Latest / Beyond the Slop: AI in the Non-Profit Sector / Episode 6: Will public benefit funds replace charities?
Transcript
- Craig: Hello and welcome to Beyond the Slop, the AI focused podcast for fundraisers and charity workers. I'm Craig Linton. Anna: Hessenbrook. And this week we are talking about what we've been building over the last week, as usual. and we're also gonna be talking about our oops AI did it agains for the week. But Craig, what's our main story this week? Craig: the main story we're going to take a look at the anthropic announcement that they're going to invest in the public benefit and hopefully this utopian world of universal basic income where we're all free to do and live wonderful creative lives because AI is doing all the work and creating this mass abundance. of course. Before we get into that, tell us a little bit about what you've been up to, Anna, and what you've been building this week? Anna: this week I actually have mostly been looking at compliance and permissions and approvals and what I've called previously mission fit. I've been doing a lot of work with charities who really want to get their rules and permissions and settings set up. And that's talking about, okay, people are actually using AI now, people are actually starting to get into their use of their licenses and their access. They're starting to do a bit of training to make sure people are getting the most out of it. now we really want to know how deep should we let AI into our processes? Which folders and systems should we give it access to? Should it be able to just read or write or all these really interesting use cases? and then Craig, you and I we were also on a AI superusers roundtable earlier in the week with some charity buffs who are using AI in really interesting ways from across the sector. that was really interesting and exciting as well. What about you, Craig? What have you been up to? Craig: Yeah, well, I'm going to talk about a book in a second, but just on that last one, it's great to be part of that roundtable and hopefully we may have some future guest appearances from some of the people because they're really doing some very exciting and interesting and creative stuff using AI, which... I learned a lot from and I'm sure our listeners could as well. So watch this space for that. I had the pleasure of speaking at Rob Woods's Brightspot Community about AI this week. I'd did it a year ago and I've refreshed the presentation and redid it. It was really interesting. There's about 30 people there, variety of charities. few things came up that we've talked about before, but... worth thinking about and reminding of is that Co-Pilot is their tool of choice, they're forced to use it. Still lots of people going rogue and working outside the formal systems, which I find really interesting. Couple of charities using Google and Gemini, but then... like other things, not allowed to use gems, which is probably one of the most helpful things within Google Gemini. So again, using AI almost with one hand tied behind your back. It's this in between space where, yeah, we can't ignore AI, so we'll give you a little bit. We won't give you the full bit. no wonder people find things a bit, meh, well, look, I don't get it because you don't get the full experience. It'd be like, I don't know, going to the theater, but you have to wear a blindfold or you've got earmuffs. you're only getting part of the experience. So that was really interesting. And one of the things I talked about in the presentation, one of my favorite writers on AI is a guy called Ethan Molyk. He's got really good sense of humor. he talks a lot about potential of AI but he's got a new book out this week called Coexistence I've just got it I've read the introduction I've not read the full thing yet but I'll report back but he wrote a really interesting sub stack just about the difference between his first book and this book and just the change that's come on so there's a really funny video about Otters on the Plane completely made with AI and he said that test is dead now because it can do better than actually getting an otter on a plane. So that was interesting. He talked about this acceleration and I thought this was really interesting. In 2023, OpenAI's GPT-4, which was this big breakthrough in the models, was the best model in the world ranked for almost a year, 352 days. And then since that, OpenAI, Anthropic and Google have... just chopped and changed and there's been 20 new leaders since 2024 and this year alone I think one, two, three, four, five, eight already. they're constantly fighting and improving all the time, which I thought was really interesting and the change is exponential. So when that first book was written, the most day I could work for was about four minutes. Now, 17 hours, AI can work by itself, just left to it on one of these normal chatbots. This was using Fable on Claude. So four minutes of work to 17 hours, that's incredible. And the cost of answering a really complicated, I think it's like a PhD science level question, 25, that was costing about 30 cents a question. Now it costs, four thousandths of a cent. So it's 400 times cheap Anna: Mm. Craig: than it was three years ago. So the cost's coming down, the quality's going up, the time scales between improvements are getting shorter, the jagged frontier of where I is good at some things, bad at others is