YouSaid · the spoken record

Brian Christian

lines on the record
64
first
2019-07-30
most recent
2019-07-30
sittings or episodes
1
sources
podcast

Every line below is reproduced as it was said and linked to the record it came from. Nothing here is summarised or generated. Directory · Search · Corrections

  1. “Yeah, and every time it would then say something to you, you'd react and it would tag that in the database as this is a human reply to this sort of thing. And you would essentially combat these bots in a totally different way. If you were trying to explore the breadth of knowledge, you would be amazed at what they could competently respond to. It's basically anything that someone on the internet talked about over an 11 year span is in there. So if you say, what's the capital of France? It would say Paris. If you said, what's 2 plus 2, it would say four. There's an example that I give in the book, which as I asked it, what is the capital of Romania? And it said Budapest? I don't know, which and the real answer is Bucharest. And so that's the kind of graceful degradation that you expect from a person but not a computer. And if you gave it popular culture things, if you said a line of Bohemian rhapsody, it would.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  2. “It into this kind of giant compendium. So I would say the software program in the 2000s that really exemplified this is what's called Cleverbot, where it just hung out on the web talking to people. And the ingenious trick that the programmer used was that everything you said to Cleverbot was going into the database of things that humans say. So if you show up and you're like, howdy, how's it going? It tags that in its database as this is a conversational opener that humans tend to use.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  3. “Program might demonstrate pretty impressive competence within a prescribed kind of channel of speech. As long as you kind of follow the predictable path through that subject matter, you were likely to encounter things for which the programmer said prepared. But if you tried to push against that envelope, the only possible thing that would happen to you would either be you'd get just nonsense or you'd get this very noticeable attempt to direct you back to the kind of main thread of conversation. That started to change in the 2000s because of the internet. Suddenly it became possible to approach the writing of a chatbot program not as if you were writing a choose your own adventure book where you had to manually key in all of these different conversational branches. But to think about it more like something like a wiki where you can kind of scrape text from the web assembly.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  4. “So, I would say there's at least one big change between the early 90s and when I participated in 2009, and then at least one major change since. So one of the biggest differences between the early 90s and the late 2000s was that in the early 90s, there was this idea that you had to make the conversation topic specific. The original Loebner Prize competitions kind of handicapped the human players by forcing every conversation to a predefined topic as a way to make it just a little bit easier for the programmers. And this led to very strange judging decisions of if you're ostensible topic is hockey, can you talk about the Cold War implications of the famous 1980s US versus Soviet Union Olympic hockey game? Are we now into politics? So it was, of course, very difficult to police the boundaries of what was or wasn't on topic. There was sort of a classic signature that”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  5. “From the perspective that I was not only competing against the machines, but I was also, in fact, competing against the other real people. And it opens up these bizarre and to me fascinating questions of what do you actually do from a tactics perspective if you have five minutes to act as human as possible in a competitive scenario? What does that actually look like? And so for me, that's a case of this philosophical question where the rubber is hitting the road. What are the distinctive qualities of the human mind? What is inimitable by software, at least at the current time? And that all has to operationalize in an actual strategy you're going to bring to what seems like this kind of high pressure small talk scenario.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  6. “There's this very bizarre aspect to the competition which you're highlighting, which is that after each conversation, the judge will not only make a binary decision about whether they thought they were talking to the human or the computer, but they will indicate a degree of confidence and basically say on a scale of zero to 100, how sure are you in making that determination? And so every year there is a computer program that gets the highest confidence score out of all of the other computer programs. And you win this title called The Most Human Computer and you get a bronze medallion and a few thousand dollars. But there is also, of course, every year a real person who gets the highest confidence score out of all of the real people. So they were better able to persuade the judges that they were a human being than the other human beings. And this person wins the so-called most human-human award. And so when I learned about this, I was riveted. And indeed, I found myself kind of approaching my role in the competition.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  7. “sequestered in a room down the hall, with five minutes on the clock to persuade a panel of judges that I was the human being, not the mechanical imposter.