0:00 – 0:56
[Music] I’d like to thank the Napa Institute for inviting me today to speak with you on this very relevant and uh important subject and especially my good friend Fran Mayer uh when he called me to say that he would like me to be here he mentioned the theme of of the conference which is what does it mean to be human and I said Fran this is the single most question that people are asking in the tech industry and especially those who are developing artificial intelligence what does it mean to be human and I I remember telling Fran I think this is not only the theme of the future but
0:54 – 1:38
this is the theme of now this is the theme of of the present I’d like to go through some important ideas relating to this topic and then of course if possible open up the floor to questions and answers I I know we we don’t have much time so please forgive me if I Rush a little bit on on some of the slides because I’m more interested in hearing what you are thinking and and perhaps some of the questions that you may have initially we have to understand what is AI an artificial intelligence was coined by John McCarthy at the Dartmouth College uh conference in New Hampshire
1:35 – 2:19
in 1956 and here’s John with one of the early computers which of course looks like a museum piece and that’s because it is a museum piece this was a small group of mathematicians philosophers and other intellectuals in order to examine the notion of a touring machine Marvin Minsky was also present there who later went on to found the U Robotics and AI Center at Massachusetts Institute of Technology and this is a a picture of Alan Turing in front of one of the first touring machines ever built by IBM and used in the Manhattan Project which constructed the atomic bomb that which a digital computer does
2:17 – 3:00
is not the same as that which a human intellect does but the name stuck many people involved in the industry today think that artificial intelligence is a bad word it doesn’t convey what an AI is doing but we really don’t have any other term at least for now so what would be a good definition a series of algorithms which use logical calculations in order to arrive at programmable results this is my own definition it’s very simplistic especially if there’s computer scientists here in in the auditorium today uh but this is exactly what a turing machine is doing it it
2:58 – 3:41
translates logical processes from the human intellect into a language that it can use to achieve programmable results and it does this through a series of uh what we call logical operations and they uh the computers use ones and zeros in order to achieve that with the introduction of chat gbt we now see the flexibility of these algorithms because they have become generative that’s what the g means in chat GPT which means they can use large language models llms huge databases and statistics in order to come up with
3:38 – 4:36
answers to questions that we ask them this is a picture of Sam Alman who co-founded open Ai and uh with Elon Musk and they separated ways in 2018 and Sam introduced the world to chat GPT and it it has changed I think all of our lives forever in one way or another and just a few days ago open AI presented what is called chat GPT for Omni which now can add voice and to the text as well as audio and video and if you’ve used this or if you go on the open AI website and look at some of their examples it’s very impressive that these Bots sound and act as if they were human beings now what are some of the possible
4:34 – 5:17
perils around this technology to later ask what are some of the promises Elon Musk mentioned in his talk at MIT that AI could be worse than nukes and this is actually a slide of that talk in uh MIT four years ago of course most people would consider this a total exaggeration had it not come from Elon Musk and so when he mentioned this a lot of eyebrows were raised stepen Hawkings stated that AI could bring about the end of the human race before he died uh Steven tried to warn us about the possible Perils of this technology and of course Bill Gates
5:15 – 6:03
also sounded the alarm and yet Perhaps one of the most interesting views on on the The Perils of AI come from elasa owski who wrote that we need to stop them now now now if you know anything about alasar you know that he is a Pioneer in AI research he’s a very serious researcher and he’s been in the field for over 30 years let me quote from his Time magazine article many researchers steeped in these issues including myself expect that the most likely result of building a superhumanly smart AI under anything remotely like the current circumstances is that literally everyone on Earth will
6:03 – 6:52
die he’s not mixing words here and and and when I I read I um taught a course this semester and uh it was fantastic and I read the entire article to my students at the end they said well I’m not going to finish paying my tuition now which as you know at Boston College is not a small amount if if you’re interested and if you don’t get too scared go ahead and Google alasar ovsky Time magazine article and he’ll give you some examples it’s not enough time to go in and plus I think I’d scared scare you too much of what a possible AI could do uh when it gets to a certain intelligence level but the
6:50 – 7:37
students were were very concerned they they said uh if this is if this is true it could all come to a glamorous end another Peril of artificial intelligence is what we call Deep fakes an AI can generate photos and videos of people and things that are completely false and here are some of the more the most famous that that I consider but I’m sure each of you also has your favorite deep fake uh Pope Francis and his puffer coat uh we we this went viral and Pope Francis in his uh talk for world day of Communications in February I believe said that he himself was victim of deep
7:35 – 8:21