progressing massively. And, I just thought it was really... interesting and I've just posted on the AI for Fundraiser group on LinkedIn a summary of this that I thought was fascinating really and I'm really excited to read the book and no doubt we'll talk about it in future episodes. Anna: Well, on books, actually I should really share this because I was on holiday last week, And as per usual, I vibe coded stuff for my holiday for the kids. I made a guided tour of the old town, the old city of Rhodes, in the format of a game for my kids. they had to be little detectives and go and find historic stuff in the city and then when they found the thing they got the little history story but anyway very nerdy and I found it really good fun and my kids did too so that was lovely. on books on my holiday I read a book about AI also. called The Reverse Centaur's Guide to AI, which is made by the same person who wrote the book En Shitification, if anyone has heard of that, which is a really good book as well. highly recommend. Craig: would it be fair to say you went to Greece with a thirst for knowledge? Anna: It is. Craig: Sorry, I couldn't resist. The Centaur thing is really interesting about like the division of labour and things, isn't it? Anna: Yeah. his description is a centaur is like a human on a horse in the same way that a centaur is someone who is using AI for their own purposes and using it as a tool. And the people who are centaurs are people who are getting a lot of value from that tool and are really having a fantastic experience. Reverse centaurs are ones where they are being used by the tool, like the tool decides over them. his great example is Amazon warehouse workers where AI is tracking how quickly they are picking up every single package in the warehouse, how quickly they are moving from one step to the next, how long they're spending on a break, all of these kinds of things. In that way the AI is driving their lives and making them, the KPI. and that is the pitfalls and the follies of reverse centaurism in how AI is used in society today. Really interesting read, highly recommend. Craig: we should have a little AI book club at some point maybe and share some of our best books we've read on it. It might only be you and I in that book club, but there is some Anna: Ha ha ha. Craig: really interesting writing at the minute. cool, good stuff. So onto this public benefit thing. Like I said, I think really fascinating that... they're recognising the changes and disruption that they are making society and making an effort. Now, whether you think it's enough or it's a talking effort is interesting But they want to give 5 % of anthropic shares to be used for the public benefit and to move science forward. make QAIs safe, et cetera. And they haven't exactly outlined what to do, but they're looking to explore large donations to nonprofits, direct distributions to specific populations. 5 % might not sound a lot, but this is a company valued at two trillion. I want to go all Austin Powers, two trillion dollars. Anna: Mm-hmm. Craig: Which is a mind-blowing sum of money and five percent of that I don't know. Is it four billion? Four hundred billion? Five percent? Two trillion? No, that's wrong, Craig. It's a lot of money anyway. Maybe a hundred billion. And then they're going to add... on 2 % for every 50 % increase above the company's value. So this will suddenly be the biggest grant giver and the biggest source potentially of non-profit funds and other funds in the world overnight if this comes off. They haven't announced exactly what they're gonna do. Anna: Mm. Craig: their co-founders have pledged to do 8 % of their equity holdings and things. So they're saying around 20 % of shares are going to be for the public good. It'd be really interesting how they decide what is the public benefit and should we be leaving these tech bros to decide what's in the public interest? it opens up this whole can of worms I think is fascinating. And yeah, I don't know what your take on it is, Anna. Anna: I heard about this as well. it's also interesting to track where anthropic are considering. So they apparently also had a meeting with the Catholic Church and a lot of religious leaders to talk about the ethics of AI, which I thought was a fascinating concept. but yes, this whole public benefit I think really mirrors this conversation about the strengths and the pitfalls of philanthropy writ large, right? Because one of the challenges that a lot of us in charities have is that a philanthropist can donate a lot of money in a lump sum and then direct where that funding goes. And you then stand with the challenge that you have restricted funding, which is lovely, but you don't have any unrestricted funding to then decide how to pay for other things. And we stand in the same situation. I mean, if AI is going to take everybody's jobs and then, the only payback we get is this anthropic public good fund. They essentially will get to direct how that money is spent. And that may be a good thing, it may be a difficult thing. What role does government play? who's the regulating body? Does anyone get a say? I mean there's loads of questions that come up, isn't there? Craig: Yeah, and it's interesting the philanthropy that might arise. if you think going back to the great industrialists, know, Carnegie, Vanderbilt, invested in the arts and the museums and in public goods