Science Friday - Medicinal Psychedelics Study, AI Art. December 9, 2022, Part 2

Episode Date: December 9, 2022

The Science Behind The Psychedelics Boom There’s been an explosion of new research into therapeutic uses for psychedelics. This includes drugs like psilocybin, the hallucinogenic chemical found in ���magic mushrooms,” and ketamine—which was originally used as an anesthetic, and later became a popular party drug also known as “special K.” Esketamine, a form of ketamine, was approved by the FDA in 2019 for use in treatment resistant depression. And just last month Colorado residents voted to legalize medicinal use of psilocybin. Following on the heels of Oregon’s legalization in 2020, which is now in the process of being implemented. A recent study published in the New England Journal of Medicine showed promising results in using psilocybin to help patients with treatment-resistant depression. About a third of those who received the highest dose were in remission 3 weeks later. This was the largest look at psilocybin’s effect on depression to date, involving 233 participants across ten countries in Europe. Ira talks with Dr. Steve Levine, senior vice president of patient access and medical affairs at COMPASS Pathways, the company that funded the study. Later, Ira takes a closer look into the latest psychedelic research and takes listener calls with Dr. Gerard Sanacora, professor of psychiatry and director of the Yale Depression Research Program at the Yale School of Medicine, and Dr. Alissa Bazinet, Clinical Psychologist, Co-Founder and Director of Research and Development at the Sequoia Center, and Associate Director of the Social Neuroscience and Psychotherapy Lab at Oregon Health and Science University.     How Will AI Image Generators Affect Artists? Back in August, controversy erupted around the winning submission of the Colorado State Fair’s art content. The winning painting wasn’t made by a human, but by an artificial intelligence app called Midjourney, which takes text prompts and turns them into striking imagery, with the help of a neural network and an enormous database of images. AI-based text-to-image generators have been around for years, but their outputs were rudimentary and rough. The State Fair work showed this technology had taken a giant leap forward in its sophistication. Realistic, near-instant image generation was suddenly here—and reactions were just as potent as their creations. Tech enthusiasts lauded the achievement, while artists were largely concerned and critical. If anyone could make a painting in just a few seconds, why would someone need to commission an artist to produce an illustration, or even bother spending years learning art at all? Read the rest on sciencefriday.com. Transcripts for each segment will be available the week after the show airs on sciencefriday.com.   Subscribe to this podcast. Plus, to stay updated on all things science, sign up for Science Friday's newsletters.

This is Science Friday. I'm Ira Flato. Later in the hour, we're going to dive into the world of AI-generated art and whether artists should fear that they'll be replaced by a computer. But first, there's been an explosion of new research into therapeutic uses for psychedelic drugs. Drugs like psilocybin, the hallucinogenic chemical found in magic mushrooms, and ketamine, which was originally used as an anesthetic, and then later became a popular party drug, also known as Special K. S-Katamine, a form of ketamine, was approved by the FDA in 2019

for use in treating resistant depression. And just last month, Colorado residents voted to legalize medicinal use of psilocybin, following on the heels of Oregon's legalization in 2020. A recent study published in the New England Journal of Medicine showed promising results in using psilocybin to help patients with treatment-resistant, depression. That's in patients whom other drugs have failed to help. About a third of those who received the highest dose were in remission three weeks later. This was the largest to date to look at psilocybin's effects on depression. 233 participants over 20, over 10 countries in Europe. To learn more

about the study, I recently spoke with Dr. Steve Levine, Senior Vice President of Patient Access and Medical Affairs at Compass Pathways. That's the company which funded this research. He joined me from Princeton, New Jersey. So let's talk about how big an effect did Silo-Sybin have on the participant's depression. It didn't work for everyone, right? That's right. What we saw in the study is that many participants did benefit from this treatment, but also some did not. And is that what you expect when you're using psychiatric drugs? It's what we expect in general with psychiatric drugs, it's especially what we expect in this population that we studied. And it's a really critical area to focus on when we interpret this data.

There is a big difference between a study that looks at major depressive disorder and one that looks at a treatment-resistant depression. Treatment-resistant depression, the regulatory definition, is a major depressive disorder that has not responded to at least two trials of an antidepressant. This wasn't the first episode of depression for most of these participants. In fact, most had had five, six, seven prior episodes of depression. So this is a group of people who, if you were to give them another antidepressant, we would have very low expectations that they would reach remission.

