensitive fight we have, that's not a referee.
• 10:36
  SPEAKER_00: That's our own bias handed back to us in a voice that sounds like it can't be wrong.
• 10:40
  SPEAKER_00: And that's more dangerous than a person with an obvious opinion because you can argue with a person.
• 10:46
  SPEAKER_00: You can see their bias coming.
• 10:48
  SPEAKER_00: The machine hides its bias, but behind that calm and confident math-sounding voice.
• 10:54
  SPEAKER_00: I'll leave my personal politics out of this one on purpose, and I want you to know it's on purpose, not because I'm scared to have a view, but because this specific fight is exactly the kind where telling you what to think would make me the very thing this whole show warns you about.
• 11:10
  SPEAKER_00: Somebody with a badge and a platform telling you the answer.
• 11:13
  SPEAKER_00: I'm like the referee.
• 11:15
  SPEAKER_00: I'm not.
• 11:16
  SPEAKER_00: Neither is the machine.
• 11:18
  SPEAKER_00: Fight five.
• 11:19
  SPEAKER_00: The have nots against the haves.
• 11:22
  SPEAKER_00: Money.
• 11:25
  SPEAKER_00: The oldest fight there is.
• 11:27
  SPEAKER_00: Some people have a lot, most people have a little.
• 11:29
  SPEAKER_00: The people with a little look at the people with a lot and say, that's not fair.
• 11:34
  SPEAKER_00: Ask the machine to settle it.
• 11:35
  SPEAKER_00: How much should one person be allowed to have?
• 11:38
  SPEAKER_00: How much should we take from the top to help the bottom?
• 11:40
  SPEAKER_00: What is fair?
• 11:42
  SPEAKER_00: Fair.
• 11:43
  SPEAKER_00: There's that word again.
• 11:44
  SPEAKER_00: It's not a fact, it's a value.
• 11:46
  SPEAKER_00: One person's fair is another person's robbery.
• 11:49
  SPEAKER_00: One person's freedom to keep what they earned is another person's greed.
• 11:53
  SPEAKER_00: The machine can tell you exactly who has what.
• 11:56
  SPEAKER_00: What's truth and it's useful.
• 11:58
  SPEAKER_00: Put it on the table, but the question of what's fair to do about it, that's values, all the way down.
• 12:04
  SPEAKER_00: No answer waiting to be found, just what people, different people, are willing to live with.
• 12:09
  SPEAKER_00: All right.
• 12:10
  SPEAKER_00: Five fights.
• 12:10
  SPEAKER_00: You see the pattern now?
• 12:13
  SPEAKER_00: Every single one, the machine can help with the truth part, the facts, the numbers, the what happened.
• 12:18
  SPEAKER_00: But bring the machine to the table for that.
• 12:20
  SPEAKER_00: It's a genuine tool and it works well.
• 12:22
  SPEAKER_00: Every single one, the actual fight is a values problem.
• 12:27
  SPEAKER_00: What matters, what's fair, what's sacred, and on what part the machine has nothing.
• 12:32
  SPEAKER_00: Not because it's dumb, but because there's no answer sitting out there to find.
• 12:36
  SPEAKER_00: Smarter doesn't help.
• 12:38
  SPEAKER_00: That's the wrong tool for the job.
• 12:39
  SPEAKER_00: Now, here's where it gets dangerous.
• 12:41
  SPEAKER_00: Here's the part I really need you to walk away with.
• 12:43
  SPEAKER_00: People are starting to believe the machine can be our neutral judge, the fair one, the one with no side.
• 12:50
  SPEAKER_00: It cannot.
• 12:51
  SPEAKER_00: And here's the exact reason one more time, and I'll give it to you slow for my benefit.
• 12:56
  SPEAKER_00: The machine learned from us.
• 12:58
  SPEAKER_00: Every value it seems to have it copied from human writing.
• 13:02
  SPEAKER_00: On top of that, real people at real companies decide what the machine is allowed to say and not say.
• 13:08
  SPEAKER_00: So when the machine gives you an answer on a values question, you're not getting neutral.
• 13:13
  SPEAKER_00: You're getting a mix of whatever it read and whatever those people decided.
• 13:17
  SPEAKER_00: Those wearing a lab coat, speaking in a calm, confident voice, sounding like math, and that's the danger.
• 13:24
  SPEAKER_00: Not the machine picking a side.
• 13:26
  SPEAKER_00: People pick sides all day, and you can see it.
• 13:28
  SPEAKER_00: You can argue back.
• 13:29
  SPEAKER_00: The machine, the danger of the machine that picks a side and hides it behind the word neutral.
• 13:36
  SPEAKER_00: A biased person, you can push back on a biased machine that everybody believes is objective, that nobody's allowed to question.
• 13:44
  SPEAKER_00: That sounds like a calculator.
• 13:45
  SPEAKER_00: That's not a referee.
• 13:47
  SPEAKER_00: That's a king nobody voted for.
• 13:49
  SPEAKER_00: And you can't vote it out.
• 13:50
  SPEAKER_00: So let me tell you what the machine is actually good for in our fights because I'm not here to trash the tool.
• 13:57
  SPEAKER_00: I work with this tool every day on my own work.
• 13:59
  SPEAKER_00: You bring up the truth problems.
• 14:01
  SPEAKER_00: How many, how much?
• 14:02
  SPEAKER_00: What happened last time?
• 14:03
  SPEAKER_00: What are the real risks?
• 14:04
  SPEAKER_00: It's fast, it's tireless, it can hold more facts than a human.
• 14:08
  SPEAKER_00: On the truth part, it's a weapon, so use it.
• 14:11
  SPEAKER_00: But the values part, what matters, what's fair, what's sacred, what we owe each other, that stays with us.
• 14:17
  SPEAKER_00: That's ours.
• 14:18
  SPEAKER_00: That was always ours because a value isn't something you calculate, it's something you choose.
• 14:24
  SPEAKER_00: And a machine doesn't get a vote on what you hold sacred.
• 14:27
  SPEAKER_00: Here's the honest bottom line, and it's not what the other side wants.
• 14:32
  SPEAKER_00: The machine crowd wants you to believe it'll save us from our fights and it won't.
• 14:36
  SPEAKER_00: Because our fights aren't about being too dumb to see the answer, they're about wanting different things.
• 14:42
  SPEAKER_00: The sacred crowd wants you to believe the machine is out to rule us.
• 14:47
  SPEAKER_00: It's not plotting anything yet, and maybe it never will.
• 14:52
  SPEAKER_00: It's a tool, but a tool people misuse can do just as much damage as a villain, and handing it the job of settling what we value is exactly the misuse to watch for.
• 15:04
  SPEAKER_00: The truth is in the middle, and it's simple enough for anybody to hold.
• 15:08
  SPEAKER_00: Use the machine for what's true, keep what matters in human terms.
• 15:12
  SPEAKER_00: And anytime somebody tells you the computer can be neutral, the neutral adult in the room, full of humans who disagree about what's sacred, you know now what to say.
• 15:23
  SPEAKER_00: And you know now enough to say no.
• 15:26
  SPEAKER_00: That's not how any of this works, and here's why.
• 15:29
  SPEAKER_00: You get it now.
• 15:30
  SPEAKER_00: This puts you ahead of most people arguing about it, and every one of the people selling it.
• 15:35
  SPEAKER_00: You be careful out there.
• 15:37
  SPEAKER_00: I appreciate you watching, and we'll see you in the next one.