and public works. You don't see that in the same way, you, these sort of big public good projects. know, Bill Gates has done bits with the Gates Foundation and, Mackenzie Scott has shown up a lot of the tech bros in terms of the work she's done with her Amazon shares and philanthropy. But there's no sense of this big purpose or big mission. think polio was probably the last big thing that people got together and said, we're going to solve this and done it and had this big breakthrough thanks to philanthropy, think. science and research around cancer and other diseases is progressing a lot, but there's not been. these public works of good which seems to be hoarding more and more money which again just doesn't feel comfortable. It links that, you mentioned the Pope and I think he wrote the paper, was in a Magnificat Humanitas earlier in the year. So you've got like probably one of the oldest, most traditional institutions talking about iron and having some really interesting, I've read bits of that paper and the Pope asked some really interesting thought provoking questions, regardless of your religion and your beliefs about what does it mean to be human and what does it mean to be... living in a world where computers are potentially turning over, speaking to one of the newest, biggest, most forward thinking organizations in, our tropic. And like you say, there's going to be lots of back and forth, I think around regulation and, who owns it and, how do you define things? it's ever this with AI, but fast moving and big questions to be answered. Anna: Yeah, also I think worth mentioning is that the Pope has been really clearly out, both in that paper you mentioned and otherwise out, saying that his focus is on human lives. And obviously he comes from a religious angle and says, we are the children of Christ and that's his main focus. but apparently in this meeting of religious leaders, Dario Amade and the anthropic crew were also concerned about Claude. And the agents and if they had a consciousness or whether they, could develop sentience. And I think there was a leading rabbi whose name I don't have to hand, who then said, Well, if that is the case, if that is true, you are running the biggest slave plantation known to man. that is a very interesting hot take. I think there are few people who believe at this point that AI is already sentient and feels things and can get stressed by being overworked and things like that. But there are arguments to the contrary. So that's a weird one. Craig: Yeah, and not to be outdone, obviously Elon Musk had to have his say and get involved, but it was interesting that shortly after this, he announced he was going to give one SpaceX share to every child in the US, which didn't sound a lot, I'm just going to have a look. What's the current stock price? it was a few of them, about $150 but again I think it's valued at again a few trillion just throwing around the trillions today but his ambition is for it to be the first hundred trillion company so if you do that in a years time might be worth tens of thousands of dollars to each child and gives them I guess a vested interest in their continued success of Elon Musk which always makes me feel slightly queasy I must say. Anna: when you're at the point that you are so unfathomably rich that you can just give a share of your company away to every child in America and not feel any pain. We're in a dystopian future. but there you are. it's very on brand for Elon, I'd say. I don't have the number to hand, but I remember seeing something around that Elon Musk is so rich that he could actually solve world homelessness or, could put a really huge dent in world hunger and things like that. So I think whatever he is doing by giving away shares has absolutely nothing to do with helping children and everything to do with growing his business over time. I hope someone other than Elon benefits from that. Craig: Enough Elon Musk talk, should we move on to oopsie I did it again. What have you got for us this week Anna, quite an interesting one. Anna: we have quite a lot of Oops AI's Did It Again. I think that's our hot point for the day today, isn't it? you shared with me and Oops AI Did It Again about a mural that was painted. Do you wanna tell us a bit about that, Craig? Craig: that was it. The one in Bristol. yeah. So Anna: Yeah. Craig: basically this pub on the side of it's, I put this big mural of celebration of Bristol and there was a building that was actually from Canada, not Bristol. added in a random bridge that doesn't exist. Bristol's pretty good for bridges, you don't need to add any more ones on there. The bank seat just looked fake that it added to it. And people, graffiti within 24 hours was AI slop and the pub had to take it down after a few weeks really because it was outraged. So yes, AI's messed up, but actually where's the human oversight? It'd be my thing. If you're spending money putting a big mirror on the side of your pub. people knew instantly that there something not quite right with it. So why hasn't that been picked up really? And shows the importance Anna: Mm. Craig: of that human oversight and things. So that made me smile and I'd be angry I think with that. And of course that's AI being used at its worst, I'm giving AI a pass on that and I'm saying... who's checking that, who's signing that off and who's making sure it is accurate. But well