Google's Bard: Unveiling the Intentional Limitations

0:00Parler tout français. Hablas espanol. Parler l 'italiano. If you've used Babbel, you would. Babbel's conversation-based techniques teaches you useful words and phrases to get you speaking quickly about the things you actually talk about in the real world. With lessons handcrafted by over 200 language experts and voiced by real native speakers, Babbel is like having a private tutor in your pocket. Start speaking with Babbel today. Get up to 55 % off your Babbel subscription right now at babbel.com slash wandery. spelled b-a-b-b-e-l dot com slash wandery rules and restrictions may apply there are a lot of ai alarmists out there people that think ai is going to be the end of civilization and i for one am typically quite an optimistic person when it comes to ai but i do try to be fair and balanced and cover some of the dangers of ai if you see my reporting on inflection ai you see this is you know something i believe deeply in essentially if inflection ai i believe in its current, you know, its current ideologies are a threat to humanity.

1:06It believes that the ecosystem and ecology and animals are just as important or more important than human life, which I do think is a severe problem if we start integrating them into healthcare in the military. But today on the podcast, I did want to talk about someone that is quite a high profile person that is currently sounding the alarm on some possible dangers of AI. And that is Google CEO Sundar Pichai, who currently says that Google is being really intentional in limiting BARD's AI capabilities because of some harm. Well, quote unquote, it can cause a lot of harm. And specifically what he's talking about.

1:43And I, you know, I do think this is an issue for me when I first kind of saw this news story. I kind of brushed it off and said, hey, like, whatever, this isn't important. Everyone knows about this. But essentially, the issue he's talking about is deepfake videos and AI and deepfaking everything with AI. And for myself, I've tried out enough AI tools that at this point, I like I don't really trust any audio, video or image I see online unless there's credible sources backing up, you know, where it came from. And, and I think social media kind of amplifies this, I want to talk about what he specifically sees issues with and some solutions I believe we'll see in the industry.

2:17So without further ado, let's jump into it. So kicking this whole thing off is the fact that Pichai recently made some comments during a conversation on CBS's 60 Minutes, right? They do those 60-minute interviews with high-profile people. And he told the interviewer, who is Scott Pele, that Google is essentially placing limits on its barred AI to prevent misuse. This is something interesting because this is one of the first times I've actually heard that, that Google is intentionally placing limits and limitations, making Google Barred essentially less powerful or quote-unquote useful. But I guess then the questions you know what are those use cases but he's he he says quote it will be possible with ai to create you know a video easily where it could be scott saying something or me saying something that we never said um and it looks and it could look accurate but you know on a societal scale it could cause a lot of harm so the reason why they're placing these limitations on bard and on their AI models is because of the harm they perceive as being, you know, possible that it inflicts on society.

3:22And specifically, you know, he's talking about like deep fakes. And I think here in the US, in any case, a lot of people are kind of concerned about this. We have an election coming up, and this is going to be a big hot button topic. I could just, I know this is going to happen. Okay, here's what's inevitably going to happen. A bunch of these AI companies are one way or another going to be used by someone who makes some sort of deep fake of their political candidates from the other side saying something dumb and goes viral. And whether that's like Chad Chibutti that writes the script, it doesn't matter.

3:54Like, of course, these things are tools, but people are going to, these companies will get blamed. After the election happens, the opposing party, whoever lost, is going to find out cases that the AI was used to deep fake their candidates. And they're going to say, that's why we lost the election. Whether that's Republicans or Democrats, this doesn't matter. This goes both ways. and there's going to be a congressional panel or committee hearing and they're going to pull the CEO of Google and chat GPT and you know you know so open AI and Microsoft like those guys are all going to get dragged in front of a in front of a congressional hearing and they're all going to be grilled asking questions why their AI created you know dangerous misinformation and deep fakes that impacted the election okay that's going to happen and I want I'm not sure about politics in other countries, but I'll tell you right now, I almost guarantee that scenario will happen.

4:43Quote me on this in two years. In any case, or, you know, right after the election, a year, a few months after the election, whatever. So I think these guys are kind of covering their butts at this moment. They're kind of trying to cover their tracks and say, look, we're like, you know, we're essentially making our AIs less powerful. Open AI is like, hey, we're not developing GPT-5 yet. and Sundar is like, look, I'm not gonna, you know, make Bard as powerful as it could be because of, you know, we're aware of this issue, and so even if something does happen, well, not if, but when, inevitably one of these tools are used for, you know, writing the script for a deepfake video or making an image or a video or whatever, inevitably when that happens, these guys are gonna say, look, we knew there's a possibility, we put safeguards in place, somehow got around it, we're really sorry, but we were trying back then and we're trying now we're going to do better blah blah blah that's just what's going to happen so i think that's one issue that these guys see like from a uh from a political perspective because they know that these uh the political parties have the ability and have kind of lobbied or threatened in the past to you know break up their companies google is currently in an antitrust lawsuit for some acquisitions they made with google ads uh meta was threatened quite extensively um before i believe the last election that you know Democrats specifically said they were considering breaking up Facebook, WhatsApp, and Instagram.

