Politics of Gen AI | Marek Tuszynski [Tactical Tech]

No, I take it. Thank you. Hi everyone. So, um, if you never been in this space in Funk House before, this is your first time. I'm not working for Funk House, but I've been here for a few uh gigs, music and and stuff.

And um this space in particular is fantastic for very intimate, very high quality performances. And it doesn't matter if you like the music or not, just come for anything here and uh and listen to it. Um the acoustics are great and it's just fantastic experience. We don't do any justice here giving the talks uh to actually how fantastic this space is. And if you have a time to go around and see all the swall rooms and so on, just do it.

I'm just wasting my time that I have only 15 minutes. Where's the clock? You're not doing the clock. Thank you. Uh okay, cool.

So I will tell you uh few things. So there supposed to be a keynote, but I don't do keynotes. It's more about I will share with you a lot of resources. If you have your mobile phone on you, take it out. It's going to be a lot of QR codes.

Every QR code will lead you to something that is way much more interesting than my talking. So just you know dive in and go into rabbit horse. This is just tactical techure code where you can test if it works. Um I will divide it into three parts. Uh I will talk about me as us.

I will talk about them and I will talk about the bubble. And this is all based on the experience of running tens of different kinds of workshops in last five years trying to engage people in conversation about the politics of of AI which is very difficult um because we are all shaped by immerse uh information that we are in that is very you know polarizing confusing and and whatnot. There's a lot of light here that is completely killing me. Um, so me is not about me. It is about you.

It's about every single individual. And I want to go back to 1967 for a second. There was a guy called Joseph Bison that kind of, you know, created this bot called Eliza. And Eliza was, if you scan it, you're going to have a reenactment by whatever BBC or channel 4, I don't remember, one of the British channels, an interview of Joseph Viserbomb. So it was hard-coded bot that you know they were testing if you can emulate a human inter inter kind of reaction between the machine and and the and the person and they use a very old school therapist uh way of dealing with conversation with somebody that you want to open them but they don't want to open instead of giving new information to them you repeat the statement as a as a question or you repeat the question as a statement without adding any new information but that makes person to add more and more and more and more and what happened to vis was that The anecdote is that the secretary fall into the spell and she asked him to leave the room because she thought she's speaking to a you know super human embedded in the machine.

That wasn't possible then and it's never going to be possible. This whole dream of a golem that you can turn clay and mat or cables and chips into general artificial intelligence is not possible. So this is a kind of an example that turned completely vis thinking about how to engage people with technology where we instantly anthropomize technology uh whenever it is able to emulate something that we recognize as human even if it is not and that's the case of the of the AI and I want to take it forward but if you want to talk more about that I think uh Rehack from Visenbomb is here speaking later talk to him more about that if you like my experience running this uh workshop is that four years ago, three years ago, most of the people I was uh talking to and running uh you know different kinds of conversations were mostly using AI as a expert kind of you know tool for translation, transcription, little bit of coding those pre code and stuff and so on and kind of enhancement of the skills that is very hard to achieve otherwise you would have to study for a long time. You would have to pay for this expertise and now you click and it's there very quickly. You can edit a video even create it and so on etc.

what happened over time in the room uh that I have more had more people that were personally kind of in relationship with the tool that started for them as a as a expert and initially that was about you know kind of a therapist like Eliza check the Eliza effect is a thing uh or more and more I I had people that would sit in the room and converse with their board while I was talking to them to verify and check what I'm up to if you know if I what I'm saying makes sense and so on, etc. And that, you know, I thought, "Wow, that's weird. Maybe I'm a bit paranoid." And aren't we all a bit paranoid here, specifically in this room now today on Sunday? Um, but then I look up different kinds of research.

So, it really changed between 24 and 25 where you can see that therapy, companionship, organizing life and finding purpose in life become the major way of using uh this different tools. You can find this graphic on the I think it is a visual capitalist but is taken from Harvard business research. Um that was like okay this is not just me I'm not paranoid here. Then then came this Washington Post research when they got hold of 100,000 conversations between individuals and chpt um they got hold of it not because it was hacked there was a glitch. There's a lot of glitch in these technologies right now because they are being you know rolled out before anybody is testing them for for anything.