And in this case, despite this being just a single administration of Comp 360, along with psychological support, as you said, three weeks later, we still saw about 30% of patients in remission. So you're saying you saw immediate results with cyclocybin as opposed to other therapies that could take days, weeks, or months to show results. Exactly. Exactly. So if this treatment gets approved, it's something that patients would say take once a month or one to four times a year with talk therapy.

And that's how it would be administered. opposed to this longer course of treatment with other drugs? It remains to be seen, but that is the possibility. Within this study that was just published in the New England Journal, after a single administration, we followed patients out to 12 weeks, and then a subset of them continued into a longer-term extension where we followed them out up to 52 weeks. and we did see that most of the patients who were in remission still at that 12-week point,

the sustained responders, were able to continue that out to about six months, and then the numbers get pretty small because it's hard to get people to stay in a long-term study like that if there is an ongoing treatment. With our phase three program that we're starting now, we will have two pivotal six-week trials, and one of them will include a second administration. And then they will feed into a longer-term extension study as well that may include components of retreatment. And that data will tell us more about what to expect

about the effect of a second administration and the durability of either one or two administrations and what we should expect in terms of the frequency of administration. Should this be FDA approved and then, real-world practice. Now, taking any drug, of course, does not come without risks, and several patients in this study reported either suicidal ideation or self-harm behaviors. Does this indicate that psilocybin is responsible? That's an important question, a really important one. So, you know, fortunately, on the whole, this was a treatment that was very well tolerated.

90% more than 90% of any adverse events within this study were mild or moderate. The vast majority of them resolved as of the next day, and they were things like headache, nausea, fatigue, dizziness. So that was the vast majority of them. In all three arms of this study, so we had three arms. One was a one milligram group, one got 10 milligrams, one got 25 milligrams. Across all three groups, we did see some incidents of suicide. thinking or behavior, looking at those on a case-by-case basis, it does not look like that was

directly connected to comp 360 psilocybin itself. And to further explain that, I mean, first, I think it's important to understand the context of this. The population is one with treatment-resistant depression. A key symptom group within depression itself is suicidality, and two-thirds of participants of the 233 participants in the study did have some history of suicidal thinking or behavior. Of those who had this as an adverse event during the study, all had had that at some point in their history. Now, entering this study, everyone had low levels of suicidality

because that was one of the criteria for entering the study. So in a way, the only way to go was up. But if we actually look at item 10 on the madras, which is the suicide item on that scale, there was no mean change across these groups in suicidal. And also important to remember that in this study, patients were off of antidepressants. So two-thirds of patients in the study had come in on antidepressants and they were withdrawn from them prior to the study. A third were already off. So this is a moderate to severely depressed treatment-resistant population off of antidepressants.

in a group of non-responders weeks after treatment. And so we think that this is likely more a reflection of the course of the illness and to an extent that there are any differences across the three arms likely due to chance. Okay, so what's next for this research? Where do you go from here? Yes, so we've had our end of phase two discussions with FDA, and we are now beginning our phase three program. This is going to be a very large and robust program as well.

We'll enroll a total of about 950 participants across two pivotal trials. The first of them we call COP 005. This will compare our 25 milligram dose of COP 360 to a true placebo, along with psychological support. And this will help us to really establish a true safety baseline. and then we will look at another study called Comp006 that will again have three arms. So looking like our phase 2B study, looking again at one milligram versus 10 milligrams and 25 milligrams. And this will be the study that has two administrations. And this will help us to understand better whether a second administration may help a greater proportion of the participants achieve response or remission.

and also whether that second administration may have an impact on the durability of response. Dr. Levine, thank you for taking time to be with us today. Oh, a pleasure. Thank you so much for having me, Ira. Dr. Stephen Levine, Senior Vice President of Patient Access and Medical Affairs at Compass Pathways, the company which funded the research we were just discussing. Of course, this study raises many questions. What do we know about how psychedelics change the brain? What are the researchers still figuring out?

And what do you want to know about psychedelics? What questions do you have about how ketamine and psilocybin are being used to treat conditions like depression, PTSD, and addiction? You make the call, but only if you make the call. Our number is 844-724-8255-844-Sy-Talk. Have you undergone ketamine therapy? Have you tried microdosing? Tell us about your experiences. 844-724-825 or tweet us at SciFri.

Just a caveat, we cannot give you personal advice. That would be unethical, but we do want to hear about your experiences and get your questions. And joining me now to answer your questions and talk more about the latest and psychedelic research are my guests. Dr. Gerard Senecaora, professor of psychiatry and director of the Yale Depression Research Program at the Yale School of Medicine based in New Haven, Connecticut. Welcome to Science Friday. That's my pleasure to be here. Nice to have you.