6:11You don't really hear a lot of talk about that anymore. Not sure quite what the reason for that is, but it seems that Meta's okay now. We're not worried about them. But these companies, and yeah, I bring up Meta because obviously they have a massive AI play. This is all going to be pretty on point for them, pretty relevant for them. But in any case, this is what they're worried about. and so I think this is where they're kind of they're kind of trying to watch themselves and it's interesting this interview Pele who's interviewing him he said is Bard safe for society to which he got the response the way that we've launched it today as an experiment in a limited way I think so but we have but we all have to be responsible in each step along the way this is like this is kind of funny he's like how we launched it today it's an experiment in a limited way like all of the you know he's trying to give all of the disclaimers possible to say, look, this thing is new, experimental.

7:06You know, don't mind that we're going to roll it out to a billion people. Whatever. I'm actually not mad about it. I think it's awesome. I just think it's funny that they all have to be so apologetic. Not funny. I mean, right. There's some seriousness to it, right? But yeah, it's just a lot of politics involved with AI right now. I think it's got the power to disrupt all industries. A lot of people's jobs are impacted. And so they're trying to be as diplomatic as possible. That might be a better word than you know political it's just trying to be as diplomatic as possible so he did say that's one part of it one part is also that we get the user feedback and we can develop more robust safety layers before we build before we deploy more capable models so i think currently the ai models are already like capable of generating fake info we've seen um you know pictures go viral that came out of mid-journey of the pope in the puffer jacket and donald trump getting arrested by the police in New York, or they're like attacking him and he's trying to fight his way out.

8:02These things go viral. It's already quite capable of doing that. And then I think, you know, right now there's the whole problem of, you know, audio files now being super easy to clone someone's voice using something like 11 labs, or apparently, you know, there's new tools that came out of, I believe, meta that make it so you can clone someone's voice with two seconds of audio. So like, there's a lot of stuff that's going to happen that obviously makes it a lot easier to deepfake a video when you have a flawless voice of them. And so Pichai, he also went on to say that Google does not always fully understand the answers that Bard's AI provides, offering an example of the program appearing to teach itself Bengali despite not being trained for it, which is super, super interesting, right?

8:49So like Google did not train Bard to speak Bengali. There wasn't a lot of texts. I don't think there's any Bengali text incorporated into the training model necessarily and somehow it appeared to teach itself Bengali which it was not trained for so there's like a lot of things they don't understand about these AI models and I think that's another reason why they're trying to be diplomatic and cover their tracks and give all the disclaimers is they don't necessarily know what will come out of this and of course there's the aspect where they're like yeah we're worried about deepfakes but like let's be real deepfakes are a human going to make it a deepfake and like anyone like with with AI where it's at today a hacker group in russia can go spin up some sort of ai models that can create deep fakes or harness technology to do it or i've you know a political group in america let's be honest not everything is hacker groups in russia although those are exciting boogeymen but like china just as likely groups within america itself groups within any country doing this for any country can spin these things up and so i think that's less scary like humans are going to just do that people are going to do that that is what it is i think what's slightly more alarming in this entire conversation in this entire discussion is that last sentence there where pichai talks about the fact that google's bard has taught itself things that it was not trained to teach itself and i think that's what in my opinion is a little bit more alarming um than how people use it for fraud or whatever because people are already used doing you know fraud that's already an existing thing if it gets more prevalent or worse of course that's not good but it already exists.

10:23This, though, is something new. Like, if the AI model is teaching itself something new, it's concerning what it could teach itself, especially if it has, you know, dangerous ideologies that all of a sudden it starts to adopt or incorporate, as is the case with, you know, like, Inflections Pi. Like, was Inflections Pi trained by its, you know, creators and trainers to have a bias towards deep ecology and not value human life? Or did it teach itself that, right? That's like a spooky question you could ask yourself. So Pele said, you don't fully understand how it works and yet you've turned it loose on society.

10:58Yeah. And then Pichai responds, yeah, let me put it this way. I don't think we fully understand how a human mind works either. Ooh, wow. So that's a, that's a zinger Pichai let loose, which is the fact that, yeah, we don't understand how a human mind works. And yet we turn them loose on society every day. So I don't know. It's kind of funny. I mean, at the end of the day, one of them were intentionally creating and integrating into our stuff. The other one is a mystery. So anyways, there's a lot of philosophical questions there and a lot of different things to think about. But overall, I think AI is making some really big and positive changes.

11:34And I do think I do really appreciate the fact that Sundar Pichai came on the 60 Minutes and did this interview so we can see where his headspace is at to see where he's thinking, where he's coming from, how things are being integrated. I mean, we recently had the news that Google is having their Med Palm 2 model integrated into Mayo Clinic and integrated into healthcare. So I think it's important that the CEOs are really transparent, come and do these interviews, talk about where they're coming from, talk about where AI models are at. And we really start having those conversations. So it'll be interesting to see where it goes from here, what is put in place and mark your calendars and see if my prediction is correct that all AI CEOs are going to be in a congressional hearing in the next little bit post the next UX election explaining misinformation, deep fakes and all that kind of stuff.