Um and they figure out of you know 47,000 they analyzed a lot of conversations 90% of them are personal personal advice kind of you know therapy and so forth etc. And that you know goes on and on and on and um rights and democracy just published a research last week here. So if Mark Bodal is here or anybody who done it uh for them they based in Berlin you should talk to them that is not only looking at this dependencies I just raised but it is also analyzing the you know dark side of it and they collect cases of people who you know suffer from different kinds of delusions and and much deeper dependency than they would like to. So that's for me like a one part of the problem. How do you talk to people uh about politics of technology of any kind and being critical where their relationship with technology is already so deep that you come across as somebody they shouldn't be talking to in the first place?

So I also look at um the research from Stanford. They just analyzed 25,000 videos from YouTube and from uh Tik Tok when they look at influencers on both side of the spectrum. They look at the you know the people who are resistors to AI and the people who are adopters and I was hoping that you know let's see how the influencers you know are engaging with AI in different kinds of ways and is kind of dissatisfying because the promoters are mostly focusing on mimes and uh effects and career and productivity that that we kind of expect uh but the resistors are mostly focusing on art creativity theft so copyright h IP and deep fake and misinformation very little about other things that we all care about if you like. Um so that's the kind of how people are using. I'm not showing you any politicians here.

I'm not showing you any other stuff that normally uh you know throwing at people in this conversation. But that's kind of us with our relationship. Now I want to talk about them. Um and you already seen some of these things here and Brad was talking about it and Tatiana um uh and so on. So I want to talk about who are they you know and because this is still this dream of Silicon Valley in the public sphere at least is very very dominant.

They are you know military contractors in the first place. um all of them. We knew that before. We knew that since Northern at least um that was very clear that every single you know big tech company had a contract with military industrial complex in the US at least in China. They don't have to because they part of the system anyway.

Um but that's a project from uh Sarak system u that was showcased in as electronica last year that actually meticulously goes after every contract that any and every of these companies across China and and US have with the military. Of course there's this entire debate right now of you know against anthropic blah blah this theatrics of who is good who is bad which is kind of irrelevant to where the contracts are who is paying for what and who is access have access to what kind of information and so on. They also are heavily influencing politics right now. I'm super happy that uh Molly White, who I follow for a long time because of her research into kind of a crypto and crypto bros and the money going to US politics and how Donald Trump is making money on the side on the regulation of you know finance and so on promoting crypto uh and people influence on you know on elections. She now expanded it to tech influence watch that is specifically focusing on the interplay between AI companies and and crypto together how they go align.

There's a lot of infighting between AI companies but there's a lot of synergies between the AI and crypto. I would recommend you just go there and and explore it. As I said every single slide is much more interesting that I'm talking about. Um I also look at okay so what's the relationship between these big tech companies that are promoting AI and big media houses no the the mainstream and there's a media capture watch another group of people that are doing really hard work in the in the back end that monitoring all the contracts and deals that the big companies are doing with with media by offering the software but offering the you know the training and so on etc. For example, Malta just signed a contract with OpenAI that every citizen of Malta will get access free access to premium CH GPT and they just require half an hour train from the company like we should be talking about this kind of things that are happening right now that are making very complicated to have any political conversation about this stuff.

If you want to see more about analysis of which is what I'm kind of in this rabbit hole right now. I'm looking at how big tech, big pharma, big oil and other companies develop a playbook of how you can influence narratives around you know what is actually happening, what they want to see that is happening or not happening in terms of impact on climate or you know what not and how they dominate uh policym through lobbying, how they dominate the narratives in the media and so on etc. And this research is actually quite fascinating because they listed this, you know, five categories of direct influence, conflict involvement, market influence, illusion of law, episteming and discourse influence into, you know, very nuance categories. But if you if you read them, it's kind of interesting because we see these things individually. If you work on privacy, you only exposed to part of it.

If you work on other things, you know, policies, etc., you exposed to to other things. And we have no collaboration across the sectors that would look at it more holistically. While from the corporate uh side they have a budget and they have a strategy and they go about it very holistically also collaborating with others. And that's just one of the cases.