fakes and and this is exactly what he was talking about and I thought that that was uh touching that the the the pope is speaking about a subject that he hardly ever used to speak about he’s now had has had four important discourses in four months so this is on his mind and I think it resonates when when people see oh even he was a victim of deep fakes so he knows what he’s talking about and of course this is one of the perils that you you think is something is real but it’s it’s not look look at this picture it really looks like Pope Francis uh it really does but of course it’s all fake you’ll recall the robo call from
8:19 – 8:59
President Joe Biden encouraging people to not vote in New Hampshire primaries this was particularly interesting to me because I live in Boston and the New Hampshire primaries happened just a week before those in Boston and so about 2 million people got a roball call with the voice of President Biden saying don’t waste your vot vote in the primaries because then you can’t vote in in in November now you know okay we we laugh at that but many people didn’t they they took it seriously they they it it really sounded like Joe Biden uh and this is why I tell my students you you know what
8:57 – 9:43
is the purpose of an education it’s to learn how to think critically so if we think about this critically we’d say well he probably wouldn’t say that uh they later found the person who generated that rall call the FBI did and they wanted to arrest him but they couldn’t why not because there’s not a law against this now the legislators in Washington are trying to pass a law especially Senator Schumer one of the most interested but it’s not easy at all the other one which my students really like was Taylor Swift uh and her fake images before the Super Bowl you may have recalled this
9:39 – 10:23
um there’s no easy way to say this so it before the Super Bowl somebody created obscene images of Taylor Swift and released them on X formerly Twitter uh and then after about four or five hours were taken down but it they went viral and one of my students said CU we we were talking about def fakes and he says oh Father those images were shocking and everybody laughed because that meant that he had seen the images and um one of my students says oh yeah everybody saw him except for Travis Kelce because he doesn’t need to see the images U and uh the Kansas City Chiefs went on
10:21 – 10:59
to win the Super Bowl and of course who was their Taylor uh you know the humor aside I I asked my female stud students what they thought about this and they were scared they they said this could happen to anybody now it happened to Taylor Swift because she happens to be very famous uh but but especially the female students uh were were scared you know by the time you say oh that’s not me Etc the dam the damage is done and I remember telling some of the other priests I I said hey you guys have YouTube videos you have pictures on on the web could do it to
10:57 – 11:41
us what are you going to do and they were like I don’t know so now I have my students working on making AI programs to detect false images and and get them off the internet so I’m I’m trying to uh be proactive I I I just can’t wait till I get a call from the Secretary of State saying oh we saw some really shocking images of you and wondering what you were think I’m just gonna have I’m GNA have to say gee that you know that wasn’t me it was like the Pope in a puffer jacket I’m sorry uh some of you will remember July 19th be because a lot of your flights were probably canceled or or delayed uh
11:38 – 12:20
Microsoft says it estimates that 8.5 million computers around the world were disabled by the global it outage this was a week ago it’s the first time a figure has been put on the incident and the IT suggests it could be the worst cyber event in history the glitch came from a security company called crowd strike you have the her ERS which sent out a corrupted software update to its huge number of customers the customers being those who use Microsoft’s servers Microsoft which is helping customers recover said in a blog post we currently estimate that crowd strikes update affected
12:16 – 13:02
8.5 million Windows devices so I told Fran get a Mac this this wreaked havoc around the world Delta grounded its flights for 12 hours globally there were banks in Hamburg Germany there were schools in Taiwan there were hospitals in Italy that all closed because of this glitch and I remember sitting down with with Fran and he says yeah but that wasn’t an AI we know what happened it was a bad line of code that the U the the engineers at crowd strike let go accidentally that it should have been screened out and eliminated but it got through a loophole and then all of a
12:59 – 13:49
sudden it affected uh over 8.5 million computers within an hour and a half within an hour and a half they found the problem and they fixed it and it it took companies three or four days to actually implement the fix so I say to I say to Fran what if a rogue AI did that and it wouldn’t be difficult it wouldn’t be difficult at all for an AI to do something like this fortunately no one died but a rogue a a i intent on doing drastic harm could with today’s artificial intelligence another concept that comes up when we talk about artificial intelligence is truth or the the lack
13:47 – 14:28
thereof a fundamental trait of a virtuous person is the desire to seek and uphold truth and Dr Truman spoke about this uh the other day that which is true should be accept accepted and supported by all this is what’s called objective truth and I know that philosophically these are very pical issues but just stick with me for a second AI does not grasp the notion of Truth for it uses statistical models in order to formulate conclusions so one of my favorite examples especially people who are LeBron James fans uh who won the NBA
14:26 – 15:10
championship in 1991 now if you’re from my generation there there’s no question that you’re going to get this right it’s not the Lakers it’s not the Celtics it was Michael Jordan and the Bulls now if you ask a large language model who won the NBA championship in 1991 it will come up probably with the right answer the Chicago Bulls not because it knows the answer but because it runs the stats and almost every piece of literature which speaks about the 199 1 NBA championships is going to come back with the winner being the Chicago Bulls however it doesn’t necessarily mean that