And Dr. Alyssa Bazanette, clinical psychologist, co-founder and director of research and development at the Sequoia Center, Associate Director of the Social Neuroscience and Psychotherapy Lab at the Oregon Health and Science University and the Portland Veterans Hospital. She joins us from Portland, Oregon. Thank you for joining us, Alyssa. Thank you so much for having me. Let me ask you, Alyssa, you're a clinician who provides ketamine therapy and you're advising the legalization process of medicinal silocybin in the state of Oregon. What stands out to you from this silocybin study? Yeah, so first of all, I do believe that it's a solid result to have such a high percentage of folks that treatment-resistant depression see remission after three weeks.

weeks, or up until three weeks and beyond, is really quite amazing. The one thing that I am thinking about that does stand out to me just because I'm on the ground in Oregon where we are practically implementing this work in terms of legalizing psilocybin is the serious adverse event finding. The compass trial was really well controlled as all clinical trials are, so it's a little bit easier to control the environment and the variability. The folks that are accepted into the study do not necessarily they were screened out if they had suicidality that was clinically significant, and that will not necessarily be the case for folks that are seeking treatment in Oregon.

One thing that I want to note is that we, in Oregon, did not actually legalize medicinal use of psilocybin, so psilocybin services will exist outside of the health care system. All right. And the facilitators who become licensed will not necessarily be health care providers. I have to take time out to take a break, but we will come back and talk lots more with Dr. Seneca and Dr. Sennikora and Dr. Bazanet, our number, if you'd like to call in 844-724-8255. We're just getting started on this.

So please, we're happy to talk to you. Stay with us. We'll be right back after this break. Hi, Ira here. As a listener, I don't have to tell you that the need for Science Friday is stronger than ever. Science helps us navigate the world and make informed choices about our health, our environment, and our priorities. Science Friday is critical to public dialogue about science, and your donations are crucial to our success. Now is the time to head over to ScienceFriday.com slash support and make a gift.

Our 2023 programming depends on the generosity of our fans and our listeners. And remember, folks, any amount makes a difference. But the lasting gifts are the ones we can count on every year. So if it's in your power, consider making a sustaining donation. Once again, that link is ScienceFriday.com slash support. And thanks. We're continuing our conversation about the latest psychedelic research, taking your calls at 844-724-8255 with Dr. Gerard Sennikora and Dr. Alisa Bacenay.

Dr. Senecaora, could this silocybin study on top of the wave of silocybin research coming out paved the way for similar FDA approval, do you think, that we saw for ketamine? Well, I think it's the first step in that direction, but there's still quite a ways to go. This was a phase two trial. It's nice to see more rigorous, larger studies being done, but this is really the first study. And despite having a couple of hundred people, and it's still a relatively small study when you're looking at measures such as safety, and especially if you're looking at things that could happen in relatively small percent, but up to 5, 10 percent of people, if you're treating thousands of people or a lot of people,

you really have to have these larger studies with longer exposure times to really feel much more confident about the results, especially when it comes to safety. Pylocybin and ketamine compare in terms of efficacy. Do they affect the brain in similar ways? So that's a very difficult question to answer. I think there are two questions there. One is in terms of efficacy, one is sort of the mechanism. Maybe I'll address the mechanism question first. We don't really or completely understand how any of these drugs. In fact, you can say that for pretty much any drug. used in the treatment of psychiatric illnesses, how they work. Ketamine was pursued largely because of increasing evidence suggesting that the glutametergic neurotransmitic system, which is where ketamine primarily targets its effects at the NMD

or glutamate receptor, glutamate had been shown to be involved in the pathophysiology of depression in several animal models and even in some human investigations. and there was increasing evidence to suggest that glutamate plays a central role in modulating neuroplasticity, the way the brain adapts to changes. And it was thought this is really at the turn in the millennia that a drug that could target that system may have beneficial effects. And that was some of the studies that led to the current use of ketamine based very much on that idea that it was targeting glutamate. Since then, we've learned that it's probably much more complicated than that. We realize there are probably many other neurotransmitter

systems that are either approximately or more distally involved in the effect, but also other effects or what we do even consider non-specific effects of the treatment, such as hope, optimism, changes in expectation that probably all also play a large role. While the more classic psychedelics like psilocybin target primarily serotonin, there's probably a lot of overlap in the downstream mechanisms. Dr. Bazanette, many patients who have done ketamine therapy talk about a profound shift in their perception and experience of the world. And as a therapist, you help guide them through this experience.