If you want to see more of it, there's a fantastic investigation done in Brazil by uh Aensia Republica with 13 different groups around the world in 13 countries where they dive super deep exactly what the lobies are doing outside of Western Europe, outside of US and how they promote AI in particular uh and and politics around it. The last part I want to talk about and this slide kind of leads to it is you know I often hear let's leave it to policy let's leave it to regulation and Europe is often highlighted as the place when we are very good at regulating things. This is a slide I taken from Johnny Ran from the talk from Republic a few weeks ago where he basically is looking at since 2019 to 26 we have the GDPR in Europe and uh it's supposed to know the main goal of GDPR is to force big companies to change how they design the tools so they are you know privacy friendly they don't collect the data all these kind of things so they have to change the system design but to do that they when they don't do they should you know pay fines they should get you be punished and this is the you know the market cap who got the fines only three of them so far Microsoft Meta and and you know Tik Tok uh Tik Tok is here as well uh who paid nobody paid you know and when you look at why this is more politics you can look at the video um is because they all based in you know Ireland except Amazon who is based in uh in Luxembourg in in Europe and one of the you know three commissioners uh that are working on this issue actually is a former meta lobbyist and so on etc. It's a very convoluted you know kind of revolving door situation uh that makes very hard implementation and efficacy of any and every policy. So about the bubble and in my last three minutes the bubble what I want to talk about and you will have people here in this room talking about it later look for Alistister Alexander who did some analysis on the actual impact and how to calculate the cost of u deploy you know building AI infrastructure if you like but this research in particular is is is very thorough the problem of doing any research into impact of AI on anything is that there's no data you know when you build a data center the first thing you do as as a company, you sign nondisclosure non-disclosure agreement with the local government and there's no data.

So you can only infer information through comparing you know how it was before and how it is after the uh you know data center started and so on etc. I would recommend that one very heavily. When you talk about the bubble, I'm not an economist, but you know, this this infographic kind of very simple one came in November last year, but then the Bloomberg turned it into interactive that you can actually follow the relationship between Nvidia, AMD, OpenAI, Microsoft and so on, etc. What really caught by you know besides is the bubble that one is buying from another from another and then they're buying from another one etc. is based on the future promise of possible income that never going to happen.

is that the valuation you know 5 trillion 4 trillion etc is insane it's absolutely insane you know valuation of 5 trillion is equal to GDP of Germany um you know we have corporations that are based on fantasy that operate at scale that state operates these days and that have never been the case that has to go either way um the thing I usually do in the workshops and so on etc is I want people to focus on three things I want to kind of then walk away to understand the AI is just pattern recognition technology is not epistemological sematic machine. You can't gain any knowledge from it. It looks like it but it isn't. You know if you're the coder you know that very well but if you engage with people who are coming from different walks of life and study they don't know that because it gives different impression. I also like to talk about the data dependency.

The fact that more and more we use these tools that the more and more data is becoming synthetic and is poisoning you know the entire system. uh there's you know lack of you know there's going to be a lot of fight for clean and new data and the privacy is on the top of that and lastly you know it's a material thing this infrastructure requires a lot of renewal you can't build data center for once and paid whatever billions they pay for it you will have to replace all the chips and everything else in three four years because they you know run out of usability if you like um the last thing I I do is to kind of talk about this entire thing what happens when a human being is using AI as enhancement to be able to address a problem and find a solution. And there's a lot of debates, a lot of research about it. But there's this insistence on using AI as some kind of, you know, for granted type of technology. You can replace AI with, you know, let's just call it a military contractor software.

Let's just call it, you know, a software as a service um by and then name the companies however you like. Then you think about it differently. the the kind of semantic dis disconnection between AI that is something that you know they maybe discover on some astereroids now bringing in into use is extremely misleading and confusing and I think we have to talk about this technology by giving it the proper name and AI is not the right name for that um I want to finish with few things so this technology is definitely you know amazing for one discovery of one noility for one but for society for groups is the very opposite When we all use the same kinds of tools with the same results, you know, individually we think we are special. But from the perspective looking at all of us, you know, when everybody is special, nobody is uh special. Um it's not a tool, it's ideology, source of influence and power uh and is a ultrarocess information.

Ultrarocess information is ultra processed food is horrible. It has no nutrition if you like. So I think on the positive side they're losing hell a lot a lot of money. This is the project that is monitoring uh you know money invested and money gain. It's going to get worse and I don't think these big companies have ever been more vulnerable than they are right now.

And I think we have to look at it holistically and you know do something together. We know their playbooks. we know who they are and let's let's work together uh to use this technology for better things uh and make it you know created with different types of thinking about you know first of all what is the problem then find the technology if you need it and then fix the problem rather than take this technology to fix any and every problem that's not going to work. Thank you.