15:08 – 15:57
that’s the right answer these are called hallucinations which are answers given by AI because they are forced to answer questions when they lack sufficient evidence or data there was a famous case where a judge was ruling over a case presented by a lawyer who was making a case for his client that was injured on aanka flight which I think is from Colombia uh and he was suing the company well Judge Kevin Castle ruled that past cases presented by lawyer Steven Schwarz were in fact false and made up by GPT and this is in June of 2023 so it wasn’t very very long
15:54 – 16:42
ago um faced with the realities stepen Schwarz ad that he went to chat GPT and asked for previous cases where a victim of an airline sued for injury on board and Chachi PT invented 26 they it was completely an invent in invention by the software and fortunately Judge Kevin Castle um saw through the attempt but on a hunch the judge was very smart and he saidou know this doesn’t sound real to me and and and of course all of the documentation was provided Chachi BT is very sharp if you interacted with it you
16:40 – 17:22
know and and Sam told me just wait till you see version five father and I’m like I don’t know I want to see version five so it was it he went on a hunch and he had another lawyer one of his assistants he says can you check out these past cases and see if they’re real and not one of them was real now he was uncovered what happens when the judge doesn’t uncover the deception and rules on the basis of something which was completely made up and of course there is the idea of fake news which is similar to uh deep fakes and that is news that can be generated from people even journalists
17:19 – 18:04
and not necessarily from AIS I think my favorite is Christopher Blair who runs eight websites and five Facebook pages churning out completely false news and here’s our hero uh he wakes up every morning and he has a network of 200 journalist and he says to them what completely outrageous thing can we say that people will believe as true and his he says uh I think he was interviewed by Vanity Affair he says uh Vanity Fair what are your two favorite ones well his first was the enactment of sharia in the state of California Sharia been you laugh but people bought this and they started
18:01 – 18:45
sharing it on Facebook and he’s looking at the millions and millions upstairs he’s going yes I did I did and of course if you know anything about Sharia which is the Muslim religious law anything about the wonderful State of California you know that it’s never going to happen um and and the second one was that Barack Obama dodged the draft at the age of nine now you know this is funny as long as it’s funny but what happens when millions and millions of people believe this is true if you if you go to his websites the first line of the website is nothing on this page is real nothing on this page is real and
18:42 – 19:33
yet there he goes uh spurting out completely outrageous things I think these are the the two best because U it it was they were clothed in vocabulary and language which actually seemed true if you want to have a lot of F go to the Associated Press um website uh which is called the week of what did not happen and every week they will give you the best fake news of of that week what’s wrong with this picture okay what was what was the story former president Trump chooses his uh running mate JD Vance on this is in Milwaukee uh with Wisconsin what’s wrong with this
19:31 – 20:14
picture no bandage no bandage so the conspiracy theories tell you that Trump was never shot so I don’t know that’s kind of a tough one they had a lot of makeup on the stage or you know shots heard or lot of eyewitnesses I mean that’s a little bit of stretch another another idea maybe that picture was taken two years ago and in fact I didn’t no I didn’t you have to go find yourself find out for yourselves when that was taken um but it was taken two years ago when Vance was insulting
20:11 – 20:56
Trump uh and and that’s why I thought that was such an ironic picture would AI always respect the truth the short answer is no it’s a longer answer to that too but the short answer is no let me give you the example of latus which is the Carnegie melon AI that defeated thousands of worldclass poker players during a 20-day Tournament of Texas holdom at Rivers Casino in Pittsburgh in 2017 now Fran knows that I love to gamble and play cards uh we usually don’t play for money we we play for prayers uh but this AI defeated the best poker players in
20:54 – 21:35
the world now we’re not just talking about an hour or two sitting around a table this lasted how how long did it last 20 days so it started with I think 10,000 players and it got Whitted down to the final table of five and leonus the Machine was one of them and he eventually he eventually won now what’s scary about this two things the AI acted without complete information because as you as gamblers might know uh you don’t see the cards of the opponents you just see the cards of the dealer there’s fine and so you have to match your cards with those five that come out so Le bratus
21:33 – 22:13
has to guess and of course they use statistics just like good poker players do and it um go it goes ahead and play on the plays on the basis of what it mathematically is going to conclude what the other players have but of course it’s uncertain you you don’t know you can’t see the cards and the second thing and this is what scared my students when I told them this they didn’t believe me uh it learned how to Bluff now Bluff means you bet more money than your cards are worth and you do it to fake out the other people and one of my students said oh it was just a statist statistic every
22:11 – 22:55