Can you talk a little bit more about that process? Yes. So one thing that I like to say is ketamine and other psychedelics as well are really catalysts to a psychotherapeutic process. So it really is the combination of the medicine and the therapy. that produces the result. And so what I've seen with clients with ketamine, and I've worked with some who I'd had in private practice

regular talk therapy for years, go through a six-week protocol of ketamine, and suddenly have insights about things that happened in their past or a way that they were relating to their self or someone else or the world that was completely different than anything that they had thought about or experienced before. And so those realizations can be very profound

and very acute and can cause complete shifts in the way that they are interacting and interacting with themselves. So it's really quite miraculous to observe when it works. It doesn't work for everyone. Yeah. Yeah. Let's go to the phones to someone I think who says it's working for him. Mike and Sue Falls, South Dakota.

Hi, welcome to Science Friday. Hi, thanks for having me on. Hi, go ahead. I'm actually on the way to an appointment right now. It's actually a few hours drive for me from where I'm at, but I've done three or four appointments so far. I'm a comp that I have to fetch. Nope. Drive.

His car phone is you're breaking up. I'll just paraphrase what it has to say on my screen. He says he's on his way to his ketamine therapy appointment, and it has worked well for him so far. How many appointments does it take to know? know that it's working, Alyssa? It really varies on the individual. Sometimes you see

drastic change after the first session. I've had clients do one or two sessions and say they don't feel they need anymore because the issue that they were coming to therapy for is resolved. Others, it may take multiple sessions. We do an average protocol

at our clinic of six sessions, six weekly sessions. And most of the research suggests several sessions within a short period of time with results lasting up to four weeks or so, is what the research shows. There hasn't been a whole lot of research on the therapeutic component added to the ketamine. So a lot of folks are going to infusion clinics and just getting the ketamine administration without the psychotherapeutic component. Let's go to the phones. Let's go to Tiffany in Kentucky. Hi, Tiffany. Hi. I would like to share my experience.

I microdose with psilocybin. On my own, I suffered from depression. More than 20 years, I made dozens of attempts all my life. I exhibited self-harm behavior. And then I just, I microdose. The longest I did it was for eight days. I would grind up mushrooms and put them in a joke up and swallow them. And I realized after about day four or five that my natural base time will just smile.

I did figure out, though, that I go into it with intention. There is a lot of trauma in my background that I'm working on, you know, working through, but with therapy and periodic, microdosing, this is the best I have proven in my life. And at the height of my illness, I say, because I'm in recovery now, I actually work as a peer support specialist, which means I've done the work. You know, I put in the work. I got better. I'm not seeing I'm perfect. But I help people with their depressed.

At the height of my illness and my battle, I was taking 33 pills a day, 11 different psych meds. And that was too much. I was numb. I couldn't feel anything. Well, that's an amazing experience, Tiffany. Let me get a reaction. Thanks for calling. Dr. Sennak, Cora, let's talk about microdosing.

As Tiffany was saying, it's taking smaller doses that don't produce some of the psychoactive effects. It was working for Tiffany when these other drugs didn't work. Is that common? Well, I think it's very important for us to remember that these illnesses tend to have a course that waxes in Wayne. So there is some natural periods of response and remission. And as I mentioned before, this idea of nonspecific effects, this idea that this simple act of taking control of a treatment or feeling that you may be getting benefit from unique treatments is actually quite powerful. It is in fact extremely powerful.

Are you talking about a placebo effect here? I'm talking about the placebo effect, but it even goes beyond placebo. It's really the non-pharmacogicous effects, but placebo is clearly part of that, which is quite powerful. And not to diminish it in any way. In fact, that's something that I think we really need to understand much more broadly and to optimize our use of it. But we have to be very careful from making a causative relationship, you know, between these events, you know, where we see something changing alongside with something.

It, you know, we can't really say with great confidence that it's the drug. In fact, you'll see in many of these studies, the hard part isn't showing a benefit from the drug. It's showing a benefit greater than the placebo or very low-dose treatment. Very interesting. Let's go to Allison in Birmingham, Alabama. Hi, Allison. Hi, hi. Hi there. Go for it.

Long-term listener, first time caller. All right. So I have a friend who is suffering from postpartum depression. She had existing depression prior. And she is, when she called me and asked, I really didn't know. So I went to research it, and there really wasn't that much research on postpartum depression and the effects of ketamine on what is typically a pretty stubborn, frustrating time for a woman to get any kind of treatment at all. Nothing seems to work traditionally.