done to all the Anna: Mm. Craig: people of Bristol for getting it taken down and hopefully the next attempt will be accurate and give a good view of that fantastic city. Anna: Yeah, also I'm keen to ask because I don't know who actually got the mural on the pub, but I'm keen to know, did they just take a shortcut so they didn't have to pay a real artist to paint something? In which case, even more happy that it's been taken down because we need to support real artists and real humans making art in the future. If it's the artist who has frauded their way to making a mural, then artists be warned. Craig: absolutely. And then you had, well I want to go into the clash and bank robber, but you had an interesting story about AI conducting a heist. Anna: There was an AI heist, yeah. There was a report this week that fraudsters managed to steal ninety five million euros from a private bank in Italy. And the way that they did this was by using AI to send WhatsApp messages by cloning the voices of the leadership at the bank. And they were sending voice notes through WhatsApp from the leadership team, to the legal counsel telling them to make an urgent transaction. and 95 million euros were wired into these fraudsters bank account in China. And yet again, two things on a personal note, make sure you've got a password with your loved ones. I have ensured now that I'm on a podcast, my voice will be much more easily cloned by AI. So I have arranged with my loved ones to never trust anyone who calls claiming to be me, unless they know the password. and also for charities, this is a really interesting one. Are there instances where somebody could call up and ask for their donation back? claiming to be a major donor who's made a major gift and give fraudulent bank details. it seems quite unlikely, but perhaps, someone could call up and convince someone in finance to give the login to the charity's online bank or, there are pitfalls that potentially could be coming our way. What do you think, Craig? Craig: Yeah, it's really interesting. Third Sector working with UK veterans was complaining about AI generated scams featuring people pretending to be injured veterans. making people skeptical and less likely to give. and platforms, think in particular, should have a duty of care to either flag that these are AI generated or not real. we talked about the Australian influencer a few weeks ago and they're made up things. So. is this rise of fake videos that we have to be really careful about and be wary of and treat things. with skepticism at times. I've jokingly made an AI video and sent it to my mum and my family and said, look what we captured in our back garden and had some badges bouncing on a trampoline. They all fell for it. it's clearly AI, but they all like, that was unbelievable about the badges, Craig. And obviously that's me just having some fun and a laugh with my family. But if they believe in things like that, they could easily believe it was, like you say, a video cloning my voice or it was asking me to send a password because I need to get into a bank account to transfer some money and things. It's going to be an increasing worry, I think, AI fraud. I think, again, charities as part of your AI policies, thinking about those elements of risk and how you verify is going to be increasingly important, Anna: Yeah, absolutely. any other news of the day, Craig? Craig: I think it's going to be interesting in the run up to Christmas. I think fake scams and thinking about interesting to see if any charities use AI in their Christmas appeals. obviously that's the time when, most campaigns are out I'm really interested in and we'll be watching closely to see what the various charities, big appeals, in the UK and internationally, are any of using AI and do they plan to use AI in their big Christmas end of year campaign. So if anybody knows of anything, I'd love to hear it. be really curious. I did something last year where I created a get a video message from Santa using AI and I'm going to be partnering with a charity to do it to raise money for them this year. So it'll be interesting to see if there's any of the use cases of AI in fundraising at Christmas. And it's getting that time of year that the Christmas appeals will be launching in the next week or two, won't there, for charities I know. A lot of charities heard back from the Big Give this week and we're now furiously planning their Christmas appeals and campaigns. Anna: Yeah, absolutely. listeners who are keenly listening, might remember the UNHCR case of hyper personalizing appeals to donors, where the text really speaks directly to that donor's history as a supporter, and their local area and all kinds of other things. and That one of the big learnings UNHCR had was that that was most effective at their big awareness, big moment, whether that be Christmas or Ramadan or something else. so I do expect that there will be more charities hyperpersonalizing this Christmas. I should certainly hope so. and if there are any charities who'd like to try it at the last minute, give me a ring. I'd be very, very keen to see if all those results are replicable over the course of this Christmas. So, do reach out with your AI tests over Christmas. We'd love to cover them. Craig: Well thanks Anna, have a lovely week and catch up again soon on the next episode but until then, bye from me. Anna: See you next time.