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  8. “Selling plastic portable roll-up lighted disco dance floors and decided to bankroll this Turing test competition and name it after himself. So it's called the Loebner Prize. Starting in the 1990s, they actually started doing these competitions. They'd convene people. They'd have people sending these texts back and forth for five minutes. And the Turing test threshold of 30% was not crossed by the year 2000. But I got very interested in the year 2008 because the top computer program at that year's competition managed to fool 25% of the judges. So it was just a single vote short of hitting this famous threshold. You were seeing newspaper headlines at the time like humanity dodges a bullet. There was this thinking that perhaps the next year would be the one that machines finally breach that. And so I got involved in my first book by being one of the human participants in the 2009 competition. So I was the guy.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  9. “That we're sending text messages back and forth with the computer, and the computer claims that it is the random human being hidden in a room, and that the other person is the computer. Of course, the human is claiming the same thing. Turing says that, in his view, by the year 2000, he predicted, we'd reach a point where after five minutes of interaction, the program could fool a panel of judges 30% of the time. And that as a result, we would come to speak of machines as being intelligent without expecting to be contradicted. He wasn't making a metaphysical point. He was making a cultural point. And this is one of the famous predictions of computer science that didn't come true. So starting in the early 1990s, a series of annual Turing test competitions were inaugurated, famously by this quirky New Jersey millionaire who had made his fortune.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  10. “I think the Turing test, I mean, Alan Turing was, of course, one of the great founding fathers of computer science and was amazingly prescient in thinking about some of these philosophical questions decades before other people were thinking about it. Already by 1950, he writes this famous paper called Computing Machinery and Intelligence. And the opening line is, can machines think? Of course, you have to remember this was back in the time when the computer, as we know it, was this room-sized thing processing punched cards. But he was already asking these philosophical questions. He was kind of seeing the long view ahead. And he answered the question not as a philosopher, but as an engineer, he said, I don't know, I don't purport to know what thinking actually is. But let's just imagine a practical test that we can run. We will sequester a random person in a room down the hall, and we'll be sending text messages back and forth with that person. We'll also have this computer.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  11. “Right, okay, so AGI is artificial general intelligence, which is a fairly recent term, but the underlying idea has been around for a long time. It basically means a system that is human-level competent or beyond at not just a specific thing that humans do, but basically everything that humans do. So we've long had systems that could outperform humans at arithmetic since the 40s, and then things like chess since the 1990s. And we're seeing these particular domain-specific milestones kind of fall into the rearview mirror, more recently Go, for example. But there's still this question of the kind of flexible general intelligence where you can ask someone to write you a poem, you can ask them to add some figures, you can ask them to ruminate on their childhood. What do we need in order to build a system capable of competence with that degree of flexibility? So that's the question of AGI.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  12. “I think it's clear that we're on the road to building AGI. And the open question that I think is on the minds of many of the people in the technical research community is how many significant breakthroughs away are we? I was just at an AI conference a few weekends ago and we were talking about this. And I said, I think we're zero to two major breakthroughs away from AGI. And that got a general consensus of, hmm, and nodding. And I think that's in some ways the holy grail. That's the biggest question hanging over the field. So I think that Is going to be one of the great things that we'll learn in the next couple decades.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  13. “For actually implementing AI, which maybe shouldn't surprise us that we're rediscovering what evolution found after millions of years of trial and error. Computation gives us a way of asking questions about what does it mean to think, what does it mean to have a mind, what does it mean to make rational decisions, and I found as I was studying these two things in undergraduate, the questions that I was interested in asking in philosophy about the mind were in some ways better answered through the tools and the vocabulary of computer science than they were in philosophy itself. So that's the sort of collision of those two fields has been kind of the overriding interest of my whole professional career.”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source

  14. “I think I have this very on paper eclectic academic background where in college I studied computer science and philosophy and then I went to graduate school for creative writing and I remember at the time raising a lot of eyebrows at family reunions and so forth people saying so you're studying computer science this very precise exacting engineering discipline discipline discipline and philosophy like this abstract vague field how do these two things connect increasingly people don't ask me that question and I think we're perhaps just coincidentally living through the great synthesis of computer science and philosophy that as the sort of computational metaphor for mind becomes increasingly literal we've developed these neural networks for example starting in the 1940s through this analogy to the way that the nervous system worked and are now discovering in the last 10 years that this is the best mechanism that we know of”

    2019-07-30 · Invest Like the Best · Brian Christian – How to Live with Computers - [Invest Like the Best, EP.140] · IDENTIFIED FROM THE TRANSCRIPT · source