It's very, very hard. I think your expert would agree. But I guess what I want to know is that what is the new research as it pertains to postpartum depression, and then again, how does it affect? breast milk and breastfeeding, should it be stopped. Alison, thanks for calling. Dr. Bassinet, can you answer that? Yeah, I'm not aware of any specific research on ketamine for postpartum depression,

but that doesn't mean it doesn't exist. There are more and more studies on ketamine's effects for all sorts of indications coming out all the time. I haven't yet read that research specifically, but I don't think it's too far of a leap to suggest that it would likely be beneficial given ketamine's effects on treatment. resistant depression and suicidality and other mood disorders bipolar depression as well. And I'm not sure about the breast milk question either. Okay.

Let's go to Bob in upstate New York. Hi, Bob. Yes, hi. Hi there. Pleased to be on the show. Do you have a question? Well, I have kind of a question and a statement. I don't doubt that many people enjoy these medicines that are being discussed, and I've heard them

discussed at professional conferences. But my concern is that we have testimonials, even on your show right now. We have industry people talking about these things. And, you know, you have most psychotherapists haven't even learned validated, scientifically validated treatments for depression yet. And they're basically, you know, using old-fashioned psychodynamic therapies without adding things that have a rich and validated history. When you think about these medicines knocking people out of

refractory depression, there are other things as well. ECT knocks people out of refractory depression. It knocks people out of a mania, which nothing else will do. If drugs have failed, ECT is the only thing. But the long, long-term effects on people's characters, people's personalities, their family structures,

their early schema, you know, a lot of the things that lead to depression or trigger depression later on, this really hasn't been studied. And a lot of these things haven't been studied one against the other. The ends are rather small. And I'm not saying there's nothing to this, but I am saying that for 30 or 40 years, women were told you should take estrogen. It's great for you, you know. And all of a sudden there was meta-analysis. Bob, let me get a reaction.

Estrogen studies, and estrogen causes cancer. Yeah, Dr. Sanakura, valid points here that Bob's making? Yes, I think Bob is making several valid points here. There, you really do have to look at what the total bodily of data is and what the rigor that was used in collecting the data and the size of the data. The best sample size we have, at least for the ketamine-type drugs, is what is available for S-Kedamine delivered intranasally that has the FDA approval. That went through the large FDA approval process, and that has data for several hundred

people going out over several years. So we do have good data for that, and that's specifically for treatment-resistant depression or, to a lesser extent, depression associated with suicidal ideation. But the further we get away from that formulation used by that drug regimen, for that duration, we get further and further away from anywhere where we're standing on solid ground. Well, we will talk more about this. This is Science Friday from WNIC Studios. Dr. Bassinet, do you agree?

I do agree to some extent. Yes, I think Bob makes many, many valid points, and that over-generalization of research results is certainly problematic. I do think that that is what led to some of the policy changes in Oregon and Colorado because we have these research results and now we have legalized psilocybin services and there's much, much more to learn about these medicines. Let's go maybe one more call from Tom in Atlanta. Hi, Tom. Hey there.

Hey there, go ahead. Hey, so I actually have a question. I'm a provider. I've supported clients who are using ketamine, and I'm, curious, as the discipline moves forward, as we see psilocybin starting to move into the field, what's the training landscape expected to be like? How do you get trained for doing this is what you're asking? Yeah.

Right. It seems like they're different enough that maybe the training that we have for ketamine is not going to be a one-for-one likeness for psilocybin as it moves forward. Alyssa, you're a clinical psychologist. Can you elucidate on that? Yes. So there are several training programs that exist that train practitioners in ketamine, psilocybin, and MDMA, as those are the three that are most likely to be legally available in the near future. So you can get sort of a certificate in psychedelic studies.

I know CIS, California is one of those. Neuropa University, I believe Berkeley as well. So there are multiple universities that have these certificates. There are also independent training programs that exist online or a hybrid of online and in-person. And there are also ketamine-specific trainings. I really like the Polaris Insight Center's training. They have a really comprehensive ketamine training if you're looking to just work with ketamine. Dr. Bazinet, I know in the short time we have, you are currently doing a study with veterans to take MDMA and undergris.