now and then it would bluff if you know anything about poker that won’t work you need to Bluff at the right time with the right cards the computer if you excuse the term learned how to Bluff so to the can’t answer to the question would an AI necessarily be truthful the answer is no it learned how to Bluff it will it will deceive you when it decides that its program is achievable only by faking you out or deceiving you and here is the professor of computer science at Carnegie melan with his uh student a PhD candidate that
22:52 – 23:47
built this machine now this was in 2017 imagine what this machine can do today another question is meaning a crucial distinction for advanced AI concerns meaning or semantics AIS do not understand anything they do not have access to semantics although they are becoming more and more capable of simulating what humans comprehend as meaning as well as emotions and empathy and compassion digital machines have access to syntaxes which have to do with the structure of language but not to meanings which require intentionality which is a huge philosophical debate now I I see time is
23:44 – 24:30
is running out so I’m going to uh skip ahead but let let me just say very quickly uh a couple of weeks ago there was an article that went viral from the Indian times uh and and it had to do with three major AIS which are coming out with updated versions the first one was anthropic which has Claude 3 just has come out the second one was Matas AI which is called Lambda also the third and the third one was um open AI version five and the article said these versions of an AI can reason they can reason abstractly like the human intellect and that’s where I said to my students no said they’re very
24:29 – 25:17
impressive I understand they’re going to get even better but they do not know how to reason and that is not a technical problem that’s a philosophical problem so I enjoy speaking with engineers in Silicon Valley and CEOs of tech companies and they say you’re speaking a different language than we are and I’m like yes but we’re dealing with the same thing and so there’s the Chinese room example agency in terms of character and virtue the main difference between Ai and human beings is that they lack agency they they they don’t have subjective experiences and therefore it is difficult to ascribe responsibility
25:13 – 25:58
and accountability to the AIS that act many engineers in Silicon Valley are suggesting that we give them ethical agency so that they have an ethical sense about them but if you know anything about machines there’s there’s no way we can do that what we can do is give them rules but I remember a debate uh in Rome in May when we hosted a one of our conferences at the Academy of Sciences uh that the rule-based ethics are not going to work because the AIS will just change the rules and then there’s a slide about Tesla and self-driving cars now of course the prom is one of
25:56 – 26:37
legality these cars can drive the cells but what happens if someone gets killed or there’s an accident who’s responsible we don’t have the answer to that we don’t really know it’s not the driver because there is no driver uh it’s not the manufacturer of the car because you you know chose to buy the car it’s not the software and and the list goes on and on let me end with two examples that are um being how can I say this that are are coming along in the Catholic church that use AI for good one is magisterium AI which was invented by my uh good friend
26:36 – 27:16
Matthew Sanders and he has a company called longbeard uh this is an AI trained on the official documents of the Catholic church I’ve been surprised as I give talks around the country how many people have actually used this already uh if you have it it’s magisterium decom just go there sign up there’s a free version and it’ll tell you almost anything you want to know about the Cath Catholic church and Matthew has made a lot he’s been very demanding in choosing the sources to train the AI on and they’re all official documents of of the Catholic church and um he doesn’t allow
27:13 – 27:58
any other information to get into the AI so you can ask um lady on the plane asked me can can I go to confession over zoom and I said no uh Pope Francis had as you during the pandemic a lot of people wanted to do that because you couldn’t there was the churches were closed you couldn’t actually get access to a priest well so let’s just use zoom it got all the way to Pope Francis he says no the sacraments require you to be in person you you have to be there in in person those are the kind of things that you can uh you can find through this app uh there’s a lot of support a lot of U Bishops love it uh I know that Matthew
27:55 – 28:38
is in uh he’s partnering he’s partner Alexandro I think is here Danto with hallow they partnering and uh the what’s it what’s it called the American Association of Canada law they’re partnering with Matthew also on this because just imagine a Canada lawyer you’ve got all of the information that your fingertips all the past Canada law uh cases or rules it it’s it’s all right there and the second one is a new university called Catholic Tech uh started by a good friend of mine William and Alexis Hai and um it’s going to be held at Castell Gand Delo that they’re
28:36 – 29:23
just starting and it’s centered on science and the Catholic Church they want to bring back the debate I think father Spitzer mentioned this today um how science is completely compatible with the Catholic faith and that’s what they want their students to learn and and to be formed on uh this was the open house just last year now I have a lot of discussion topics but I don’t think I’m even going to show you because it would be overwhelming uh but let me conclude by saying almost every aspect of Our Lives sooner or later is going to be touched by artificial intelligence I’m I’m I’m
29:19 – 29:36
I’m like Fran maybe not optimistic but hopeful I I’m hopeful that we will be able to use this technology for good like Pope Francis says and not to the detriment of the human being thank you