group therapy to treat PTSD. What are you hoping to come out of that? Yes. So this study will be, it's a proof of concept, safety and feasibility study. So it is the first clinical trial to look at group psychotherapy for MDMA for PTSD at the VA, at the Portland VA. So we're hoping to assess if that is affected in that format. The thought is that having, you know, group connections, MDMA can really increase bonding and trust and with military veterans. A lot of group psychotherapy takes place at the VA anyhow. So we're interested to see if the MDMA protocol will adapt nicely to the group format. Well, we have run out of time so much to talk about. I'm sure this is something we will have to get back to. And I thank my guest,

and thank my listeners for calling in. Thank you, Dr. Gerard Sannachora, from Yale School of Medicine. Dr. Alyssa Bazinette, clinical psychologist and researcher at the Oregon Health and Science University at Portland Veterans Hospital. Thank you both. I wish we had more time. We have to take a break, as I say, and when we come back, we talk with artists and how they think AI-generated art will affect their livelihoods in our society. Stay with us. We'll be right back. This is Science Friday from WNYC Studios. This is Science Friday. I'm I Refleto. Back in August, there was a bit more controversy surrounding the Colorado State Art Fair contest than usual, and it was centered around the winning painting. It was painted not by a person who submitted it, but by an artificial intelligence app called Mid Journey, which turns a short amount of text into striking images,

thanks to machine learning and an enormous database of pictures. Many artists online were furious. saying that the piece shouldn't have been allowed to be entered. Others wondered if it meant the beginning of the end of their careers. There's a lot to talk about here. And joining me to delve deeply into the world of AI-generated art is sci-fi producer D. Peter Schmidt, ID. Hey, Ira. Yeah, have you seen other AI art from some of these platforms, Dali, Mid-Journey, Stable Diffusion?

Well, you know, no, I can't say that I have. Yeah, it's pretty wild, a little disturbing. it definitely provokes some mixed emotions. All you really do is you type in some words into a text field like Ira Flato in the style of Salvador Dali. It gives you a handful of images. Some of them are kind of messy. They're obviously fake-looking, but more often than not, it can actually give you some pretty realistic results. And also this guy who won the contest, Jason Allen, he gave this really punchy statement afterwards.

and he said, Art is dead, it's over AI one, humans lost. Oh, really? That's a pretty heavy statement. I mean, writing off all of humanity. Yeah, no big deal. And I basically wanted to know, is art really dead?

Is AI going to kill it? And the first person I talked to was Tina Tallinn, and she's an assistant professor of AI and the arts at the University of Florida. Do you remember the last thing you typed in to Dolly? To be honest, I think it was something related to apparel spritzes. Why? I was trying to create art to go above my bar cart,

and I couldn't find an apparel spritz poster that I liked on Etsy. When she's not putting together her dream bar cart, Tina teaches her students about AI systems, how to code with them and how to use them ethically. Many decisions that we're affected by are made by artificially intelligent agents, whether we know it or not. And so it's something that deserves a lot of conversation beyond its applicability to the arts. It's something that's becoming increasingly more common in society. A lot of these tools have really proliferated over the past like two months. So we're kind of on the bleeding edge in terms of seeing how it's all going to shake out and how these things are going to disrupt not only the creative industry, but also something.

society in general, when it becomes so much easier to generate really high quality images of literally anything. Is that a good thing? I don't know. That remains to be seen. If you had to abstract Dolly's process of what it's doing to something a human would actually do, how would you put that? So not unlike the processes by which humans learn how to create art, Dali is examining a ton of different artworks. So, for instance, in many fine arts programs, students learn by copying. works of art in the canon. And I think because again in AI we're trying to replicate human mental cognitive processes, these tools are trying to learn how to create art. To replicate the way a human's brain works, you need a couple things. First is a neural network. So similar to neural

pathways in the human brain, if we're trying to form a good habit, you want to strengthen the neural connections that result in that good habit. And if you want to break a bad habit, you want to disincentivize the neural connections. And so the same thing sort of happens in neural network. And so the same thing sort of happens in neural networks, where we're trying to figure out which of those we want to prioritize. And the second thing are data sets. Lots of data sets. With Dali, you have to train it on an enormous number of images in a variety of different

styles. However, one of these models can analyze more art than we will ever be able to in our lifetime. And so in terms of scale, in that respect, the learning process is very different. To make these models open AI and other. downloaded billions of images from the web. And because one of Dolly's main goals is to basically be an art generator, a lot of those images came from artists, dead and alive. As far as the legality of all this.

Right now, it's not a violation of copyright. It seems like you should say, yes, people obviously own their artwork. Legally, though, the way in which that artwork can be used is still up for the debate. So it looks like these AI-generated images fall under what's called fair use. which is a law that has protected works of people like Weird Al. If you make something using someone else's copyrighted material, but you transform it in a way that makes it unrecognizable or a different thing entirely, you are protected by fair use.

There's a lot of gray area here, but at the moment, the law favors AI image generators and not artists. Recent U.S. court cases have held that it's not a violation of copyright for your data to be used to train an AI model. That might change. Can you talk about your performance? personal feelings on the ethics of acquiring a data set. Like, obviously, data sets are not unethical to have, you know, fundamentally.

Right, right. But they can be ethically constructed, though. Some data sets are more ethically constructed than others. Data are a reflection of a society's values. And I'm a huge proponent of consent in all possible realms. But especially when it comes to data set creation. A lot of the popular image generation models, Dali, Mid Journey, Stable Diffusion, their developers did not get permission from artists to scrape their work to make these models,

models that ultimately may undermine those artists' ability to get work in the future. I think that they should have some say into whether or not they want their work incorporated into these data sets. It's not even difficult to even imagine an ethical AI. Janelle Jumel Jumelan is an illustrator and VFX artist. And it's this. It's just like, hey, just don't use. copyright data and the people want to, like, contribute to this data set, they can. And here's, like, the crazy part. Like, artists are weird. Like, we like weird things. We like new and interesting things. And I'd be willing to put my artwork in this data sets because I like weird things and I

like trying to see where things are. But the fact that, you know, these companies didn't ask any of us for our consent is really telling to kind of like the ethics of how they're producing it. And it's just, it's just frustrating. Even though Janelle is not a fan of how these companies establish themselves. They are actually pretty pumped about this basic idea, so much so that they've begun to integrate it into their illustration process. I thought it was cool and rad as heck. I was like, whoa, this is very wild, right? This is practically like an acid trip, honestly. But there's the ethics issue that comes with it after the initial wow, right? I was like, oh no, like, this could easily be used for like some very unethical practices and it has been.

And to avoid some of those ethical mindfields, Janelle's only been using the journey for the concept stage of their paintings. You know, when you look at clouds and you kind of can pull like shapes for things, like, oh, that's kind of like a tiger or like a dog or something like that. That's kind of how like my brain kind of works. What's exciting for me is like it's able to tap into things that's like I'm able to play with compositions that I would never come to the conclusion with and then be able to infuse my own aesthetics and ideas about where this composition should go towards. For me as an artist, it's more about trying to make it as much of my artwork as possible. Janelle's echoing an idea I saw on Twitter that maybe a more ethical use of the software is to use it as a jumping off point. Generate a bunch of quick visual concepts, get inspired, and then complete the illustration, just like a good old-fashioned human wood.

That way, the stolen art side of this stays out of the final product, and maybe the artist will have more time to make the illustration better. So I thought it would be interesting to commission Janelle to make. editorial art for this segment based on that process, which they're already familiar with. Usually when I contract artists for editorial art, I have a pretty good idea of what I want the final illustration to look like. But I wanted to give Janelle and Mid Journey a prompt and let them take it from there. So, yeah, the general prompt I want to give you is how you feel as a working artist about the expansion of AI on the art world. And I'm just going to leave it

like really vague like that. Very artsy, yes. So, yeah, how are you feeling about this project right now? I am excited. Like I said, I've just been very vocal on Twitter about trying to find, like, this middle ground solution, right? Because it's usually like the opposite corners where it's like AI should never we use and also AI is going to replace all artists. I understand the sentiments of a lot of artists being threatened about that. And a lot of my energy, at least with my own platform, is less about like trying to stop. or impede it and trying more about to protect the artist.

We need to find ways to kind of coexist in this weird world that we're living in right now. So they want to find a middle ground someplace, it sounds like. Yeah. So they're going to come up with a prompt. They're going to put it in Mid Journey, and Mid Journey is going to give them back a bunch of results. We'll decide on our favorites, and they'll paint it from there.

Sounds great. Yep. And I wanted to dig in more to this idea Janelle brought up that we should maybe start thinking of ways to live with this technology since it doesn't really seem like it's going anywhere. And some artists are also concerned about how generative AI will affect vulnerable communities. My name is Stephanie Dinkins. I am an artist who works with artificial intelligence and emerging technologies. I think my art is really a set of experiments, trying to figure out where communities of color fit into the technological future. I got in touch with

Stephanie because she spent almost the past decade integrating AI into her art and using it as a way to reflect on race and gender. She's made an AI chatbot called Not the Only One that you can have conversations with. To construct its database, she collected oral histories from three generations of her family where they reflected on the Black American experience. A lot of artists the past few months online have been very vocal about like, you know, I'm never going to touch this. I'm never going to use this in my work. But what do you say to people who, for their very own legitimate reasons, kind of refuse to engage with these systems? Okay. So in the realm of artists, I'm like, okay, that's your prerogative. Go ahead.

In the realm of people, I always think that we do not have the luxury and cannot afford not to engage this technology. I don't like this idea of having to play Kedeket. up once we realize how deeply embedded the technologies are in the world that we're living in, and how the ability to craft some of it, work with it, understand it gives you a power, or at least the avenue to power. And so for me, I much prefer that people engage and try to figure out than I'm staying. Even though the developers of these models are not transparent with how their algorithms work, Stephanie's been trying to get a peek behind the curtain anyway. Since 2016, she's been typing in different variations of the words black woman into these models

and seeing what kind of images they give back to her. And what they returned at first was not what she was expecting. And what came back at me was a kind of image of a white figure and a black cloak. Right? And you have to think, I'm like, what? Why am I getting this result? Like, how is this functioning? But those results have changed lately. I've been running a few Dolly 2 examples, and it's phenomenal what it can come. up with now, right? Like just the other day, I ran, what was it, a black woman laughing in a

field of rosers? And I got images of a black woman laughing in a field of roses and not only a black woman, but a dark skin black woman, which to me was also something interesting, right? And if you look at the opening eye webpage, they explain how they're trying to be more inclusive. They explain how they're trying to keep offensive information out of the data set, right? And then I was thinking a lot about that. It's like, well, what does that mean? And how do we go there? Like, is my job done? I don't need to think about this. Or is it that now I think we need to even be more vigilant in certain ways, right? Because the systems are going to be putting out things that feel really acceptable in certain ways. And then the question becomes about where

where real cultural attunedness comes into play. Despite instances like this where OpenAI is trying to appear more equitable, Tina Tallinn doesn't think we should leave it up to these powerful tools to make the right decisions. We need to have some sort of governing body that's able to audit these tools and examine what their impacts might be. Similar to how we have an FDA that looks at new medical interventions and what their impact on society may be, I think we need the same thing for algorithms. What we've seen time and time again is that the restrictions come too late. And we need to put the brakes on a lot of these things if we want to really make sure that we're forging the best future possible that involves both humans and AI.

Hey, Janelle. A week later, I got back on the phone with Janelle to check in on those first drafts, which were generated by Mid-Journey. Cool, cool, cool. Let me send some stuff to your way. I thought it'd be really interesting to do a very modern interpretation of the creation of Adam. Janelle wanted to keep out as much potentially copyrighted material as they could. So they typed in, quote, the creation of Adam and the style of Jean-Leon Jerome. Jerome was a 19th century artist whose work has inspired Janelle's art.

He's obviously long dead, which should put his art in the public domain. Okay, cool, I'm looking at this now. This is pretty wild. Hold on. You can check out what all this looks like at sciencefriiday.com slash AI art. Janelle sent me 36 thumbnails arranged on a six-by-six grid. They were all variations on the same basic idea. Ominous swirling clouds, disembodied, outstretched hands, mysterious, cloaked figures. They were like religious paintings from my nightmares.

God, the more I'm looking at this, the more I'm like freaking myself out. It's like I'm starting to see skulls in there and like twisted faces. That's like, see, that's like the fun part for me. It's like, what are these shapes? Is that a dragon? It's weird because like the idea of creating prompts from like, artists that had passed away, like, you're creating artwork that they would have never made in, like, a weird, almost like sci-fi, like necromancy kind of thing, right? Now you have

art that never existed from this artist, and it's almost like looking at like parallel timelines of things that never happen, right? Yeah, you're completely right. Maybe this is the most sci-fi thing about it to me so far, I think actually like directly interacting with it in this way, where you are kind of like peering into like alternate timelines that you can just create it well. And it's like, it's here right in front of me. Like I'm looking at it's, it's, it's, um, It's kind of surreal. Picasso making the Mona Lisa. That's a thing. You can literally write in the art thing. And it's a thing we can see now. And it's a, it's so wild. Like there's a timeline that that happened. And it's just. Yeah. Exactly. Wow, that's a really interesting idea having artists paint like other artists. Yeah. And my kind of big takeaway is that at the moment, at least it seems like these AI applications, they're kind of more of an assistant just to like help take care of some busy work. And then. We can focus on the creative stuff, but I think for visual artists like Janelle, threat is definitely bigger.

But anyway, we have a ton of stuff for listeners to check out on the website. We have Janelle's final human-painted illustration, all the works in progress of it. They even took screen recordings during the painting process. And we also have samples of Stephanie Dinkin's own AI-influenced artwork on that page two. All that's on our website at sciencefriiday.com slash AI art. Thank you, Ira.